ProgrammableMarketing/zalos.ai (opens in a new tab)
PDF

Views

  • Plan
  • Roadmap
  • Research

On this page

  • The brief
  • Start here
  • Idea portfolio
  • Effort vs impact
  • This week · 5
  • This quarter · 11
  • Big bets · 3
  • Wildcards · 3
  • How we can help
Generated Nov 10, 2026
  1. Programmable
  2. Playbooks
  3. Zalos
Zalos logo

growth playbook · Nov 10, 2026

zalos.ai (opens in a new tab)

Zalos growth playbook

Diagnosis

Zalos is a ten-person YC F25 company with a $3.6M seed, paying enterprise customers and a clear story: agents that work on top of the finance systems companies already have. In November 2026 it relaunched its site as "the finance agent factory" with four solution pages, and its engineering blog now publishes strong numbers (96.5% of harmful finance prompts caught, 94 of 100 browser tasks completed).

The gaps are in how buyers find, trust and contact it. The version of the site that AI assistants read still shows last year's pitch, and nine of its sixteen sitemap pages returned 404 to our AI reader. The contact page is a founder's inbox. There are no case studies or customer numbers, while rivals show named results.

This playbook fixes the AI-reader layer and the contact path first. It then turns the YC customer story, the benchmarks and the audit trail into proof, adds ERP pages, a paid process sprint and partner channels, and ends with big bets on self-serve, a workflow library and an annual finance-agent benchmark.

Growth Score

6/10

Biggest Opportunity

Zalos already has the raw material that finance buyers need: measured benchmarks, an audit trail on every run and a customer who got a day a week back. Packaged as pages, scorecards and a paid sprint, it can convert enterprise interest faster than a ten-person team could by selling one deal at a time.

Biggest Risk

Hyphenate, Ledge, Kognitos and UiPath sell to the same finance teams with named customers, hard numbers, free trials or much larger budgets. Without public proof, Zalos depends on founder-led selling and investor introductions, and that will not reach "12 million in 12 months".

01

Start here

The three moves to make first. Each one is written up in full below.

  1. 1

    Publish the first case study: from four days a month of transaction work to a day a week back

    Ask the NetSuite customer whether they can be named or prefer anonymity.

    High impact First result in 2 weeks
  2. 2

    Fix the version of zalos.ai that AI assistants read: nine sitemap pages return 404 and the Markdown homepage still shows last year's pitch

    Request /platform and /solutions/order-to-cash with an "Accept: text/markdown" header and confirm the 404.

    High impact First result in 1 week
  3. 3

    Replace "email the founder" with a two-minute form that routes by process and ERP

    Write the five form questions and add them before the calendar link.

    High impact First result in 2 weeks

02

Your idea portfolio

22 ideas, grouped by when to run them. Open any idea for how to run it: the steps, who to approach, copy you can use, and how to measure it.

  • 7 ideas are marked Unexpected: plays you are unlikely to have tried.

    1 of these are things Programmable can build and run for you.

Every idea placed by the effort it takes against the impact it could have. Quick wins sit top left; select a number to open that idea. The filters below narrow the list by status, timing, category and what Programmable can build.

Effort vs impactEach number is an idea. Select one to open it.
Impact →
Quick wins · 5Big bets · 9Fill-ins · 4Reconsider · 4
Effort →

Showing 22 of 22 ideas

This week 5

Quick wins with a clear first action.

  • 01

    Zalos built an AI-reader layer: an llms.txt and a Markdown homepage served to tools that ask for it. That layer has not kept up with the November relaunch. The Markdown homepage still leads with the old headline "Computer Agents for Finance Operations", not "The finance agent factory". llms.txt lists only the homepage and careers. When we fetched the site with an AI reading tool on November 9, 2026, nine of the sixteen sitemap URLs returned "404 Not Found": /platform, all four solution pages (order-to-cash, procure-to-pay, record-to-report, cost-to-serve), /about, /security and /newsroom. The homepage, blog, contact and careers pages loaded. A CFO who asks an AI assistant "what does Zalos do for procure-to-pay?" or "is Zalos secure?" gets an answer built from a page that is a year out of date, or from nothing.

    High impactLow effort First result in 1 week Founding Product Engineer

    First action

    Request /platform and /solutions/order-to-cash with an "Accept: text/markdown" header and confirm the 404.

    How to measure it

    Sitemap URLs answering Markdown requests with 200

    Now
    –
    Target
    16 of 16, with assistant answers naming the four solution areas
    By
    2 weeks

    Steps

    1. 1

      Reproduce

      Request all 16 sitemap URLs as HTML and as Markdown and log the status codes.

      Founding Product Engineer 1 hour

    2. 2

      Add the fallback

      Serve HTML (or generated Markdown) whenever no Markdown version exists, so no page returns 404.

      Founding Product Engineer 1 day

    3. 3

      Regenerate Markdown

      Build Markdown versions of the homepage, platform, four solutions, security and about from the live content, with the new headline.

      Founding Product Engineer 2 days

    4. 4

      Rewrite llms.txt

      List every page and post with a one-line summary, the supported ERPs and how to book a call.

      Co-Founder and CEO 2 hours

    5. 5

      Guard the build

      A deploy check that fails on any 404 or on mismatched titles between the HTML and Markdown versions.

      Founding Product Engineer half a day

    6. 6

      Test the answers

      Ask three AI assistants ten buyer questions (security, P2P, SAP, pricing) before and after, and keep the answers.

      Co-Founder and CEO 2 hours

    7. 7

      Repeat monthly

      Re-run the questions after each site change.

      Founding Product Engineer monthly

    First result
    1 week
    Cost
    Two to three days of engineering
    Owner
    Founding Product Engineer

    Smallest test

    Make /security and /solutions/order-to-cash answer Markdown requests today and ask an assistant about Zalos security tomorrow.

    Why it works

    Make every URL in the sitemap answer AI readers. Either serve a Markdown version of each page or fall back to the normal HTML when a Markdown version doesn't exist, so these pages never return 404. Regenerate the Markdown homepage from the live page so the headline, description and four solution areas match. Rewrite llms.txt to list every page with a one-line summary: platform, the four solutions, security, about, newsroom, blog posts and contact. Add a check to the deploy pipeline that requests every sitemap URL both as HTML and as Markdown and fails the build on any 404.

    Enterprise finance buyers increasingly shortlist vendors by asking AI assistants, and those assistants read what the site serves to them. Zalos sells agents that operate software for finance teams. Its own site should be the clearest example of a site that agents can read. The fix is a few hours of engineering on pages the team relaunched four days earlier.

    Ready to use

    ChecklistAI-reader site check

    [ ] Every sitemap URL returns 200 for Accept: text/html
    [ ] Every sitemap URL returns 200 for Accept: text/markdown (or falls back to HTML)
    [ ] Markdown homepage headline = "The finance agent factory"
    [ ] llms.txt lists platform, 4 solutions, security, about, newsroom, blog posts, contact
    [ ] llms.txt names supported ERPs and how to book a call
    [ ] Build fails on any 404
    [ ] Ten buyer questions re-asked monthly

    Risks and how to handle them

    • Two versions of every page drift apart again

      Generate the Markdown from the same source as the HTML and let the build check catch mismatches.

    • Assistants update their answers slowly

      Measure fetch-time answers first; indexed answers follow.

    Evidence

    • Your websiteSitemap (16 URLs, updated November 5, 2026): homepage, /platform, four solution pages (order-to-cash, procure-to-pay, record-to-report, cost-to-serve), /about, /security, /newsroom, /blog and two posts, /contact-us, /careers, /privacy-policy and /terms-conditions. Fetched with an AI reading tool on November 9, 2026 (/about twice), /platform, all four solution pages, /about, /security and /newsroom returned "404 Not Found"; the homepage, /blog, both posts, /contact-us, /careers, /privacy-policy, robots.txt and llms.txt loaded.Source
    • Your websiteAI-reader files (fetched November 9, 2026): llms.txt lists only the homepage and careers as key pages and offers "a Markdown version of the homepage (request with Accept: text/markdown)". That Markdown homepage still leads with the old headline "Zalos: Computer Agents for Finance Operations", describes Connect, Act and Audit stages, names Oracle, NetSuite, SAP and QuickBooks, and ends with the contact page and a founder's email. Neither mentions the finance agent factory, the four solution areas, the platform, security or the blog.Source
    • Your websiteHomepage (as served to browsers, checked November 9, 2026): title "Zalos - The finance agent factory"; description "Build AI workforces that run finance workflows across fragmented systems. Zalos computer agents log into your ERP, map your processes, then automate them and get better every month." The headline changed from "Computer Agents for Finance Operations" to "The finance agent factory." and the site moved from Webflow and HubSpot to Next.js. No pricing page, case studies, testimonials or live chat; a strip of 12 logos.Source
  • 02

    The contact page has no form. A buyer can email a founder or book a generic discovery call. Zalos sells on mapping each customer's workflows with forward deployed engineers, so the first call goes better when it already knows the process, the ERP and the volume. Today a founder learns all of that live on the call.

    High impactLow effort First result in 2 weeks Co Founder and CEO

    First action

    Write the five form questions and add them before the calendar link.

    How to measure it

    Call-to-pilot rate

    Now
    –
    Target
    +30%
    By
    60 days

    Steps

    1. 1

      Write the questions

      Process, systems, volume, size, optional recording.

      Co-Founder and CEO 2 hours

    2. 2

      Build the form

      On /contact-us, storing answers in the CRM.

      Founding Product Engineer 1 day

    3. 3

      Route

      Enterprise ERPs to the CEO, midmarket to a forward deployed engineer.

      Founding Product Engineer half a day

    4. 4

      Prepare demos

      One recorded demo per process to open each call.

      Forward deployed engineer 1 week

    5. 5

      Confirm by email

      An automatic reply that repeats the answers and the agenda.

      Founding Product Engineer 2 hours

    6. 6

      Review weekly

      Demand by process and ERP, and call-to-pilot rate.

      Co-Founder and CEO weekly

    First result
    2 weeks
    Cost
    Two days of engineering
    Owner
    Co Founder and CEO

    Smallest test

    Add three required questions to the calendar booking today.

