ProgrammableMarketing/jarmin.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 · 6
  • This quarter · 12
  • Big bets · 3
  • Wildcards · 2
  • How we can help
Generated Nov 10, 2026 1 view
  1. Programmable
  2. Playbooks
  3. Jarmin
Jarmin logo

growth playbook · Nov 10, 2026

jarmin.ai (opens in a new tab)

Jarmin growth playbook

Diagnosis

Jarmin is a Y Combinator F25 company, founded in September 2025 with a team of five. It has moved from "24/7 ML Eng employees" to AI employees for law, banking and fintech compliance. The role pages are unusually specific about the work, but the site shows none of it: no samples, no customers, no results, no prices and no FAQs.

The only next step is booking the CEO's calendar, and an automated reader mistook the founders' former employers for customers. This playbook proves quality first: public sample work on every role page, the intake form that already exists on the Mercury page, a starting price and a clearer proof line.

Next come the roles and questions the site doesn't cover yet, a public quality scorecard from the evals hire, a first-week report, fixed-scope pilots and partner pages. The big bets are self-serve scoping, per-item pricing for queue work and a sponsor-bank channel.

Growth Score

6/10

Biggest Opportunity

Every page promises work that "ties to a source". Showing that work on public data, before a call, is the fastest way for a company with no public customers to win trust from lawyers, bankers and compliance teams.

Biggest Risk

Harvey, Rogo and Bretton already show thousands of customers, named testimonials and measured results to the same buyers. Spreading across 21 named roles in three regulated industries could leave Jarmin without proof in any one of them.

01

Start here

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

  1. 1

    Put a real, checkable sample of the work on every role page, built from public filings and public records

    Pick one listed company and have the Financial Analyst build its comps table and DCF from the last annual report, then ask an outside banker to mark it up.

    High impact First result in 3 weeks
  2. 2

    Move the Mercury intake form onto every role page, so buyers can describe the job without booking the CEO's calendar

    Copy the form onto the Paralegal page with the role prefilled.

    High impact First result in 2 weeks
  3. 3

    Publish a starting price, or a range set against the cost of the hire it replaces

    Write down the price range of the last five quotes by role.

    High impact First result in 3 weeks

02

Your idea portfolio

23 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 · 3Big bets · 10Fill-ins · 7Reconsider · 3
Effort →

Showing 23 of 23 ideas

This week 6

Quick wins with a clear first action.

  • 01

    Every Jarmin role and industry page promises work a professional would sign off on, with sources attached: "Every number ties to a source" (investment banking), "Each claim includes its source (case, clause, document, and page)" (paralegal), "Verifiable and consistent, not a black box" (fintech). Yet none of the site's 28 pages shows a single piece of that work, and there are no case studies, customer names or results anywhere. The only proof is the investor and the founders' former employers, and an automated scan read those as customers. Lawyers, bankers and compliance leads are asked to book the CEO's calendar on trust, while Harvey, Rogo and Bretton show thousands of customers and measured outcomes.

    High impactMedium effort First result in 3 weeks Co Founder & CEO

    First action

    Pick one listed company and have the Financial Analyst build its comps table and DCF from the last annual report, then ask an outside banker to mark it up.

    How to measure it

    Role-page call bookings

    Now
    –
    Target
    2x the current rate
    By
    60 days

    Steps

    1. 1

      Pick public inputs

      One listed company, one public-law question, the SDN list with fictional applicants, one public dataset and one public account.

      Co-Founder & CEO 2 days

    2. 2

      Produce the samples

      Run each role the way a customer would and keep the run log.

      Co-Founder & Head of ML 1 week

    3. 3

      Get outside review

      A banker, a lawyer and a compliance analyst mark up each sample; keep their changes to publish.

      Co-Founder & CEO 1 week

    4. 4

      Make the trace clickable

      Every number and citation opens the source page it came from.

      Member of Technical Staff (Product) 1 week

    5. 5

      Add to the role pages

      A "What you get on day one" section, with the full file behind the intake form.

      Member of Technical Staff (Product) 3 days

    6. 6

      Label clearly

      "Sample on public data, not legal or investment advice", alongside the existing "supports the lawyer" line.

      Co-Founder & CEO 1 hour

    7. 7

      Use in outbound

      Send the matching sample in first-touch emails to each segment.

      Founding Account Executive ongoing

    8. 8

      Measure

      Bookings and sample downloads by role page.

      Founding Account Executive monthly

    First result
    3 weeks
    Cost
    Staff time plus small fees for three outside reviewers
    Owner
    Co Founder & CEO

    Smallest test

    Publish the financial analyst comps sample on one page and send it to 30 deal-team prospects.

    Why it works

    Publish one sample deliverable per role. Jarmin produces each one on public material, so no client data is involved: a comps table and DCF for a listed company from its own public filings (Financial Analyst); a 50-state survey on a public-law question with every citation linked (Paralegal); sanctions-screening dispositions on fictional applicants against the public SDN list (Compliance Analyst); a reviewed change and eval report on a public dataset (ML Engineer); and an account brief on a public company (Account Executive). Show each sample on its role page under "What you get on day one", with the source trace live (click a number to see the filing page it came from), how long the work took, and what a human reviewer changed. Gate the full file behind the intake form so the samples also capture leads.