    Why it works

    Add a short form above the calendar: process (order-to-cash, procure-to-pay, record-to-report, other), systems (ERP, banks, portals), monthly volume, company size and an optional screen recording or process note. Route enterprise ERPs (SAP, Oracle) to the CEO, and route midmarket NetSuite and QuickBooks teams to a forward deployed engineer with a prepared demo of the same process. Keep the email for people who prefer it.

    Qualifying questions cut wasted calls, and buyers prefer a first meeting that shows their own process. The answers also record which processes and ERPs drive demand, which the solution pages and roadmap need.

    Risks and how to handle them

    • Enterprise buyers dislike forms

      Keep it to five fields and leave the direct email visible.

    Evidence

    • Your websiteContact page: no form; two options, email a founder directly or book a "discovery call" through a linked calendar. llms.txt adds that Zalos offers no self-serve public API or autonomous purchasing.Source
    • News coverageHg Orbit podcast with the CEO: Zalos was "already deployed with major enterprise customers" five weeks after launch, including "companies with ... billions in revenue"; one customer is "likely to get rid of a point solution SaaS"; delivery uses forward deployed engineering to map each customer's workflows; the plan is to start with the biggest companies and become "increasingly self-serve to come down the market"; the goal is "12 million in 12 months".Source
    • News coverageSeed round (tech.eu, March 24, 2026): $3.6 million led by 14Peaks Capital with Cohen Circle and 20VC, plus angels including the CTO of business bank Tide and the CFO of FedEx; the money is to "improve its tech and target new customers", moving from midmarket ERPs to enterprise ERPs; Zalos wants to "sit on top of what is already there".Source
  • 03

    The blog's two posts are Zalos's strongest public proof: 96.5% of harmful finance prompts caught against 56.9% for a general filter, and 94 of 100 browser tasks completed. The harness post closes by previewing future evaluations, with no demo or contact link. Zalos already runs a "bring us a finance process" challenge, but readers of its best content never see it.

    Medium impactLow effort First result in 2 weeks Co Founder and CTO

    First action

    Write the three-line closing block and add it to the Safety Gate post.

    How to measure it

    Blog-sourced demo requests

    Now
    –
    Target
    5 a month
    By
    60 days

    Steps

    1. 1

      Write the block

      Result, finance use and the bring-a-process offer.

      Co-Founder and CEO 1 hour

    2. 2

      Build a component

      A reusable post footer with tagged links.

      Founding Product Engineer 2 hours

    3. 3

      Add to both posts

      Safety Gate and browser harness.

      Founding Product Engineer 1 hour

    4. 4

      Banner on /blog

      One line above the post list.

      Founding Product Engineer 1 hour

    5. 5

      Share the posts

      Post each on the founders' LinkedIn with the offer.

      Co-Founder and CTO 1 hour

    6. 6

      Track

      Leads by post, monthly.

      Co-Founder and CEO monthly

    First result
    2 weeks
    Cost
    $0
    Owner
    Co Founder and CTO

    Smallest test

    Add the footer to the Safety Gate post today.

    Why it works

    Add a closing block to both posts and to every future post: one sentence tying the result to finance work ("this is the harness that runs your reconciliations"), the bring-a-process offer and a link to the new contact form. Add a slim banner on /blog. Tag the links so the form shows which post a lead read.

    Technical readers who finish a benchmark post are evaluators, often the engineers or finance-systems owners a CFO will ask. A specific next step turns readership into pipeline at no cost.

    Risks and how to handle them

    • Engineering readers find sales copy off-putting

      Keep it to one factual line and a link.

    Builds on

    Evidence

    • Your websiteBlog, "A better browser agent, without a new browser" (October 27, 2026): Zalos completed 94 of 100 tasks on the BU-Bench-V1 browser benchmark and showed a desktop demo entering a split expense in GnuCash on a remote Windows desktop; the post previews future evaluations and ends without a demo or contact call to action. The blog index lists two posts.Source
    • Your websiteBlog, "How Zalos catches 96.5% of harmful finance prompts" (November 3, 2026): the Safety Gate caught 747 of 774 harmful finance prompts (96.5%) against 56.9% for a general-purpose moderation filter, with no false alarms on 234 benign finance prompts; it covers insider trading, market manipulation, accounting fraud, tax evasion, AML and sanctions evasion and prompt injection, and separately flags sensitive data such as payroll records and bank details. No customers named.Source
    • News coverageFondo (October 30, 2026): Zalos launched AI workers for order-to-cash, procure-to-pay and record-to-report that log into browser billing tools and desktop applications, including on-premise ERPs, run in a customer's virtual private cloud or on an air-gapped device, log decisions and learn from resolved exceptions. Zalos did not disclose customer numbers, "measured exception rates, process-completion rates" or before-and-after costs. Offer: bring a finance process to automate, or get AirPods or a charity donation.Source
  • 04

    zalos.ai has no case studies or testimonials, and press coverage notes that Zalos has published no customer numbers. Yet its YC launch already tells a complete story: a Bay Area finance manager spent four days a month downloading, uploading and categorizing transactions, and Zalos now runs weekly in NetSuite, categorizes more than 60% of outstanding transactions and gives back a day a week. Rivals lead with named results.

    High impactLow effort First result in 2 weeks Co Founder and CEO

    First action

    Ask the NetSuite customer whether they can be named or prefer anonymity.

    How to measure it

    Published case studies with numbers

    Now
    –
    Target
    3
    By
    90 days

    Steps

    1. 1

      Ask permission

      Named or anonymous, with a sign-off step.

      Co-Founder and CEO 2 days

    2. 2

      Pull the numbers

      Hours before, run frequency, share categorized, exceptions.

      Forward deployed engineer 1 day

    3. 3

      Write it

      Before, how it runs, after, one quote.

      Co-Founder and CEO 1 day

    4. 4

      Publish

      A /customers page and a homepage strip with the headline numbers.

      Founding Product Engineer 1 day

    5. 5

      Repeat for enterprise

      Two anonymous enterprise stories by industry and revenue band.

      Co-Founder and CEO 6 weeks

    6. 6

      Use in sales

      Attach to every proposal and follow-up.

      Co-Founder and CEO ongoing

    First result
    2 weeks
    Cost
    Founder time
    Owner
    Co Founder and CEO

    Smallest test

    Add the three YC-launch numbers to the homepage this week.

    Why it works

    Turn that story into a one-page case study, anonymized if the customer prefers: the before (systems, hours, errors), the agent's weekly run, the after (60%+ categorized, a day a week back) and a short audit-trail screenshot. Ask enterprise customers for the same in anonymous form ("a $2B-revenue logistics company"). Put the numbers on the homepage under the logo strip and in every proposal.

    Finance buyers want a peer who did it first. A specific before-and-after beats any claim about agents, and an anonymous case study is easy for a customer to approve.

    Risks and how to handle them

    • Enterprise customers refuse publicity

      Offer anonymity and describe them by industry and size.

    Evidence

    • News coverageYC launch post: problems are "Legacy ERPs with limited AI", "Authentication-protected websites & portals" and "Too many Finance Systems that don't speak to one another"; a Bay Area finance manager spent 4 days a month downloading, uploading and categorizing transactions, and Zalos now runs weekly and "operates NetSuite to categorize >60% outstanding transactions", giving back a day a week. The ask: "Send our website / YouTube to 5 friends in Midmarket / Enterprise Finance & Operation teams."Source
    • Your websiteHomepage (as served to browsers, checked November 9, 2026): title "Zalos - The finance agent factory"; description "Build AI workforces that run finance workflows across fragmented systems. Zalos computer agents log into your ERP, map your processes, then automate them and get better every month." The headline changed from "Computer Agents for Finance Operations" to "The finance agent factory." and the site moved from Webflow and HubSpot to Next.js. No pricing page, case studies, testimonials or live chat; a strip of 12 logos.Source
    • Competitor researchRivals' own homepages: Hyphenate (maximor.ai now redirects there), "The autonomous finance platform for the office of the CFO", shows named customer results such as a 60% faster month-end close (3 weeks to 5 days) and 98% straight-through cash processing, and a $45M Series A; Ledge, "AI accountants that prepare your close", shows nine customer logos, SOC 2 and ISO 42001 badges and "Go live in one day".Source
  • 05

    Enterprise finance deals stall on controls: who approved this action, what stops an agent from moving money wrongly, and where the evidence is. Zalos already has the answers spread across pages. Every step is screen-recorded and logged, human review is optional, a finance-specific gate catches 96.5% of harmful prompts with no false alarms on benign ones, sensitive data such as payroll and bank details is flagged, and it can run in a private cloud. There is no trust center, and the /security page did not load for AI readers.

    High impactLow effort First result in 1 month Co Founder and CTO

    First action

    List the twenty questions the last three enterprise security reviews asked.

    How to measure it

    Security review cycle time

    Now
    –
    Target
    -30%
    By
    90 days

    Steps

    1. 1

      Collect questions

      From recent security reviews and questionnaires.

      Co-Founder and CTO 1 day

    2. 2

      Draft the brief

      Approvals, audit trail, Safety Gate, data, deployment, SOC 2.

      Co-Founder and CTO 3 days

    3. 3

      Add evidence

      A sample audit log and recording frame, with synthetic data.

      Senior AI Engineer 1 day

    4. 4

      Review

      With a friendly controller or auditor from the investor network.

      Co-Founder and CEO 1 week

    5. 5

      Publish

      Summary on /security, full PDF after the first call, documents in a trust center.

      Founding Product Engineer 2 days

    6. 6

      Update quarterly

      New benchmark results and certifications.

      Co-Founder and CTO quarterly

    First result
    1 month
    Cost
    Founder time; trust-center tool optional
    Owner
    Co Founder and CTO

    Smallest test

    Send a two-page draft to the next prospect in security review.

    Why it works

    Write a four-page controls brief for the CFO, controller and internal audit. It covers segregation of duties and approvals, the audit trail (recording, logs, retention), the Safety Gate policy and its measured results, data handling and deployment options (VPC, air-gapped) and SOC 2 status. Publish a short version on /security and give the full PDF after the first call. Add a lightweight trust center for documents under NDA.

    Controls questions come up in every enterprise finance deal. Answering them before they are asked shortens security review and gives the CFO's champion a document they can forward internally.

    Risks and how to handle them

    • Over-claiming on controls

      State only what is tested, with dates and test sizes.

    Builds on

    Evidence

    • Your websiteBlog, "How Zalos catches 96.5% of harmful finance prompts" (November 3, 2026): the Safety Gate caught 747 of 774 harmful finance prompts (96.5%) against 56.9% for a general-purpose moderation filter, with no false alarms on 234 benign finance prompts; it covers insider trading, market manipulation, accounting fraud, tax evasion, AML and sanctions evasion and prompt injection, and separately flags sensitive data such as payroll records and bank details. No customers named.Source
    • Your websiteAI-reader files (fetched November 9, 2026): llms.txt lists only the homepage and careers as key pages and offers "a Markdown version of the homepage (request with Accept: text/markdown)". That Markdown homepage still leads with the old headline "Zalos: Computer Agents for Finance Operations", describes Connect, Act and Audit stages, names Oracle, NetSuite, SAP and QuickBooks, and ends with the contact page and a founder's email. Neither mentions the finance agent factory, the four solution areas, the platform, security or the blog.Source
    • Your websiteBrand footprint (research, November 9, 2026): domain registered April 2025; no LinkedIn, X, YouTube or other social profile links found on the site; no newsletter; the site mentions SOC 2 but has no trust center.Source
    • News coverageFondo (October 30, 2026): Zalos launched AI workers for order-to-cash, procure-to-pay and record-to-report that log into browser billing tools and desktop applications, including on-premise ERPs, run in a customer's virtual private cloud or on an air-gapped device, log decisions and learn from resolved exceptions. Zalos did not disclose customer numbers, "measured exception rates, process-completion rates" or before-and-after costs. Offer: bring a finance process to automate, or get AirPods or a charity donation.Source

This quarter 11

The core plays for the next three months.

  • 06

    Press coverage of the October launch pointed out exactly what Zalos has not disclosed: "measured exception rates, process-completion rates" and before-and-after costs. The Senior AI Engineer role is hired to build evals, replay testing and regression detection, so the data will exist inside the company. Zalos already publishes benchmark results on its blog, and the scorecard would apply the same habit to production.

    High impactMedium effort First result in 6 weeks Senior AI Engineer

    First action

    Define "completed", "exception" and "hours returned" in one page.

    How to measure it

    Deals where the scorecard was opened

    Now
    –
    Target
    50% of active deals
    By
    1 quarter

    Steps

    1. 1

      Define metrics

      Completion, exception, resolution time, hours returned.

      Senior AI Engineer 2 days

    2. 2

      Instrument runs

      Log the fields on every production run.

      Senior AI Engineer 2 weeks

    3. 3

      Get consent

      Customer contracts allow anonymized aggregates.

      Co-Founder and CEO 2 weeks

    4. 4

      Build the page

      Monthly numbers with definitions and breakdowns.

      Founding Product Engineer 1 week

    5. 5

      Launch

      Blog post and founder posts with the first month.

      Co-Founder and CTO 1 day

    6. 6

      Update monthly

      Automated where possible, reviewed by the CTO.

      Senior AI Engineer monthly

    First result
    6 weeks
    Cost
    Engineering time
    Owner
    Senior AI Engineer

    Smallest test

    Share one month of numbers privately with three prospects.

    Why it works

    Aggregate anonymized production runs into a public scorecard page updated monthly: runs completed, share completed without a human, exceptions per 100 runs and how fast they were resolved, hours returned, and breakdowns by process (O2C, P2P, R2R) and system type (browser, desktop, legacy). Explain the definitions. Link it from the homepage, controls brief and proposals.

    Reliability is the main doubt buyers have about agents in finance. A number that updates every month answers that doubt better than any demo, and none of the rivals reviewed publish one.

    Risks and how to handle them

    • A bad month is public

      Show trends and explain dips; honesty builds trust with finance buyers.

    • Too few runs to be meaningful

      Publish counts alongside percentages.

    Evidence

    • News coverageFondo (October 30, 2026): Zalos launched AI workers for order-to-cash, procure-to-pay and record-to-report that log into browser billing tools and desktop applications, including on-premise ERPs, run in a customer's virtual private cloud or on an air-gapped device, log decisions and learn from resolved exceptions. Zalos did not disclose customer numbers, "measured exception rates, process-completion rates" or before-and-after costs. Offer: bring a finance process to automate, or get AirPods or a charity donation.Source
    • HiringSenior AI Engineer (London, £70k-£140k, 0.20-1.50% equity, open on Y Combinator's job board): "We have paying enterprise customers, strong seed funding, and a small team that moves fast"; builds evals, replay testing and regression detection, computer-use agents for legacy interfaces such as SAP GUI and thick clients, and cross-system workflows across SAP, Salesforce and ServiceNow; estimates 30-40% of enterprise processes lack reliable APIs; works directly with founders and customers.Source
    • Your websiteBlog, "How Zalos catches 96.5% of harmful finance prompts" (November 3, 2026): the Safety Gate caught 747 of 774 harmful finance prompts (96.5%) against 56.9% for a general-purpose moderation filter, with no false alarms on 234 benign finance prompts; it covers insider trading, market manipulation, accounting fraud, tax evasion, AML and sanctions evasion and prompt injection, and separately flags sensitive data such as payroll records and bank details. No customers named.Source
  • 07

    Zalos names Oracle, NetSuite, SAP and QuickBooks and is moving from midmarket to enterprise ERPs. Its engineers are building agents for SAP GUI and thick clients and estimate that 30-40% of enterprise processes lack reliable APIs. The sixteen-page sitemap has no page for any ERP, so a controller searching "automate SAP GUI reconciliation" or "NetSuite bank feed categorization without API" finds nothing from Zalos.

    High impactMedium effort First result in 2 months Founding Product Engineer

    First action

    List the ten most common tasks per system from customer deployments.

    How to measure it

    Demo requests from system pages

    Now
    –
    Target
    8 a month
    By
    1 quarter

    Steps

    1. 1

      List tasks

      Top ten per system from deployments.

      Forward deployed engineer 2 days

    2. 2

      Record demos

      One short video per system with synthetic data.

      Senior AI Engineer 1 week

    3. 3

      Write pages

      Tasks, video, controls, proof.

      Co-Founder and CEO 1 week

    4. 4

      Build

      /systems/netsuite, /sap, /oracle, /quickbooks with Markdown versions.

      Founding Product Engineer 1 week

    5. 5

      Link

      From solution pages, llms.txt and the sitemap.

      Founding Product Engineer 1 day

    6. 6

      Measure

      Visits, assistant answers and demo requests per page.

      Founding Product Engineer monthly

    First result
    2 months
    Cost
    About three weeks of team time
    Owner
    Founding Product Engineer

    Smallest test

    Publish the NetSuite page first, built from the YC-launch story.

    Why it works

    Build four system pages, each listing the specific tasks Zalos runs there (for example, NetSuite: download card and bank transactions, categorize, reconcile; SAP GUI: post journal entries and run reports with no API), a short video of the agent working in that system with synthetic data, the controls involved and the matching case study. Link them from the solution pages and the Markdown layer.

    Finance buyers search by the system they are stuck with. System pages capture that intent and show Zalos's real advantage, which is working where APIs don't exist.

    Risks and how to handle them

    • Trademark use of ERP names

      Use names descriptively and add a non-affiliation note.

    Builds on

    Evidence

    • Your websiteAI-reader files (fetched November 9, 2026): llms.txt lists only the homepage and careers as key pages and offers "a Markdown version of the homepage (request with Accept: text/markdown)". That Markdown homepage still leads with the old headline "Zalos: Computer Agents for Finance Operations", describes Connect, Act and Audit stages, names Oracle, NetSuite, SAP and QuickBooks, and ends with the contact page and a founder's email. Neither mentions the finance agent factory, the four solution areas, the platform, security or the blog.Source
    • HiringSenior AI Engineer (London, £70k-£140k, 0.20-1.50% equity, open on Y Combinator's job board): "We have paying enterprise customers, strong seed funding, and a small team that moves fast"; builds evals, replay testing and regression detection, computer-use agents for legacy interfaces such as SAP GUI and thick clients, and cross-system workflows across SAP, Salesforce and ServiceNow; estimates 30-40% of enterprise processes lack reliable APIs; works directly with founders and customers.Source
    • News coverageSeed round (tech.eu, March 24, 2026): $3.6 million led by 14Peaks Capital with Cohen Circle and 20VC, plus angels including the CTO of business bank Tide and the CFO of FedEx; the money is to "improve its tech and target new customers", moving from midmarket ERPs to enterprise ERPs; Zalos wants to "sit on top of what is already there".Source
    • Your websiteSitemap (16 URLs, updated November 5, 2026): homepage, /platform, four solution pages (order-to-cash, procure-to-pay, record-to-report, cost-to-serve), /about, /security, /newsroom, /blog and two posts, /contact-us, /careers, /privacy-policy and /terms-conditions. Fetched with an AI reading tool on November 9, 2026 (/about twice), /platform, all four solution pages, /about, /security and /newsroom returned "404 Not Found"; the homepage, /blog, both posts, /contact-us, /careers, /privacy-policy, robots.txt and llms.txt loaded.Source
  • 08

    Zalos's launch offer is playful: bring a finance process, or get AirPods or a charity donation. That gets attention but sets the price of a pilot at zero. Delivery runs on forward deployed engineers mapping each workflow, so every free pilot uses the scarcest people on a ten-person team, and the CEO's goal is "12 million in 12 months".

    High impactMedium effort First result in 1 month Co Founder and CEO

    First action

    Write the one-page sprint offer with scope, fee, timeline and exit criteria.

    How to measure it

    Sprint-to-contract conversion

    Now
    –
    Target
    60%
    By
    2 quarters

    Steps

    1. 1

      Define scope

      One process, systems in scope, success criteria.

      Co-Founder and CEO 2 days

    2. 2

      Set the fee

      A fixed fee credited against year one.

      Co-Founder and CEO 1 day

    3. 3

      Standardize delivery

      A day-by-day sprint plan and report template.

      Forward deployed engineer 1 week

    4. 4

      Build the report

      Runs, completion, exceptions, hours, go/no-go.

      Senior AI Engineer 1 week

    5. 5

      Offer it

      To every qualified call from the new form.

      Co-Founder and CEO ongoing

    6. 6

      Review

      Conversion and delivery time per sprint.

      Co-Founder and CEO monthly

    First result
    1 month
    Cost
    Founder and engineer time
    Owner
    Co Founder and CEO

    Smallest test

    Offer the paid sprint to the next three qualified prospects.