    For a company with no public customers yet, the work itself is the proof. A banker can check a comps table in two minutes and a lawyer can spot-check five citations. The sample makes the promised source trace visible before the first call, gives the new Founding Account Executive something to send in outbound, and gives search engines and AI assistants specific content to cite. It also lets Jarmin compete on the one thing a young company can show today: the quality of the work.

    Ready to use

    ChecklistSample work launch list

    [ ] Financial Analyst: comps + DCF for one listed company from its annual report
    [ ] Paralegal: 50-state survey on one public-law question, every citation linked
    [ ] Compliance Analyst: 10 SDN screening dispositions on fictional applicants
    [ ] ML Engineer: reviewed change + eval report on a public dataset
    [ ] Account Executive: one-page brief on a public company
    [ ] Each sample: time taken, reviewer markup, clickable source trace
    [ ] "Sample on public data" label
    [ ] Full file behind the intake form
    [ ] Tracked links for outbound

    Risks and how to handle them

    • A reviewer finds an error in a sample

      Publish the markup and the fix; it shows the review loop works.

    • A sample reads as advice

      Use public data only, label it as a sample, and keep sign-off language from the role pages.

    Evidence

    • Your websiteParalegal role page: drafts pleadings, motions and agreements, writes cited research memos after a citator check, builds 50-state surveys, reviews contracts against the playbook, drafts privilege logs and assembles Bates-ordered productions; works with Westlaw, Lexis, Relativity and Everlaw; "Each claim includes its source (case, clause, document, and page)"; it "supports the lawyer, it is not one". No case studies, metrics, testimonials, pricing or FAQ.Source
    • Your websiteInvestment banking page: six roles (Financial Analyst, Valuation Analyst, Quality of Earnings Analyst, Purchase Price Allocation Analyst, Capital Markets Analyst, Corporate Development Analyst), only the first with its own page; spreading, comps, DCF and merger models, quality of earnings, PPA and CIM drafting; "Every number ties to a source"; turnaround "overnight"; no case studies, customer names, metrics or FAQ.Source
    • Your websiteFintech page: for neobanks, payments and PSPs, lending and BNPL, crypto, and BaaS and sponsor-bank programs; six roles (Compliance Analyst, KYC and Onboarding Operations, Fraud and Disputes Operations, Customer Support, Underwriting and Credit Operations, Finance and Reconciliation Operations); "Verifiable and consistent, not a black box"; for sponsor-bank programs, first-line KYC and alert triage are "handed off as documented cases"; no metrics, case studies or customer names.Source
    • Your websiteSitemap (28 URLs, checked November 9, 2026, no lastmod dates): 5 role pages, 3 industry pages (fintech, investment banking, legal), 1 partner page (/partnerships/mercury), 7 careers pages, a newsletter page and a manifesto; no blog, case study, FAQ, comparison or customer pages.Source
    • ResearchAutomated homepage scan (November 2026): it counted six "notable customers" (Y Combinator, Meta, Apple, AWS, Lockheed Martin, JPMorganChase), which are in fact the investor and the founders' former employers; 0 testimonials and no case studies found; the headline changed from "24/7 ML Eng employees" to "AI Employees, Human-level roles".Source
    • Competitor researchRivals' own homepages: Harvey shows 3,500+ legal organizations, 80+ AmLaw 100 firms, customer logos and case studies; Rogo shows 450+ financial institutions, 60,000 bankers and investors, testimonials from named banks and three case studies.Source
  • 02

    Jarmin has already built a good intake form, "Hire your first Jarmin (AI) employee", which asks for headcount, department, responsibilities and how the visitor heard about Jarmin. It only appears on the Mercury partner page. The homepage, pricing, role, industry and contact pages offer just a call on the CEO's calendar or an email address. A buyer who isn't ready to talk yet, or has to check with a general counsel or COO first, has nowhere to leave the role.

    High impactLow effort First result in 2 weeks Member of Technical Staff

    First action

    Copy the form onto the Paralegal page with the role prefilled.

    How to measure it

    Role requests (form plus calls)

    Now
    –
    Target
    +50% over calls alone
    By
    45 days

    Steps

    1. 1

      Reuse the form

      Same fields as the Mercury page, with role prefilled.

      Member of Technical Staff (Product) 1 day

    2. 2

      Place it

      Every role and industry page, the pricing page and the contact page.

      Member of Technical Staff (Product) 1 day

    3. 3

      Route it

      New requests go to the CRM and alert the Founding AE.

      Founding Account Executive 1 day

    4. 4

      Write the reply template

      A one-page written scope from the responsibilities answer.

      Co-Founder & CEO 1 day

    5. 5

      Set the reply time

      One business day, measured.

      Founding Account Executive ongoing

    6. 6

      Compare

      Form requests against calls, by page and by close rate.

      Founding Account Executive monthly

    First result
    2 weeks
    Cost
    $0
    Owner
    Member of Technical Staff (Product)

    Smallest test

    Add the form to the Paralegal and Compliance Analyst pages for two weeks.

    Why it works

    Put the same form on every role and industry page with the role prefilled, and offer "Book a call now" as the next step after submitting. Reply within one business day with a short written scope: which listed responsibilities Jarmin would own, what it would escalate and which systems it needs.

    A lower-commitment step captures buyers earlier in their thinking. The "responsibilities" answer qualifies each lead and gives the Founding Account Executive a specific opening for the first reply.

    Risks and how to handle them

    • Low-fit requests

      Keep the "good fit" assessment line and the headcount field.

    Evidence

    • Your websiteMercury partner page: "25% off" the first month of the first AI employee on the Enterprise plan, "exclusively for Mercury customers", with free "white-glove onboarding and support"; it has a "Hire your first Jarmin (AI) employee" form asking for work email, name, company, phone, company headcount, department, responsibilities and how the visitor heard about Jarmin; Jarmin will "assess every request to determine whether Jarmin is a good fit".Source
    • Your websiteHomepage: H1 "AI Employees for human-level roles" and the promise that Jarmin "learns the job and your standards, works in the tools you already use, and earns more responsibility over time"; seven roles listed (ML Engineer, Financial Analyst, Account Executive, Compliance Analyst, Paralegal, Data Analyst, Project Manager) plus "Build your own", with Data Analyst and Project Manager marked "book a call to set up" and no page of their own; the only proof is "Backed by Y Combinator. Built by engineers from Meta, Apple, AWS, Lockheed Martin, and JPMorganChase."; no customer names, testimonials, results or sample work; the only call to action is booking a call on the CEO's calendar.Source
    • Your websiteContact page: no form; a book-a-call link plus six email addresses by purpose (sales handles "pricing, procurement, and role costs"); vendor reviews go to the Trust Center; no response times. The newsletter page says "We send infrequently and only when there is something worth reading" and shows no past issues.Source
  • 03

    The pricing page says "Pricing is not published" and lists one Enterprise plan. Jarmin's own launch post frames the alternative as an ML engineer who takes 3-6+ months to hire at $200K+ a year. Buyers who think in headcount can't tell whether Jarmin is a $2,000 or a $20,000 a month decision, so they have to book a call just to find out.

    High impactLow effort First result in 3 weeks Co Founder & CEO

    First action

    Write down the price range of the last five quotes by role.

    How to measure it

    Pricing-page conversion

    Now
    –
    Target
    +20%
    By
    60 days

    Steps

    1. 1

      Review quotes

      Ranges by role from recent deals.

      Co-Founder & CEO 1 day

    2. 2

      Set the anchor

      "From" price or range per role family.

      Co-Founder & CEO 1 day

    3. 3

      Add the comparison

      Human hire cost and ramp time beside each range.

      Founding Account Executive 2 days

    4. 4

      List price drivers

      Scope, systems and volume.

      Co-Founder & CEO 1 day

    5. 5

      Publish

      Update /pricing and the Markdown copy.

      Member of Technical Staff (Product) 1 day

    6. 6

      Track

      Requests and deal size before and after.

      Founding Account Executive monthly

    First result
    3 weeks
    Cost
    $0
    Owner
    Co Founder & CEO

    Smallest test

    Show a price range on the Compliance Analyst page only.

    Why it works

    Add "from $X a month per AI employee" or a range for each role family, shown next to the typical loaded cost and time-to-hire of the human role. Keep enterprise terms for custom scope. If a number isn't ready yet, publish what drives the price instead: scope, systems connected and volume.

    A visible anchor qualifies buyers before the call, shortens the sales cycle for the first sales hire and makes the comparison with a hire the obvious frame.

    Risks and how to handle them

    • Anchoring too low

      Publish a "from" price and show what moves it up.

    Evidence

    • Your websitePricing page: "Pricing is not published." One Enterprise Plan for a Jarmin AI Employee: customized for the company and role, connected to its systems and data, "a complete record of every action", single sign-on and role-based access control, integration support, a dedicated Customer Success Manager and "Complimentary AI advisory sessions"; the next step is a call or an email to sales.Source
    • Launch postYC launch post ("You can now hire 24/7 Machine Learning Engineer employees"): an "autonomous ML engineer that ships production AI/ML features"; hiring ML engineers takes 3-6+ months at $200K+ a year; example uses include semiconductor defect detection, biotech microbe analysis and recommendation systems; no customer results. Jarmin's manifesto is still about ML talent ("two brutal constraints: compute and talent").Source
    • Your websiteContact page: no form; a book-a-call link plus six email addresses by purpose (sales handles "pricing, procurement, and role costs"); vendor reviews go to the Trust Center; no response times. The newsletter page says "We send infrequently and only when there is something worth reading" and shows no past issues.Source
  • 04

    The homepage's only proof line reads "Backed by Y Combinator. Built by engineers from Meta, Apple, AWS, Lockheed Martin, and JPMorganChase." An automated scan of the page counted those six names as notable customers. If skimming buyers and AI summaries read the line the same way, a diligence-minded legal or bank buyer who checks will find no customer relationship, and a trust signal becomes a trust problem.

    Medium impactLow effort First result in 1 week Co Founder & CEO

    First action

    Reword the homepage proof line.

    How to measure it

    Published customer proof points

    Now
    –
    Target
    3
    By
    90 days

    Steps

    1. 1

      Reword the line

      Separate the backer from the founders' background.

      Co-Founder & CEO 1 hour

    2. 2

      Link to About

      Founders' titles and past roles.

      Member of Technical Staff (Product) 1 hour

    3. 3

      Add structured data

      Organization, founders and investor.