    Why it works

    Package a fixed-scope sprint: one process, two weeks, a fixed fee credited against the annual contract. It includes process mapping from recordings and files, an agent running in a sandbox or VPC, a measured result (runs, completion, exceptions, hours) and a go/no-go report for the CFO. Keep the AirPods challenge as the top-of-funnel hook that leads into the sprint.

    A paid sprint filters for serious buyers, funds forward deployed time, and produces the numbers the CFO needs to sign. A credit against the contract removes most of the objection to paying.

    Risks and how to handle them

    • Enterprise procurement slows a small payment

      Keep the fee under common card or sign-off limits.

    Builds on

    Evidence

    • News coverageFondo (October 30, 2026): Zalos launched AI workers for order-to-cash, procure-to-pay and record-to-report that log into browser billing tools and desktop applications, including on-premise ERPs, run in a customer's virtual private cloud or on an air-gapped device, log decisions and learn from resolved exceptions. Zalos did not disclose customer numbers, "measured exception rates, process-completion rates" or before-and-after costs. Offer: bring a finance process to automate, or get AirPods or a charity donation.Source
    • News coverageHg Orbit podcast with the CEO: Zalos was "already deployed with major enterprise customers" five weeks after launch, including "companies with ... billions in revenue"; one customer is "likely to get rid of a point solution SaaS"; delivery uses forward deployed engineering to map each customer's workflows; the plan is to start with the biggest companies and become "increasingly self-serve to come down the market"; the goal is "12 million in 12 months".Source
    • HiringFounding Product Engineer (London, £70k-£130k, 0.20-1.00% equity): will "work directly with users, iterating in real time" and "help shape the product, culture, and company from the ground up"; team size 10; stack Next.js, React, TypeScript, Python, FastAPI and Postgres.Source
  • 09

    Zalos's pitch is to "sit on top of what is already there", and the CEO expects most CFOs to keep their ERPs, which leaves "a lot of automation to be built on top". The firms that implement and support those ERPs hear these automation requests every week and often have no answer short of custom code. A ten-person team cannot meet every finance team itself.

    High impactMedium effort First result in 1 quarter Co Founder and CEO

    First action

    Ask each current customer which firm implements or supports their ERP.

    How to measure it

    Partner-sourced qualified opportunities

    Now
    –
    Target
    10
    By
    2 quarters

    Steps

    1. 1

      Map partners

      Firms behind current customers' ERPs.

      Co-Founder and CEO 1 week

    2. 2

      Design terms

      Referral fee or margin, joint sprint model.

      Co-Founder and CEO 1 week

    3. 3

      Sign the first two

      Pilot agreements.

      Co-Founder and CEO 1 month

    4. 4

      Train consultants

      What Zalos does and how to spot a fit.

      Forward deployed engineer 2 days each

    5. 5

      Run joint sprints

      One per partner.

      Forward deployed engineer 2 weeks each

    6. 6

      Publish

      A partner page and joint case study.

      Founding Product Engineer 1 week

    First result
    1 quarter
    Cost
    Referral fees from closed revenue
    Owner
    Co Founder and CEO

    Smallest test

    Run one joint sprint with one partner.

    Why it works

    Recruit three to five mid-size implementation or managed-accounting partners (NetSuite and SAP first). Offer referral fees or a reseller margin, joint process sprints, training for their consultants and co-branded case studies. Start with partners already working with Zalos customers.

    Implementation partners have trust and access inside the finance teams Zalos wants. Agents that avoid custom integration help partners deliver more without new code.

    Risks and how to handle them

    • Partners see agents as a threat to services revenue

      Position Zalos as capacity that lets them take on more clients.

    Builds on

    Evidence

    • News coverageSeed round (tech.eu, March 24, 2026): $3.6 million led by 14Peaks Capital with Cohen Circle and 20VC, plus angels including the CTO of business bank Tide and the CFO of FedEx; the money is to "improve its tech and target new customers", moving from midmarket ERPs to enterprise ERPs; Zalos wants to "sit on top of what is already there".Source
    • News coverageHg Orbit podcast with the CEO: Zalos was "already deployed with major enterprise customers" five weeks after launch, including "companies with ... billions in revenue"; one customer is "likely to get rid of a point solution SaaS"; delivery uses forward deployed engineering to map each customer's workflows; the plan is to start with the biggest companies and become "increasingly self-serve to come down the market"; the goal is "12 million in 12 months".Source
    • Competitor researchKognitos ("Deterministic Agents, Hallucination-Free AI") names PepsiCo ($1M+ annual cost reduction) and DISH (30,000 hours a year) and offers "Start for Free"; UiPath's finance page positions agents and RPA as an execution layer above ERPs, with Canon USA saving 6,000 hours a year and Thermo Fisher cutting invoice processing time 70%.Source
  • 10

    Zalos's YC launch asked readers to send its site to five friends in midmarket and enterprise finance teams. Its seed round brought in a finance network: 14Peaks Capital, Cohen Circle, 20VC, and angels who are the CFO of FedEx and the CTO of a business bank. That network has never been given a structured ask, a target list or anything to forward.

    High impactLow effort First result in 1 month Co Founder and CEO

    First action

    Send each investor and angel a list of ten target companies and ask which they know.

    How to measure it

    Investor-sourced first calls

    Now
    –
    Target
    15
    By
    1 quarter

    Steps

    1. 1

      Build the target list

      By ERP, size and region.

      Co-Founder and CEO 2 days

    2. 2

      Write the forwardable

      One page plus case study.

      Co-Founder and CEO 1 day

    3. 3

      Make the ask

      Three intros per investor and angel.

      Co-Founder and CEO 1 week

    4. 4

      Plan the roundtable

      Twelve finance leaders, one investor host.

      Co-Founder and CEO 1 month

    5. 5

      Follow up

      Within 24 hours, with a sprint offer.

      Co-Founder and CEO ongoing

    6. 6

      Report back

      Monthly update to investors on intros and outcomes.

      Co-Founder and CEO monthly

    First result
    1 month
    Cost
    Event costs only
    Owner
    Co Founder and CEO

    Smallest test

    Ask one angel for three introductions this week.

    Build the prospect list. The list compiled from public sources, verified, deduplicated and delivered to your CRM. Usually live within days.

    Book a working session

    Why it works

    Build a one-page forwardable brief (problem, one case study, the controls brief and the paid sprint). Ask each investor and angel for three specific CFO or controller introductions from their portfolio and network, using a shared target list by ERP and size. Host a quarterly closed-door CFO roundtable on "automating the processes your ERP can't" with an investor as host.

    Warm introductions from a known CFO or investor are the fastest route into enterprise finance, and investors want to help but need a specific ask.

    Risks and how to handle them

    • Investor fatigue

      Keep asks small and report outcomes.

    Builds on

    Evidence

    • News coverageYC launch post: problems are "Legacy ERPs with limited AI", "Authentication-protected websites & portals" and "Too many Finance Systems that don't speak to one another"; a Bay Area finance manager spent 4 days a month downloading, uploading and categorizing transactions, and Zalos now runs weekly and "operates NetSuite to categorize >60% outstanding transactions", giving back a day a week. The ask: "Send our website / YouTube to 5 friends in Midmarket / Enterprise Finance & Operation teams."Source
    • News coverageSeed round (tech.eu, March 24, 2026): $3.6 million led by 14Peaks Capital with Cohen Circle and 20VC, plus angels including the CTO of business bank Tide and the CFO of FedEx; the money is to "improve its tech and target new customers", moving from midmarket ERPs to enterprise ERPs; Zalos wants to "sit on top of what is already there".Source
  • 11

    Finance buyers compare Zalos with three alternatives. RPA vendors such as UiPath now pitch agents as an execution layer above the ERP. Close and finance platforms such as Ledge and Hyphenate show named results. The third alternative is building it in-house, which the CEO says buyers increasingly do ("buy and build"). The sitemap has no comparison pages, so buyers and AI assistants make these comparisons without Zalos's input.

    Medium impactLow effort First result in 1 month Co Founder and CEO

    First action

    Ask the last five prospects what else they evaluated.

    How to measure it

    Comparison-page demos

    Now
    –
    Target
    5 a month
    By
    1 quarter

    Steps

    1. 1

      Ask prospects

      What else they looked at and why.

      Co-Founder and CEO 1 week

    2. 2

      Check facts

      Rivals' current claims from their own sites.

      Co-Founder and CEO 2 days

    3. 3

      Write three pages

      Vs RPA, vs platforms, vs in-house.

      Co-Founder and CEO 1 week

    4. 4

      Legal review

      Claims and trademarks.

      Legal counsel 3 days

    5. 5

      Publish

      With Markdown versions and llms.txt entries.

      Founding Product Engineer 1 day

    6. 6

      Refresh

      Quarterly.

      Co-Founder and CEO quarterly

    First result
    1 month
    Cost
    Founder time and a legal review
    Owner
    Co Founder and CEO

    Smallest test

    Publish the "agents vs RPA" page first.

    Why it works

    Write three category comparisons rather than attacks on one vendor: computer agents vs RPA bots (maintenance when screens change, 2FA, exceptions), agents on top of your ERP vs adopting a new finance platform, and Zalos vs building with general-purpose agents (finance safety, audit trail, evals). Use fair tables, date them, link the benchmarks and give each one a Markdown version.

    Comparison searches come from buyers late in the process. Category framing lets Zalos define the choice on its strengths: no migration, no API needed and an audit trail by default.

    Risks and how to handle them

    • Stale or unfair claims

      Date each page and cite rivals' own pages.