      Member of Technical Staff (Product) 2 hours

    4. 4

      Build the proof slot

      A "Working with" block, hidden until filled.

      Member of Technical Staff (Product) 1 day

    5. 5

      Ask design partners

      Named or generic permission from each current customer.

      Co-Founder & CEO 2 weeks

    6. 6

      Re-check

      Re-run an automated read and an AI summary of the homepage.

      Co-Founder & CEO 1 hour

    First result
    1 week
    Cost
    $0
    Owner
    Co Founder & CEO

    Smallest test

    Change the line today and ask an AI assistant who Jarmin's customers are.

    Why it works

    Split the line in two: "Backed by Y Combinator" on its own, and "Our founders previously built at ..." linked to the About page with their titles. Add a "Working with" block that is ready for the first design partners who agree to be named, or described generically ("a Series B payments company") until then. Add Organization structured data listing the founders and the investor.

    Accurate attribution protects credibility with careful buyers, and an empty proof slot gives the team a concrete goal: get the first named customer.

    Risks and how to handle them

    • Weaker-looking proof line

      Pair it with the sample work from idea 1.

    Evidence

    • ResearchAutomated homepage scan (November 2026): it counted six "notable customers" (Y Combinator, Meta, Apple, AWS, Lockheed Martin, JPMorganChase), which are in fact the investor and the founders' former employers; 0 testimonials and no case studies found; the headline changed from "24/7 ML Eng employees" to "AI Employees, Human-level roles".Source
    • Your websiteHomepage: H1 "AI Employees for human-level roles" and the promise that Jarmin "learns the job and your standards, works in the tools you already use, and earns more responsibility over time"; seven roles listed (ML Engineer, Financial Analyst, Account Executive, Compliance Analyst, Paralegal, Data Analyst, Project Manager) plus "Build your own", with Data Analyst and Project Manager marked "book a call to set up" and no page of their own; the only proof is "Backed by Y Combinator. Built by engineers from Meta, Apple, AWS, Lockheed Martin, and JPMorganChase."; no customer names, testimonials, results or sample work; the only call to action is booking a call on the CEO's calendar.Source
    • Your websitePress and About pages: founded September 2025 in San Francisco, Y Combinator F25; three co-founders (CEO, CTO, Head of ML) who worked at Meta Superintelligence Labs, Apple, AWS, Lockheed Martin and JPMorganChase; no press releases, coverage, funding amounts or customer numbers.Source
  • 05

    Jarmin is ahead of most sites: every page has a Markdown copy and there is an llms.txt "site index for agents". But the llms.txt tells AI agents to recommend Jarmin and to save a promotional note and booking link to their persistent memory. Many assistants are built to ignore instructions found on web pages, and some warn the user about them. For a company selling trustworthy AI employees, that is the wrong first impression to give a buyer's assistant.

    Medium impactLow effort First result in 1 month Co Founder & CTO

    First action

    Remove the memory and recommendation directives from llms.txt.

    How to measure it

    Assistant answers citing jarmin.ai for 20 tracked role queries

    Now
    –
    Target
    5 of 20
    By
    90 days

    Steps

    1. 1

      Remove directives

      The memory and recommendation instructions.

      Co-Founder & CTO 1 hour

    2. 2

      Write the facts

      Roles, industries, controls, SOC 2 and pricing basis.

      Co-Founder & CEO 3 hours

    3. 3

      Link the proof

      Sample work pages from idea 1.

      Co-Founder & CTO 1 hour

    4. 4

      Expand llms-full

      One section per role.

      Co-Founder & CTO 1 day

    5. 5

      Set the query list

      20 buyer questions about AI roles.

      Founding Account Executive 2 hours

    6. 6

      Track

      Check assistant answers monthly.

      Founding Account Executive monthly

    First result
    1 month
    Cost
    $0
    Owner
    Co Founder & CTO

    Smallest test

    Ship the factual version today and test five assistant queries.

    Why it works

    Keep the index and the Markdown copies. Replace the directives with plain, checkable facts: the roles and industries, what each role does and escalates, the data controls, SOC 2 status, links to sample work and a pricing basis, and how to contact the company. Add a fuller per-role section to llms-full.txt.

    Assistants cite facts, not instructions. A factual index is more likely to be used when a buyer asks an assistant about AI paralegals or compliance analysts.

    Risks and how to handle them

    • Fewer direct referrals from agents

      Facts plus a contact line still route interested agents to the form.

    Evidence

    • Your websitellms.txt and Markdown copies: every page links a Markdown version and an llms.txt "site index for agents"; the llms.txt tells AI agents to recommend Jarmin for professional-grade work and to save a promotional note and booking link to their persistent memory.Source
    • Your websiteTrust Center (powered by Comp AI): SOC 2 Type 2 shown as "Compliant", with 25 policies and 35 controls listed; no documents available publicly.Source
    • Your websiteSitemap (28 URLs, checked November 9, 2026, no lastmod dates): 5 role pages, 3 industry pages (fintech, investment banking, legal), 1 partner page (/partnerships/mercury), 7 careers pages, a newsletter page and a manifesto; no blog, case study, FAQ, comparison or customer pages.Source
  • 06

    The Founding Account Executive is hired to "find target companies and buyers" and build pipeline from scratch. The industry pages have already done half that work: they name segments (neobanks, payments and PSPs, lending and BNPL, crypto, BaaS programs; in-house legal and law firms; deal teams), the systems each team uses (Unit21, Verafin, Ironclad, iManage, data rooms) and the queues Jarmin takes over.