    Builds on

    Evidence

    • Competitor researchKognitos ("Deterministic Agents, Hallucination-Free AI") names PepsiCo ($1M+ annual cost reduction) and DISH (30,000 hours a year) and offers "Start for Free"; UiPath's finance page positions agents and RPA as an execution layer above ERPs, with Canon USA saving 6,000 hours a year and Thermo Fisher cutting invoice processing time 70%.Source
    • Competitor researchRivals' own homepages: Hyphenate (maximor.ai now redirects there), "The autonomous finance platform for the office of the CFO", shows named customer results such as a 60% faster month-end close (3 weeks to 5 days) and 98% straight-through cash processing, and a $45M Series A; Ledge, "AI accountants that prepare your close", shows nine customer logos, SOC 2 and ISO 42001 badges and "Go live in one day".Source
    • News coverageHg Orbit podcast with the CEO: Zalos was "already deployed with major enterprise customers" five weeks after launch, including "companies with ... billions in revenue"; one customer is "likely to get rid of a point solution SaaS"; delivery uses forward deployed engineering to map each customer's workflows; the plan is to start with the biggest companies and become "increasingly self-serve to come down the market"; the goal is "12 million in 12 months".Source
    • Your websiteSitemap (16 URLs, updated November 5, 2026): homepage, /platform, four solution pages (order-to-cash, procure-to-pay, record-to-report, cost-to-serve), /about, /security, /newsroom, /blog and two posts, /contact-us, /careers, /privacy-policy and /terms-conditions. Fetched with an AI reading tool on November 9, 2026 (/about twice), /platform, all four solution pages, /about, /security and /newsroom returned "404 Not Found"; the homepage, /blog, both posts, /contact-us, /careers, /privacy-policy, robots.txt and llms.txt loaded.Source
  • 12

    The new homepage promises agents that "get better every month", and press describes agents that learn from resolved exceptions. The customer who buys has no regular evidence that this is happening. The CEO says stickiness comes from modelling each customer's workflows in depth. A finance leader's renewal decision rests on hours saved, exceptions avoided and audit readiness, so those should be what each customer sees.

    High impactMedium effort First result in 6 weeks Forward deployed engineer

    First action

    Agree a baseline of hours per process with each current customer.

    How to measure it

    Customers adding a second process

    Now
    –
    Target
    50% within 6 months
    By
    2 quarters

    Steps

    1. 1

      Set baselines

      Hours per process before Zalos, agreed with each customer.

      Forward deployed engineer 1 week

    2. 2

      Design the report

      One page, six numbers, one recommendation.

      Founding Product Engineer 3 days

    3. 3

      Automate it

      From run logs.

      Senior AI Engineer 2 weeks

    4. 4

      Send

      To the CFO and process owner monthly.

      Forward deployed engineer monthly

    5. 5

      Quarterly review

      Results and the next process to automate.

      Co-Founder and CEO quarterly

    6. 6

      Feed the scorecard

      Anonymized totals to the public page.

      Senior AI Engineer monthly

    First result
    6 weeks
    Cost
    Engineering time
    Owner
    Forward deployed engineer

    Smallest test

    Hand-build one report for the largest customer this month.

    Why it works

    Generate a one-page monthly report per customer: runs, hours returned (using the customer's own baseline), exceptions and how they were resolved, what the agent learned this month, processes ready to automate next and a link to the audit trail. Send it to the CFO and the process owner, and review it quarterly in a business review that proposes the next process.

    A monthly record of value builds the renewal case continuously and turns each report into an expansion conversation. It also produces the data behind the public scorecard and case studies.

    Risks and how to handle them

    • Disputed baselines

      Agree them in writing at kickoff.

    Evidence

    • Your websiteHomepage (as served to browsers, checked November 9, 2026): title "Zalos - The finance agent factory"; description "Build AI workforces that run finance workflows across fragmented systems. Zalos computer agents log into your ERP, map your processes, then automate them and get better every month." The headline changed from "Computer Agents for Finance Operations" to "The finance agent factory." and the site moved from Webflow and HubSpot to Next.js. No pricing page, case studies, testimonials or live chat; a strip of 12 logos.Source
    • News coverageHg Orbit podcast with the CEO: Zalos was "already deployed with major enterprise customers" five weeks after launch, including "companies with ... billions in revenue"; one customer is "likely to get rid of a point solution SaaS"; delivery uses forward deployed engineering to map each customer's workflows; the plan is to start with the biggest companies and become "increasingly self-serve to come down the market"; the goal is "12 million in 12 months".Source
    • News coverageFondo (October 30, 2026): Zalos launched AI workers for order-to-cash, procure-to-pay and record-to-report that log into browser billing tools and desktop applications, including on-premise ERPs, run in a customer's virtual private cloud or on an air-gapped device, log decisions and learn from resolved exceptions. Zalos did not disclose customer numbers, "measured exception rates, process-completion rates" or before-and-after costs. Offer: bring a finance process to automate, or get AirPods or a charity donation.Source
  • 13

    The Senior AI Engineer will build replay testing and regression detection, and the role says "it worked in the demo" is not good enough. Finance teams fear silent changes: an agent update that categorizes differently mid-quarter. Change control is something controllers already require from every system that touches the ledger.

    Medium impactMedium effort First result in 2 months Senior AI Engineer

    First action

    Pick one customer process and replay last month against the current version.

    How to measure it

    Updates with replay evidence

    Now
    –
    Target
    100%
    By
    2 quarters

    Steps

    1. 1

      Build replay

      Re-run recorded inputs on a new version in a sandbox.

      Senior AI Engineer 3 weeks

    2. 2

      Diff results

      Flag runs whose outcome would change.

      Senior AI Engineer 1 week

    3. 3

      Write the note

      A template for what changed and why.

      Founding Product Engineer 2 days

    4. 4

      Add approval

      Approve or defer for the process owner.

      Founding Product Engineer 1 week

    5. 5

      Store as evidence

      In the audit trail.

      Senior AI Engineer 2 days

    6. 6

      Pilot

      With two customers, then roll out.

      Forward deployed engineer 1 month

    First result
    2 months
    Cost
    Engineering time already in the hiring plan
    Owner
    Senior AI Engineer

    Smallest test

    Send one hand-made change note with the next update.

    Why it works

    Before each agent or policy update, replay the customer's last month of runs against the new version and produce a short change note: what changed, how many past runs would now behave differently and why, and an approve or defer option for the process owner. Keep each note in the audit trail as change-management evidence.

    It turns an internal engineering habit into a control that finance and audit can rely on. It lowers the risk of churn after a bad update and is hard for less disciplined rivals to copy.

    Risks and how to handle them

    • Approvals slow down fixes

      Allow urgent fixes with a note sent afterwards.

    Evidence

    • HiringSenior AI Engineer (London, £70k-£140k, 0.20-1.50% equity, open on Y Combinator's job board): "We have paying enterprise customers, strong seed funding, and a small team that moves fast"; builds evals, replay testing and regression detection, computer-use agents for legacy interfaces such as SAP GUI and thick clients, and cross-system workflows across SAP, Salesforce and ServiceNow; estimates 30-40% of enterprise processes lack reliable APIs; works directly with founders and customers.Source
    • Your websiteBlog, "A better browser agent, without a new browser" (October 27, 2026): Zalos completed 94 of 100 tasks on the BU-Bench-V1 browser benchmark and showed a desktop demo entering a split expense in GnuCash on a remote Windows desktop; the post previews future evaluations and ends without a demo or contact call to action. The blog index lists two posts.Source
  • 14

    Zalos's blog is written for engineers, and its best market insight sits in a job post: 30-40% of enterprise processes lack reliable APIs. The site links to no LinkedIn, X or YouTube profile and has no newsletter, although the YC launch asked people to share "our website / YouTube". Finance leaders mostly discover vendors through peers and LinkedIn.

    Medium impactMedium effort First result in 1 month Co Founder and CEO

    First action

    Add the company LinkedIn and YouTube links to the footer.

    How to measure it

    Finance-role subscribers

    Now
    –
    Target
    500
    By
    2 quarters

    Steps

    1. 1

      Add links

      LinkedIn, YouTube and newsletter in the footer.

      Founding Product Engineer 1 hour

    2. 2

      List episodes

      Twelve no-API tasks from deployments.

      Forward deployed engineer 2 days

    3. 3

      Record clips

      Synthetic data, 60 seconds each.

      Senior AI Engineer 2 hours each

    4. 4

      Write posts

      Today's cost, the agent, the controls.

      Co-Founder and CEO fortnightly

    5. 5

      Start the newsletter

      Collect the series with a sign-up form.

      Founding Product Engineer 1 day

    6. 6

      Measure

      Followers, subscribers and inbound from finance roles.

      Co-Founder and CEO monthly

    First result
    1 month
    Cost
    Founder time and a newsletter tool
    Owner
    Co Founder and CEO

    Smallest test

    Post one clip of a 2FA portal download on LinkedIn this week.

    Why it works

    Publish a fortnightly series for controllers and finance-ops leads. Each episode takes one no-API task (portal downloads with 2FA, SAP GUI postings, payslip reconciliation), shows how teams do it today, what it costs in hours and a 60-second agent clip with synthetic data. Post it on the founders' LinkedIn, collect it into a newsletter, and add social and newsletter links to the site footer.

    Specific, recognizable pain draws the right audience, and the clips double as sales material. A newsletter builds an owned list for a company that has none.

    Risks and how to handle them

    • Founder time competes with selling

      Batch-record clips once a month.

    Evidence

    • HiringSenior AI Engineer (London, £70k-£140k, 0.20-1.50% equity, open on Y Combinator's job board): "We have paying enterprise customers, strong seed funding, and a small team that moves fast"; builds evals, replay testing and regression detection, computer-use agents for legacy interfaces such as SAP GUI and thick clients, and cross-system workflows across SAP, Salesforce and ServiceNow; estimates 30-40% of enterprise processes lack reliable APIs; works directly with founders and customers.Source
    • Your websiteBrand footprint (research, November 9, 2026): domain registered April 2025; no LinkedIn, X, YouTube or other social profile links found on the site; no newsletter; the site mentions SOC 2 but has no trust center.Source
    • News coverageYC launch post: problems are "Legacy ERPs with limited AI", "Authentication-protected websites & portals" and "Too many Finance Systems that don't speak to one another"; a Bay Area finance manager spent 4 days a month downloading, uploading and categorizing transactions, and Zalos now runs weekly and "operates NetSuite to categorize >60% outstanding transactions", giving back a day a week. The ask: "Send our website / YouTube to 5 friends in Midmarket / Enterprise Finance & Operation teams."Source
  • 15

    Zalos builds agents from screen recordings, process instructions and historical files, and its forward deployed engineers already map workflows by hand. A finance manager can record a painful process in five minutes. The analysis Zalos does after a sale could be the thing that starts one.

    High impactMedium effort First result in 1 month Forward deployed engineer

    First action

    Write the plan template from two existing customer mappings.