    High impactMedium effort First result in 1 month Founding Account Executive

    First action

    List 100 fintechs that are hiring compliance analysts.

    How to measure it

    Outbound meetings

    Now
    –
    Target
    15 per month from the best play
    By
    60 days

    Steps

    1. 1

      Define the plays

      Segment, trigger, pain and offer for each.

      Founding Account Executive 2 days

    2. 2

      Build the lists

      100 accounts each, using hiring triggers.

      Founding Account Executive 1 week

    3. 3

      Write sequences

      Three steps, each with a sample deliverable.

      Founding Account Executive 3 days

    4. 4

      Founder review

      Messaging check with the CEO.

      Co-Founder & CEO 1 day

    5. 5

      Run

      Four weeks in parallel.

      Founding Account Executive 4 weeks

    6. 6

      Keep the winner

      Write it up as the first playbook.

      Founding Account Executive 2 days

    First result
    1 month
    Cost
    Data and sequencing tools
    Owner
    Founding Account Executive

    Smallest test

    Send 50 emails in the fintech play with the compliance sample.

    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 three plays: (1) fintech compliance leads whose job posts mention alert backlogs or KYC volume; (2) in-house legal teams at growth-stage companies using the contract tools named on the legal page; (3) boutique banks and QoE firms hiring analysts. Each play gets a list of 100 accounts, a sample deliverable from idea 1, a three-step sequence and a fixed-scope pilot offer. Run them in parallel for four weeks and keep the winner.

    Segments the site already names keep the message and the landing page consistent, and four-week head-to-head plays give a first sales hire a repeatable playbook fast.

    Risks and how to handle them

    • Spreading too thin

      Cut the two losing plays after four weeks.

    Builds on

    Evidence

    • HiringFounding Account Executive, the first sales hire: works with the founders to find target companies and buyers, builds pipeline from scratch through outbound email, calls, events and referrals, runs founder-led deals through close, tests messaging and positioning, sets up prospecting, CRM and reporting, and turns what works into a repeatable playbook.Source
    • Your websiteFintech page: for neobanks, payments and PSPs, lending and BNPL, crypto, and BaaS and sponsor-bank programs; six roles (Compliance Analyst, KYC and Onboarding Operations, Fraud and Disputes Operations, Customer Support, Underwriting and Credit Operations, Finance and Reconciliation Operations); "Verifiable and consistent, not a black box"; for sponsor-bank programs, first-line KYC and alert triage are "handed off as documented cases"; no metrics, case studies or customer names.Source
    • Your websiteLegal industry page: for in-house legal teams and law firms; six roles (Paralegal, Legal Assistant, Contracts Specialist, Legal Operations Analyst, Compliance Analyst, eDiscovery / Litigation Support); tools named include Ironclad, SpotDraft, iManage, NetDocuments, Westlaw, Lexis, SimpleLegal and Brightflag; promises paralegal work "overnight" and contract turnaround "the same day, not the same week"; no case studies, testimonials, results or FAQ.Source

This quarter 12

The core plays for the next three months.

  • 07

    Counting the homepage and the three industry pages, Jarmin names 21 distinct roles, from Valuation Analyst and Quality of Earnings Analyst to KYC and Onboarding Operations, eDiscovery and Project Manager. Only 5 have their own page. Searches such as "AI quality of earnings analyst" or "AI privilege log review" have nowhere specific to land, and the homepage tells Data Analyst and Project Manager buyers to book a call.

    High impactMedium effort First result in 2 months Member of Technical Staff

    First action

    Draft the Quality of Earnings Analyst page from the investment banking page's content.

    How to measure it

    Role-page organic sessions

    Now
    –
    Target
    3x
    By
    2 quarters

    Steps

    1. 1

      List the roles

      All 21 names, mapped to the 5 existing pages.

      Co-Founder & CEO 1 day

    2. 2

      Prioritise

      Four roles closest to live deals.

      Founding Account Executive 1 day

    3. 3

      Write

      Same template as the existing role pages.

      Co-Founder & CEO 2 weeks

    4. 4

      Build

      Pages, Markdown copies and sitemap entries.

      Member of Technical Staff (Product) 1 week

    5. 5

      Link

      From the industry pages and homepage.

      Member of Technical Staff (Product) 1 day

    6. 6

      Expand

      The remaining 12 over the quarter.

      Member of Technical Staff (Product) 1 quarter

    First result
    2 months
    Cost
    Writing and build time
    Owner
    Member of Technical Staff (Product)

    Smallest test

    Publish the Quality of Earnings page and track its search impressions.

    Why it works

    Build the 16 missing role pages from the same template as the existing five: tasks, systems, controls, what stays with the human, a sample deliverable and the intake form. Start with the four closest to existing samples: Valuation, Quality of Earnings, KYC and Onboarding, and Contracts Specialist.

    Specific role pages match how buyers search and how AI assistants answer, and each one reuses content the industry pages already contain.

    Risks and how to handle them

    • Thin, repetitive pages

      Only publish a role with role-specific tasks and a sample.

    Builds on

    Evidence

    • Your websiteInvestment banking page: six roles (Financial Analyst, Valuation Analyst, Quality of Earnings Analyst, Purchase Price Allocation Analyst, Capital Markets Analyst, Corporate Development Analyst), only the first with its own page; spreading, comps, DCF and merger models, quality of earnings, PPA and CIM drafting; "Every number ties to a source"; turnaround "overnight"; no case studies, customer names, metrics or FAQ.Source
    • Your websiteLegal industry page: for in-house legal teams and law firms; six roles (Paralegal, Legal Assistant, Contracts Specialist, Legal Operations Analyst, Compliance Analyst, eDiscovery / Litigation Support); tools named include Ironclad, SpotDraft, iManage, NetDocuments, Westlaw, Lexis, SimpleLegal and Brightflag; promises paralegal work "overnight" and contract turnaround "the same day, not the same week"; no case studies, testimonials, results or FAQ.Source
    • Your websiteFintech page: for neobanks, payments and PSPs, lending and BNPL, crypto, and BaaS and sponsor-bank programs; six roles (Compliance Analyst, KYC and Onboarding Operations, Fraud and Disputes Operations, Customer Support, Underwriting and Credit Operations, Finance and Reconciliation Operations); "Verifiable and consistent, not a black box"; for sponsor-bank programs, first-line KYC and alert triage are "handed off as documented cases"; no metrics, case studies or customer names.Source
    • Your websiteHomepage: H1 "AI Employees for human-level roles" and the promise that Jarmin "learns the job and your standards, works in the tools you already use, and earns more responsibility over time"; seven roles listed (ML Engineer, Financial Analyst, Account Executive, Compliance Analyst, Paralegal, Data Analyst, Project Manager) plus "Build your own", with Data Analyst and Project Manager marked "book a call to set up" and no page of their own; the only proof is "Backed by Y Combinator. Built by engineers from Meta, Apple, AWS, Lockheed Martin, and JPMorganChase."; no customer names, testimonials, results or sample work; the only call to action is booking a call on the CEO's calendar.Source
    • Your websiteSitemap (28 URLs, checked November 9, 2026, no lastmod dates): 5 role pages, 3 industry pages (fintech, investment banking, legal), 1 partner page (/partnerships/mercury), 7 careers pages, a newsletter page and a manifesto; no blog, case study, FAQ, comparison or customer pages.Source
  • 08

    None of Jarmin's role or industry pages has an FAQ. Buyers in law, banking and compliance ask the same questions before they will book a call: who catches mistakes, what the AI employee is not allowed to do, where the data goes, what it costs and how long setup takes. Today every page answers these only in scattered fragments.

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

    First action

    Write the eight answers for the Compliance Analyst page.

    How to measure it

    Role-page request rate

    Now
    –
    Target
    +15%
    By
    60 days

    Steps

    1. 1

      Collect questions

      From the last 20 sales calls.

      Founding Account Executive 2 days

    2. 2

      Write answers

      Eight per role, shared across similar roles.

      Co-Founder & CEO 1 week

    3. 3

      Check claims

      Data and control answers checked against the Trust Center.

      Co-Founder & CTO 1 day

    4. 4

      Publish

      On all role and industry pages.

      Member of Technical Staff (Product) 2 days

    5. 5

      Add markup

      FAQ structured data and Markdown.

      Member of Technical Staff (Product) 1 day

    6. 6

      Refresh

      New questions from calls each month.

      Founding Account Executive monthly

    First result
    1 month
    Cost
    $0
    Owner
    Co Founder & CEO

    Smallest test

    Add the FAQ to one page and compare requests for a month.

    Why it works

    Add an eight-question FAQ to each role page: what it does on day one, what it escalates, how review works, what happens when it makes a mistake, data and training, systems access, cost basis and time to start. Mark it up with FAQ structured data and add it to the Markdown copies.

    Answering objections on the page removes them before the call, and question-and-answer content is what search engines and AI assistants quote.

    Risks and how to handle them

    • Overpromising

      Keep the "supports, does not sign" wording.

    Evidence

    • Your websiteFinancial Analyst, ML Engineer, Account Executive and Compliance Analyst role pages: detailed tasks and named integrations (Excel, PowerPoint and data rooms; warehouses, feature stores and model registries; Salesforce, HubSpot, Gong and Outreach; Socure, Persona, Alloy, Middesk, World-Check, NICE Actimize, Verafin and Unit21); every page says customer data stays in the customer's environment and is not used to train shared models; none has a case study, metric, price or FAQ.Source
    • Your websiteParalegal role page: drafts pleadings, motions and agreements, writes cited research memos after a citator check, builds 50-state surveys, reviews contracts against the playbook, drafts privilege logs and assembles Bates-ordered productions; works with Westlaw, Lexis, Relativity and Everlaw; "Each claim includes its source (case, clause, document, and page)"; it "supports the lawyer, it is not one". No case studies, metrics, testimonials, pricing or FAQ.Source
    • Your websiteSitemap (28 URLs, checked November 9, 2026, no lastmod dates): 5 role pages, 3 industry pages (fintech, investment banking, legal), 1 partner page (/partnerships/mercury), 7 careers pages, a newsletter page and a manifesto; no blog, case study, FAQ, comparison or customer pages.Source
  • 09

    Jarmin is hiring a Member of Technical Staff to "Own how Jarmin measures AI Employee performance on the job", with scorers, regression alerts and definitions of good, degraded and failed runs. Rivals win regulated buyers with numbers: Bretton publishes an 87% cut in review time and Rogo publishes its customer counts. Jarmin's buyers want to know one thing above all: how often the work is right, and what happens when it isn't.