    How to measure it

    Assessment-to-sprint rate

    Now
    –
    Target
    25%
    By
    1 quarter

    Steps

    1. 1

      Template the plan

      Agent steps, approvals, hours, controls, quote.

      Forward deployed engineer 2 days

    2. 2

      Build the upload page

      Secure upload with redaction guidance.

      Founding Product Engineer 1 week

    3. 3

      Draft with tooling

      Use the internal mapping pipeline to draft each plan.

      Senior AI Engineer 2 weeks

    4. 4

      Review

      An engineer checks every plan before it's sent.

      Forward deployed engineer 1 hour each

    5. 5

      Promote

      Blog footers, the content series and investor intros.

      Co-Founder and CEO ongoing

    6. 6

      Follow up

      Offer the sprint within 48 hours.

      Co-Founder and CEO ongoing

    First result
    1 month
    Cost
    Engineering time; about an hour per assessment
    Owner
    Forward deployed engineer

    Smallest test

    Offer the assessment by email to ten prospects.

    Why it works

    Offer a page: upload a screen recording (with synthetic or redacted data) or describe the process, name the systems and monthly volume, and receive within 48 hours a short plan covering steps an agent can run, steps needing human approval, expected hours returned, controls and a sprint quote. Use the internal mapping tools to draft it, with an engineer reviewing.

    It gives the buyer something concrete and personal before any call, qualifies demand by process and system, and leads into the paid sprint. It also shows what the agent factory does.

    Risks and how to handle them

    • Sensitive data in uploads

      Ask for synthetic or redacted recordings, delete after 30 days and cover it in the controls brief.

    • Too many low-fit requests

      Require company size and systems.

    Builds on

    Evidence

    • Your websiteHomepage (as served to browsers, checked November 9, 2026): title "Zalos - The finance agent factory"; description "Build AI workforces that run finance workflows across fragmented systems. Zalos computer agents log into your ERP, map your processes, then automate them and get better every month." The headline changed from "Computer Agents for Finance Operations" to "The finance agent factory." and the site moved from Webflow and HubSpot to Next.js. No pricing page, case studies, testimonials or live chat; a strip of 12 logos.Source
    • News coverageHg Orbit podcast with the CEO: Zalos was "already deployed with major enterprise customers" five weeks after launch, including "companies with ... billions in revenue"; one customer is "likely to get rid of a point solution SaaS"; delivery uses forward deployed engineering to map each customer's workflows; the plan is to start with the biggest companies and become "increasingly self-serve to come down the market"; the goal is "12 million in 12 months".Source
    • News coverageFondo (October 30, 2026): Zalos launched AI workers for order-to-cash, procure-to-pay and record-to-report that log into browser billing tools and desktop applications, including on-premise ERPs, run in a customer's virtual private cloud or on an air-gapped device, log decisions and learn from resolved exceptions. Zalos did not disclose customer numbers, "measured exception rates, process-completion rates" or before-and-after costs. Offer: bring a finance process to automate, or get AirPods or a charity donation.Source
  • 16

    zalos.ai has no pricing page, and llms.txt states there is no self-serve purchasing. Finance leaders are the people who have to fit a new line into a budget, and a CFO comparing Zalos with Kognitos (which offers "Start for Free") or with in-house builds has no anchor. The CEO wants to move down-market over time, and that will need public pricing.

    Medium impactLow effort First result in 1 month Co Founder and CEO

    First action

    Write the pricing model in three sentences and a typical range.

    How to measure it

    Pricing-page demo requests

    Now
    –
    Target
    4 a month
    By
    90 days

    Steps

    1. 1

      Decide the model

      Per process or agent, plus the sprint.

      Co-Founder and CEO 2 days

    2. 2

      Set ranges

      Midmarket and enterprise typical totals.

      Co-Founder and CEO 1 day

    3. 3

      Build the calculator

      Hours times cost against price.

      Founding Product Engineer 3 days

    4. 4

      Publish

      /pricing with a Markdown version and llms.txt entry.

      Founding Product Engineer 1 day

    5. 5

      Prefill the form

      Calculator inputs flow into the contact form.

      Founding Product Engineer 1 day

    6. 6

      Review

      Win rate and deal size before and after.

      Co-Founder and CEO quarterly

    First result
    1 month
    Cost
    A week of engineering
    Owner
    Co Founder and CEO

    Smallest test

    Add a "How pricing works" section to the contact page.

    Why it works

    Add a /pricing page that explains the model rather than a full rate card. Pricing is per process (or per agent) per year, plus the paid sprint, with a typical range for midmarket and enterprise. Add a simple calculator: hours a month on the process times the loaded cost, compared with the annual price. Put the controls and deployment options beside it.

    Price transparency removes a reason to delay the first call and filters out buyers who don't fit. The calculator makes the business case in the buyer's own numbers.

    Risks and how to handle them

    • Enterprise deals anchor low

      Show ranges and what drives them (systems, volume, deployment).

    Builds on

    Evidence

    • Your websiteContact page: no form; two options, email a founder directly or book a "discovery call" through a linked calendar. llms.txt adds that Zalos offers no self-serve public API or autonomous purchasing.Source
    • Your websiteHomepage (as served to browsers, checked November 9, 2026): title "Zalos - The finance agent factory"; description "Build AI workforces that run finance workflows across fragmented systems. Zalos computer agents log into your ERP, map your processes, then automate them and get better every month." The headline changed from "Computer Agents for Finance Operations" to "The finance agent factory." and the site moved from Webflow and HubSpot to Next.js. No pricing page, case studies, testimonials or live chat; a strip of 12 logos.Source
    • Competitor researchKognitos ("Deterministic Agents, Hallucination-Free AI") names PepsiCo ($1M+ annual cost reduction) and DISH (30,000 hours a year) and offers "Start for Free"; UiPath's finance page positions agents and RPA as an execution layer above ERPs, with Canon USA saving 6,000 hours a year and Thermo Fisher cutting invoice processing time 70%.Source

Big bets 3

Larger plays that can change the trajectory.

  • 17

    The CEO plans to become "increasingly self-serve to come down the market". Zalos's first public success was midmarket: a finance manager using NetSuite who got a day a week back. Rivals such as Kognitos already offer "Start for Free". Forward deployed engineering cannot serve thousands of smaller teams, but a handful of standard workflows on NetSuite and QuickBooks could run with little hand-holding.

    High impactHigh effort First result in 2 quarters Founding Product Engineer

    First action

    Measure how much engineer time the last three NetSuite deployments needed, step by step.

    How to measure it

    Self-serve customers

    Now
    –
    Target
    50
    By
    3 quarters

    Steps

    1. 1

      Pick workflows

      The three most repeated midmarket tasks.

      Forward deployed engineer 1 week

    2. 2

      Standardize

      Templates that need only credentials and one recorded run.

      Senior AI Engineer 6 weeks

    3. 3

      Build onboarding

      Credential vault, recording, first-five-runs approval.

      Founding Product Engineer 6 weeks

    4. 4

      Add billing

      Monthly card billing.

      Founding Product Engineer 2 weeks

    5. 5

      Private beta

      Twenty teams from the waitlist and partners.

      Co-Founder and CEO 1 month

    6. 6

      Launch

      Public launch with case studies.

      Co-Founder and CEO 1 month

    First result
    2 quarters
    Cost
    Roughly a quarter of two engineers
    Owner
    Founding Product Engineer

    Smallest test

    Onboard five midmarket teams on one workflow with a written guide and no engineer calls.

    Why it works

    Package three standard workflows (card and bank transaction download and categorization, bank reconciliation, vendor-portal invoice download) for NetSuite and QuickBooks. A team connects credentials in a secure vault, records one run, approves the agent's first five runs and then lets it run weekly. Price it monthly with a card. Route anything non-standard to the sales path.

    Midmarket finance teams feel the same pain with smaller budgets and faster decisions. A self-serve tier builds volume, data and word of mouth while enterprise deals take months.

    Risks and how to handle them

    • Support load from many small teams

      Limit to three workflows and two ERPs at launch.

    • Distraction from enterprise deals

      Gate on the enterprise pipeline being healthy.

    Builds on

    Evidence

    • News coverageHg Orbit podcast with the CEO: Zalos was "already deployed with major enterprise customers" five weeks after launch, including "companies with ... billions in revenue"; one customer is "likely to get rid of a point solution SaaS"; delivery uses forward deployed engineering to map each customer's workflows; the plan is to start with the biggest companies and become "increasingly self-serve to come down the market"; the goal is "12 million in 12 months".Source
    • News coverageYC launch post: problems are "Legacy ERPs with limited AI", "Authentication-protected websites & portals" and "Too many Finance Systems that don't speak to one another"; a Bay Area finance manager spent 4 days a month downloading, uploading and categorizing transactions, and Zalos now runs weekly and "operates NetSuite to categorize >60% outstanding transactions", giving back a day a week. The ask: "Send our website / YouTube to 5 friends in Midmarket / Enterprise Finance & Operation teams."Source
    • Competitor researchKognitos ("Deterministic Agents, Hallucination-Free AI") names PepsiCo ($1M+ annual cost reduction) and DISH (30,000 hours a year) and offers "Start for Free"; UiPath's finance page positions agents and RPA as an execution layer above ERPs, with Canon USA saving 6,000 hours a year and Thermo Fisher cutting invoice processing time 70%.Source
  • 18

    Zalos now calls itself "the finance agent factory", but delivery is still one forward-deployed mapping per customer. Every NetSuite categorization or SAP GUI posting it builds is a workflow the next customer on the same system also needs. A reusable library is what would let a ten-person team reach "12 million in 12 months".

    High impactHigh effort First result in 1 quarter Co Founder and CTO

    First action

    Tag every live workflow by process and system and find the overlaps.

    How to measure it

    Time to first live run

    Now
    –
    Target
    -50%
    By
    3 quarters

    Steps

    1. 1

      Inventory

      All live workflows by process and system.

      Forward deployed engineer 1 week

    2. 2

      Define a template

      Steps, approvals, exceptions, evals, version.

      Co-Founder and CTO 2 weeks

    3. 3

      Convert the top ten

      Most repeated workflows into templates.

      Senior AI Engineer 6 weeks

    4. 4

      Shared evals

      Every template ships with a replay set.

      Senior AI Engineer 4 weeks

    5. 5

      Public catalogue

      Workflow list on the solution pages.