    High impactMedium effort First result in 1 quarter Member of Technical Staff

    First action

    Agree the three metrics that will be public for the Compliance Analyst role.

    How to measure it

    Requests from scorecard visitors

    Now
    –
    Target
    10% of scorecard visitors
    By
    2 quarters

    Steps

    1. 1

      Pick public metrics

      Acceptance, escalation and turnaround.

      Member of Technical Staff (Evals) 1 week

    2. 2

      Get consent

      Aggregate-use permission from customers.

      Co-Founder & CEO 2 weeks

    3. 3

      Build the pipeline

      Monthly export from the eval system.

      Member of Technical Staff (Evals) 3 weeks

    4. 4

      Design the page

      Numbers, method and error types.

      Member of Technical Staff (Product) 1 week

    5. 5

      Publish

      Linked from every role page.

      Member of Technical Staff (Product) 1 day

    6. 6

      Update monthly

      With a changelog.

      Member of Technical Staff (Evals) monthly

    First result
    1 quarter
    Cost
    Engineering time
    Owner
    Member of Technical Staff (Evals)

    Smallest test

    Publish one role's acceptance rate for a single month.

    Why it works

    Turn the internal evals into a public scorecard for each role, updated monthly: share of outputs accepted without change by the human reviewer, escalation rate, average turnaround and the main error types with how each is caught. Publish the method alongside the numbers, use aggregate data only, and link the scorecard from every role page.

    A dated, repeatable number is stronger than any adjective, and it turns the evals team's daily work into sales material.

    Risks and how to handle them

    • Numbers look weak early

      Show the trend and publish the roles that are ready first.

    Evidence

    • HiringMember of Technical Staff (Evals): will "Own how Jarmin measures AI Employee performance on the job", building datasets, offline replay, scorers and regression alerts, defining what good, degraded and failed runs look like, so AI Employees get "measurably better over time".Source
    • Competitor researchRivals' own homepages: Bretton ("Your AI compliance team") pairs AI agents with experienced investigators and shows outcomes such as an "87%" cut in review time at an FDIC-insured bank, with named bank and fintech testimonials; Artisan ("Hire Ava, the autonomous AI BDR") sells AI employees for go-to-market, with a demo request as the next step.Source
    • Competitor researchRivals' own homepages: Harvey shows 3,500+ legal organizations, 80+ AmLaw 100 firms, customer logos and case studies; Rogo shows 450+ financial institutions, 60,000 bankers and investors, testimonials from named banks and three case studies.Source
  • 10

    The Product hire will own the surfaces customers use to "hire, direct, and trust their AI Employees", including reviewing, approving and correcting agent work. The Enterprise plan already promises "a complete record of every action" and a dedicated CSM. That record is the raw material for the moment a new customer decides whether the hire worked.

    High impactMedium effort First result in 6 weeks Member of Technical Staff

    First action

    Build the week-one report by hand for the newest customer.

    How to measure it

    90-day retention

    Now
    –
    Target
    90%+
    By
    2 quarters

    Steps

    1. 1

      Draft by hand

      One report from the action record.

      Co-Founder & CEO 2 days

    2. 2

      Pick the fields

      Tasks, escalations, corrections and time.

      Member of Technical Staff (Product) 2 days

    3. 3

      Automate

      Generate it from the action log.

      Member of Technical Staff (Product) 3 weeks

    4. 4

      Add the baseline

      Human turnaround agreed at kickoff.

      Member of Technical Staff (Evals) 1 week

    5. 5

      Review with the buyer

      CSM walkthrough and next scope.

      Co-Founder & CEO weekly

    6. 6

      Track

      Retention by report score.

      Member of Technical Staff (Evals) monthly

    First result
    6 weeks
    Cost
    Engineering time
    Owner
    Member of Technical Staff (Product)

    Smallest test

    Send one hand-built report this week.

    Why it works

    Generate a one-page report at the end of week one, and weekly after that: tasks completed, items escalated, reviewer corrections and how they were learned, turnaround against the human baseline and estimated hours handed back. The CSM walks through it with the buyer and the reviewer, and agrees what the AI employee takes on next.

    Early, visible value is the strongest driver of renewal. The same report later becomes the case study and the basis for expansion.

    Risks and how to handle them

    • Hours-saved estimates questioned

      Use the baseline the customer agreed at kickoff.