      Founding Product Engineer 2 weeks

    6. 6

      Measure reuse

      Share of new deployments starting from a template.

      Co-Founder and CTO monthly

    First result
    1 quarter
    Cost
    A quarter of engineering focus
    Owner
    Co Founder and CTO

    Smallest test

    Reuse one NetSuite workflow for the next NetSuite customer and time it.

    Why it works

    Turn each deployed workflow into a versioned template: the systems, steps, approval points, exception rules and eval set. When a new customer arrives, start from the closest template and adapt it. Track reuse rate and time to first live run. Show the library publicly as a catalogue ("42 finance workflows ready to run") on the solution pages.

    Reuse cuts deployment time and cost on every new customer, makes quality consistent through shared evals, and gives buyers a concrete menu instead of a blank page.

    Risks and how to handle them

    • Customer-specific logic leaks into templates

      Keep customer data and rules out of the template layer and review each one.

    Evidence

    • Your websiteHomepage (as served to browsers, checked November 9, 2026): title "Zalos - The finance agent factory"; description "Build AI workforces that run finance workflows across fragmented systems. Zalos computer agents log into your ERP, map your processes, then automate them and get better every month." The headline changed from "Computer Agents for Finance Operations" to "The finance agent factory." and the site moved from Webflow and HubSpot to Next.js. No pricing page, case studies, testimonials or live chat; a strip of 12 logos.Source
    • News coverageHg Orbit podcast with the CEO: Zalos was "already deployed with major enterprise customers" five weeks after launch, including "companies with ... billions in revenue"; one customer is "likely to get rid of a point solution SaaS"; delivery uses forward deployed engineering to map each customer's workflows; the plan is to start with the biggest companies and become "increasingly self-serve to come down the market"; the goal is "12 million in 12 months".Source
    • HiringSenior AI Engineer (London, £70k-£140k, 0.20-1.50% equity, open on Y Combinator's job board): "We have paying enterprise customers, strong seed funding, and a small team that moves fast"; builds evals, replay testing and regression detection, computer-use agents for legacy interfaces such as SAP GUI and thick clients, and cross-system workflows across SAP, Salesforce and ServiceNow; estimates 30-40% of enterprise processes lack reliable APIs; works directly with founders and customers.Source
  • 19

    Zalos has already done the hard part twice. It assembled finance-safety datasets and tested a finance-specific gate against a general filter (96.5% vs 56.9%), and it ran a 100-task browser benchmark (94 of 100), while excluding a dataset whose licence forbids commercial use. The browser post promises more evaluations. There is no recognized benchmark for agents doing real finance work in real finance software, and whoever publishes one sets the terms buyers use to judge.

    High impactHigh effort First result in 2 quarters Co Founder and CTO

    First action

    Write the task list and scoring rules and ask two finance leaders to review them.

    How to measure it

    Qualified leads from the benchmark launch

    Now
    –
    Target
    40
    By
    first edition + 60 days

    Steps

    1. 1

      Design tasks

      Thirty to fifty finance tasks across systems.

      Senior AI Engineer 1 month

    2. 2

      Build sandboxes

      Synthetic companies and data.

      Senior AI Engineer 6 weeks

    3. 3

      Advisory review

      Two controllers and an auditor review realism.

      Co-Founder and CEO 2 weeks

    4. 4

      Run systems

      Zalos and other public agent systems, with methods disclosed.

      Co-Founder and CTO 1 month

    5. 5

      Publish

      Report, open task set, leaderboard and Markdown version.

      Founding Product Engineer 2 weeks

    6. 6

      Launch

      Webinar for finance leaders and press outreach.

      Co-Founder and CEO 2 weeks

    First result
    2 quarters
    Cost
    A quarter of one engineer plus launch costs
    Owner
    Co Founder and CTO

    Smallest test

    Publish a 20-task finance browser set with Zalos's own score.

    Why it works

    Publish a yearly benchmark of finance tasks in realistic sandboxes: NetSuite-style categorization, portal downloads behind 2FA, reconciliations, SAP-style GUI postings and harmful-request refusals. Release the task set and scoring openly, run named agent systems including Zalos, and report completion, accuracy, exceptions and safety. Launch with a report, a webinar for finance leaders and a leaderboard page.

    A neutral, rigorous benchmark earns press, analyst attention and AI-assistant citations, and positions Zalos as the standard-setter in a crowded agent market. It also doubles as a recruiting magnet for AI engineers.

    Risks and how to handle them

    • Seen as self-serving

      Open tasks and scoring, external reviewers, and report Zalos's own misses.

    • Licensing of test data

      Use synthetic data, as the team already does.

    Builds on

    Evidence

    • Your websiteBlog, "How Zalos catches 96.5% of harmful finance prompts" (November 3, 2026): the Safety Gate caught 747 of 774 harmful finance prompts (96.5%) against 56.9% for a general-purpose moderation filter, with no false alarms on 234 benign finance prompts; it covers insider trading, market manipulation, accounting fraud, tax evasion, AML and sanctions evasion and prompt injection, and separately flags sensitive data such as payroll records and bank details. No customers named.Source
    • Your websiteBlog, "A better browser agent, without a new browser" (October 27, 2026): Zalos completed 94 of 100 tasks on the BU-Bench-V1 browser benchmark and showed a desktop demo entering a split expense in GnuCash on a remote Windows desktop; the post previews future evaluations and ends without a demo or contact call to action. The blog index lists two posts.Source
    • HiringSenior AI Engineer (London, £70k-£140k, 0.20-1.50% equity, open on Y Combinator's job board): "We have paying enterprise customers, strong seed funding, and a small team that moves fast"; builds evals, replay testing and regression detection, computer-use agents for legacy interfaces such as SAP GUI and thick clients, and cross-system workflows across SAP, Salesforce and ServiceNow; estimates 30-40% of enterprise processes lack reliable APIs; works directly with founders and customers.Source

Wildcards 3

Unexpected experiments, each with a cheap way to test it.

  • 20

    Every Zalos run is recorded with screen recordings and logs, with optional human review and "an audit trail for regulators and stakeholders". External auditors spend a lot of time testing controls over manual finance processes. If an audit firm accepts Zalos's run evidence as a standard format, the audit trail becomes a reason for a CFO to buy rather than only a reassurance.

    Medium impactMedium effort First result in 2 quarters Co Founder and CEO

    First action

    Ask one current customer's external auditor what evidence they would need from an agent-run process.

    How to measure it

    Customers using the evidence pack in an audit

    Now
    –
    Target
    3
    By
    3 quarters

    Steps

    1. 1

      Interview auditors

      Three audit and advisory firms.

      Co-Founder and CEO 1 month

    2. 2

      Define the pack

      Fields, retention and export format.

      Co-Founder and CTO 2 weeks

    3. 3

      Build export

      Per workflow and period.

      Founding Product Engineer 3 weeks

    4. 4

      Pilot

      With one customer's year-end audit.

      Forward deployed engineer 1 quarter

    5. 5

      Joint guide

      Auditing agent-run finance processes.

      Co-Founder and CEO 1 month

    6. 6

      Referral terms

      Where independence rules allow.

      Co-Founder and CEO 2 weeks

    First result
    2 quarters
    Cost
    Founder and engineering time
    Owner
    Co Founder and CEO

    Smallest test

    Show one auditor a sample evidence pack with synthetic data.

    Why it works

    Work with one or two mid-tier audit or advisory firms to define an evidence pack per workflow: inputs, actions, approvals, exceptions, recordings and change notes, exportable per period. Publish a joint guide on auditing agent-run finance processes. Offer the firms' clients a preferred sprint and the firms a referral arrangement where independence rules allow.

    Auditors influence what controllers trust. An accepted evidence format lowers the perceived risk of agents for every prospect, and it is hard for rivals without full recordings to match.

    Risks and how to handle them

    • Auditor independence limits on referrals

      Keep the joint work to guidance and formats; take legal advice on referral terms.

    Builds on

    Evidence

    • Your websiteAI-reader files (fetched November 9, 2026): llms.txt lists only the homepage and careers as key pages and offers "a Markdown version of the homepage (request with Accept: text/markdown)". That Markdown homepage still leads with the old headline "Zalos: Computer Agents for Finance Operations", describes Connect, Act and Audit stages, names Oracle, NetSuite, SAP and QuickBooks, and ends with the contact page and a founder's email. Neither mentions the finance agent factory, the four solution areas, the platform, security or the blog.Source
    • News coverageFondo (October 30, 2026): Zalos launched AI workers for order-to-cash, procure-to-pay and record-to-report that log into browser billing tools and desktop applications, including on-premise ERPs, run in a customer's virtual private cloud or on an air-gapped device, log decisions and learn from resolved exceptions. Zalos did not disclose customer numbers, "measured exception rates, process-completion rates" or before-and-after costs. Offer: bring a finance process to automate, or get AirPods or a charity donation.Source
  • 21

    Zalos's workers can run in a customer's virtual private cloud or on an air-gapped local device, and they handle on-premise ERPs and desktop applications. Banks, insurers, utilities and defence suppliers often run exactly those legacy systems and rule out cloud agents. The site does not offer this option as its own path, and the /security page did not load for AI readers.

    Medium impactMedium effort First result in 2 quarters Co Founder and CEO

    First action

    Write a two-page architecture note for private and air-gapped deployment.

    How to measure it

    Private-deployment pipeline

    Now
    –
    Target
    5 opportunities
    By
    2 quarters

    Steps

    1. 1

      Document architecture

      What runs where, updates, logs.

      Co-Founder and CTO 1 week

    2. 2

      Check update path

      How the Safety Gate policy updates offline.

      Senior AI Engineer 1 week

    3. 3

      Build the page

      Deployment options with a Markdown version.

      Founding Product Engineer 3 days

    4. 4

      Set pricing

      A premium for private deployment.

      Co-Founder and CEO 2 days

    5. 5

      Target accounts

      Regulated mid-size companies on on-premise ERPs.

      Co-Founder and CEO 2 weeks

    6. 6

      Outreach

      Through partners and investor introductions.

      Co-Founder and CEO ongoing

    First result
    2 quarters
    Cost
    Founder and engineering time
    Owner
    Co Founder and CEO

    Smallest test

    Pitch the private option to three regulated prospects already in the pipeline.

    Why it works

    Create a deployment page and a short architecture note for regulated buyers: what runs where, what never leaves the network, how updates and the Safety Gate policy are delivered without external access, and how audit evidence is stored. Target finance-operations leaders at regulated mid-size companies with on-premise ERPs through partners and investor introductions. Price the private deployment at a premium.

    Most agent vendors are cloud-only, so a credible air-gapped option wins deals where competition is thin and budgets for compliance are larger.

    Risks and how to handle them

    • Private deployments are costly to support

      Charge a premium and limit the number per year.

    • The hosted Safety Gate doesn't fit air-gapped setups

      Ship signed policy updates on a schedule.

    Builds on

    Evidence

    • News coverageFondo (October 30, 2026): Zalos launched AI workers for order-to-cash, procure-to-pay and record-to-report that log into browser billing tools and desktop applications, including on-premise ERPs, run in a customer's virtual private cloud or on an air-gapped device, log decisions and learn from resolved exceptions. Zalos did not disclose customer numbers, "measured exception rates, process-completion rates" or before-and-after costs. Offer: bring a finance process to automate, or get AirPods or a charity donation.Source
    • HiringSenior AI Engineer (London, £70k-£140k, 0.20-1.50% equity, open on Y Combinator's job board): "We have paying enterprise customers, strong seed funding, and a small team that moves fast"; builds evals, replay testing and regression detection, computer-use agents for legacy interfaces such as SAP GUI and thick clients, and cross-system workflows across SAP, Salesforce and ServiceNow; estimates 30-40% of enterprise processes lack reliable APIs; works directly with founders and customers.Source
    • Your websiteSitemap (16 URLs, updated November 5, 2026): homepage, /platform, four solution pages (order-to-cash, procure-to-pay, record-to-report, cost-to-serve), /about, /security, /newsroom, /blog and two posts, /contact-us, /careers, /privacy-policy and /terms-conditions. Fetched with an AI reading tool on November 9, 2026 (/about twice), /platform, all four solution pages, /about, /security and /newsroom returned "404 Not Found"; the homepage, /blog, both posts, /contact-us, /careers, /privacy-policy, robots.txt and llms.txt loaded.Source
  • 22

    Real visitors already pass Core Web Vitals on zalos.ai (LCP 1.41 s at the 75th percentile, 31% faster than before), so speed is not a crisis. But the mobile lab test still scores 65 with a 10.4 s LCP, and the largest third-party cost is the Facebook pixel: 279 KB and 283 ms of blocking time. That is on a site selling to CFOs, which also loads LinkedIn Ads, Apollo and PostHog.

    Low impactLow effort First result in 1 week Founding Product Engineer

    First action

    Check whether any Facebook campaign has run in the last 90 days.

    How to measure it

    PageSpeed mobile score

    Now
    –
    Target
    80+
    By
    30 days

    Steps

    1. 1

      Check usage

      Facebook campaigns and audiences in the last 90 days.

      Co-Founder and CEO 1 hour

    2. 2

      List tags

      Every marketing and analytics script, owner and purpose.

      Founding Product Engineer 2 hours

    3. 3

      Remove or defer

      Drop unused tags; load the rest after idle or consent.

      Founding Product Engineer half a day

    4. 4

      Fix caching

      Longer cache lifetimes on static assets (about 1.7 s of lab savings).

      Founding Product Engineer 2 hours

    5. 5

      Re-test

      PageSpeed mobile and desktop.

      Founding Product Engineer 1 hour

    6. 6

      Review quarterly

      Tag list against current campaigns.

      Founding Product Engineer quarterly

    First result
    1 week
    Cost
    $0
    Owner
    Founding Product Engineer

    Smallest test

    Defer the pixel until page idle and compare the lab score.

    Why it works

    Decide whether Facebook advertising is actually used for enterprise finance buyers. If not, remove the pixel. If it is, load it after consent or after the page is idle. Review the other marketing and analytics tags at the same time: confirm each has an owner and a purpose and that no two do the same job. Re-run PageSpeed afterwards.

    Fewer tags make the page faster on slower phones and corporate networks, keep data collection clear for privacy-conscious finance buyers, and leave less to maintain.

    Risks and how to handle them

    • Losing retargeting audiences

      Keep LinkedIn, which matches the B2B audience, and confirm Facebook isn't in use first.

    Evidence

    • PageSpeedHomepage, Google PageSpeed (November 9, 2026): real Chrome users pass Core Web Vitals (LCP p75 1.41 s, 31% faster than before; INP 62 ms; CLS 0), although the mobile lab test scores 65 with a 10.4 s LCP. Heaviest third-party: the Facebook pixel, 279 KB and 283 ms of blocking time, on an enterprise B2B page that also loads LinkedIn Ads, Apollo and PostHog; 1,408 KB total; efficient cache lifetimes would save about 1.7 s in the lab.Source

03

How Programmable can help

Where an idea in this playbook is something Programmable builds or runs, it's listed here with what we'd do. Everything else is yours to run.

Have it built

ProgrammableAgency

1 of the 22 ideas are things Programmable builds and runs for its customers: pages, outreach, prospect lists, booking flows and monitors. Pick one, or send the whole list.

Prospect lists1The list compiled from public sources, verified, deduplicated and delivered to your CRM.

A 30-minute session with the engineer who would build it, not a sales call. You leave with a plan whether or not you hire us.

Book a working session

Usually live within days.

Tools for the rest

ProgrammableInfo

Turn the research into a list you can act on

These ideas start with finding the right companies. Info finds them and the right person at each, researches every one and drafts the outreach, then asks before it sends.

  • Build this list in Info for idea #10 (opens programmable.info in a new tab)

New accounts: 100,000 free credits with code BETA100K

Find the companies (opens programmable.info in a new tab)
More example playbooks and reviews

More growth playbooks

Other companies we have researched, each read from what they are hiring for.

  • Lightspeed Systems growth strategyCloud software for K-12 schools: web filtering, student safety monitoring, classroom management, digital hall passes, device management…21 growth ideas
  • Chapman Bright growth strategyA Breda-based B2B consultancy that redesigns marketing and sales operations with automation, AI and agents, specialised in Adobe Marketo…24 growth ideas
  • elefante RevOps growth strategyA revenue operations agency and HubSpot Diamond Solutions Partner: RevOps as a Service delivered by small pods, a fixed-scope GTM…24 growth ideas
  • OLAPLEX growth strategyPrestige hair care built on patented bond-building technology: treatments, shampoos, conditioners, masks, stylers and new lash, brow and…21 growth ideas
  • Reformation growth strategyA Los Angeles sustainable fashion brand selling women's dresses, denim, knits, shoes, bags and bridal online, in 70 of its own stores and…22 growth ideas
  • Starknet Foundation growth strategyThe non-profit steward of Starknet, a validity rollup that scales Ethereum and now settles on Bitcoin too; it funds builders through…21 growth ideas
  • CNTXT AI growth strategyAn Abu Dhabi data and AI company that captures and verifies data from real work (including through its KLKT contributor app), trains…22 growth ideas
  • Avon Anglers growth strategyA small-batch menswear and accessories brand made in New York, with outdoor and angling roots, plus Re-Wear, a line of reworked and…20 growth ideas

See every playbook and website review

Independent research. Programmable wrote this playbook from public sources, as they stood on November 10, 2026, and cites them throughout. Zalos did not commission, review or endorse it, and Programmable is not affiliated with Zalos. Company and product names are trademarks of their owners, used only to identify them. The ideas are suggestions and opinion, not professional, legal or financial advice, and not statements about how Zalos runs its business.

Something wrong, or do you represent Zalos? Email [email protected] and we will correct it or take it down.

30/60/90 roadmap

What to focus on in each month, and the ideas that carry it.

  1. Days 1 to 306 ideas

    Fix the AI-reader site, replace the inbox-only contact page, add product CTAs to the blog, publish the first case study and a controls brief, and trim tracking scripts.

  2. Days 31 to 6011 ideas

    Public scorecard, ERP pages, the paid process sprint, partner and investor channels, comparison pages, monthly CFO reports, replay results, the content series, the screen-recording audit and a published starting price.

  3. Days 61 to 905 ideas

    Commit to the big bets and wildcards: a self-serve midmarket tier, a workflow library, an annual finance-agent benchmark, audit-firm partnerships and an air-gapped offer for regulated companies.

The research behind the plan

3 research areas, each summarised in a few lines. Open any of them for the full detail.

Customers

What customers say

Customers value automation without replacing or migrating the ERP. The most cited problem: legacy ERPs with limited AI.

  • Agents that handle logins, 2FA and portals that have no API
  • Too many finance systems that don't talk to each other
  • In their words: “fragmented systems”

Market

Competitive landscape

4 competitors compared: Hyphenate, Ledge, Kognitos and UiPath.

  • Gap to exploit: Finance agents for the 30-40% of processes with no reliable API
  • Hyphenate: A platform to adopt rather than agents on top of existing tools
  • Ledge: Close-focused; NetSuite is the only ERP it names

Hiring signals

Zalos is a ten-person team hiring engineers only, in London, and both posts point to the same next stage.

  • Senior AI Engineer: Evals, replay testing, SAP GUI computer-use agents; paying enterprise customers. Ideas 6, 13, 18, 19.
  • Founding Product Engineer: Works directly with users on a Next.js stack. Ideas 1, 2, 17.

From the team that wrote this playbook

All products
  • ProgrammableAgencyWorking AI systems in days, not quarters (opens programmable.agency in a new tab)
  • ProgrammableInfoAn agent that finds, researches and asks firstNew accounts get 100,000 free enrichment credits with code BETA100K (opens programmable.info in a new tab)
  • ProgrammableDigitalScreenshots in, an animated product demo out (opens programmable.digital in a new tab)
  • ProgrammableWatchKnow the day a supplier falls out of compliance (opens programmable.watch in a new tab)
  • ProgrammableTalkGet registrants for your events, and know who’s coming (opens programmable.talk in a new tab)
  • ProgrammableMediaAn AI note taker, interpreter and webinar host for your calls (opens programmable.media in a new tab)
  • ProgrammablePageText what you need, your customer gets a page that works (opens programmable.page in a new tab)
  • ProgrammableCommunityForums with labelled AI members, never empty (opens programmable.community in a new tab)
ProgrammableMarketing
ExamplesLeaderboardsProductsTermsPrivacyCookiesDPALegalContact

© 2026 AI-Powered Growth Intelligence