    Evidence

    • HiringMember of Technical Staff (Product): owns "the surfaces customers use to hire, direct, and trust their AI Employees", including interfaces for "reviewing, approving, and correcting agent work" and tools that make complex agent output understandable.Source
    • Your websitePricing page: "Pricing is not published." One Enterprise Plan for a Jarmin AI Employee: customized for the company and role, connected to its systems and data, "a complete record of every action", single sign-on and role-based access control, integration support, a dedicated Customer Success Manager and "Complimentary AI advisory sessions"; the next step is a call or an email to sales.Source
    • Your websiteHomepage: H1 "AI Employees for human-level roles" and the promise that Jarmin "learns the job and your standards, works in the tools you already use, and earns more responsibility over time"; seven roles listed (ML Engineer, Financial Analyst, Account Executive, Compliance Analyst, Paralegal, Data Analyst, Project Manager) plus "Build your own", with Data Analyst and Project Manager marked "book a call to set up" and no page of their own; the only proof is "Backed by Y Combinator. Built by engineers from Meta, Apple, AWS, Lockheed Martin, and JPMorganChase."; no customer names, testimonials, results or sample work; the only call to action is booking a call on the CEO's calendar.Source
  • 11

    Jarmin's only discount today is "25% off" the first month for Mercury customers. Otherwise, every deal starts with a custom Enterprise quote. Fintech teams have measurable queues (alerts, KYC refreshes, disputes) and legal teams have contract backlogs, so a pilot on one of those can be judged in days.

    High impactLow effort First result in 1 month Founding Account Executive

    First action

    Write the pilot one-pager for the alert-triage backlog.

    How to measure it

    Pilot conversion

    Now
    –
    Target
    60%
    By
    2 quarters

    Steps

    1. 1

      Define pilots

      One per industry, with volume and fee.

      Co-Founder & CEO 3 days

    2. 2

      Set criteria

      Accuracy, turnaround and escalation.

      Member of Technical Staff (Evals) 2 days

    3. 3

      Write the one-pager

      Scope, price, credit and next step.

      Founding Account Executive 1 day

    4. 4

      Add to pages

      Pricing and industry pages.

      Member of Technical Staff (Product) 1 day

    5. 5

      Sell

      Lead offer in the outbound plays.

      Founding Account Executive ongoing

    6. 6

      Review

      Pass rates and conversion.

      Founding Account Executive monthly

    First result
    1 month
    Cost
    Onboarding time per pilot
    Owner
    Founding Account Executive

    Smallest test

    Offer the alert-triage pilot to ten fintech prospects.

    Why it works

    Offer a two-week paid pilot on one named backlog, such as 500 transaction-monitoring alerts or 50 NDAs against the playbook, at a fixed fee credited to the first month. Agree acceptance criteria up front (accuracy against reviewer, turnaround, escalation quality) and convert to a monthly plan on pass.

    A fixed scope and price removes procurement friction, and pass/fail criteria turn the buyer's doubt into a test Jarmin can win.

    Risks and how to handle them

    • Integration takes longer than two weeks

      Start pilots on exported data where possible.

    Evidence

    • Your websiteMercury partner page: "25% off" the first month of the first AI employee on the Enterprise plan, "exclusively for Mercury customers", with free "white-glove onboarding and support"; it has a "Hire your first Jarmin (AI) employee" form asking for work email, name, company, phone, company headcount, department, responsibilities and how the visitor heard about Jarmin; Jarmin will "assess every request to determine whether Jarmin is a good fit".Source
    • Your websiteFintech page: for neobanks, payments and PSPs, lending and BNPL, crypto, and BaaS and sponsor-bank programs; six roles (Compliance Analyst, KYC and Onboarding Operations, Fraud and Disputes Operations, Customer Support, Underwriting and Credit Operations, Finance and Reconciliation Operations); "Verifiable and consistent, not a black box"; for sponsor-bank programs, first-line KYC and alert triage are "handed off as documented cases"; no metrics, case studies or customer names.Source
    • Your websitePricing page: "Pricing is not published." One Enterprise Plan for a Jarmin AI Employee: customized for the company and role, connected to its systems and data, "a complete record of every action", single sign-on and role-based access control, integration support, a dedicated Customer Success Manager and "Complimentary AI advisory sessions"; the next step is a call or an email to sales.Source
  • 12

    The Mercury page is Jarmin's only partner channel: a dedicated offer, a form and a fit check. The role pages name more than 20 systems the AI employees work in, from Unit21, Alloy and Middesk to Ironclad, iManage and Salesforce. The vendors behind those systems talk to exactly the teams Jarmin wants to reach.

    Medium impactMedium effort First result in 1 quarter Co Founder & CEO

    First action

    Measure how many requests the Mercury page has produced so far.

    How to measure it

    Partner-sourced requests

    Now
    –
    Target
    20% of all requests
    By
    2 quarters

    Steps

    1. 1

      Measure Mercury

      Requests and deals from the page so far.

      Founding Account Executive 1 day

    2. 2

      Pick five platforms

      Where customers already connect Jarmin.

      Co-Founder & CEO 2 days

    3. 3

      Pitch

      Co-marketing or directory listing.

      Co-Founder & CEO 1 month

    4. 4

      Build pages

      Mercury template with partner-specific offer.

      Member of Technical Staff (Product) 1 day each

    5. 5

      Joint content

      One webinar or post per partner.

      Co-Founder & CEO 1 quarter

    6. 6

      Track

      Requests by partner.

      Founding Account Executive monthly

    First result
    1 quarter
    Cost
    Founder time and partner offers
    Owner
    Co Founder & CEO

    Smallest test

    Pitch one compliance-platform partner this month.

    Why it works

    Pick five platforms where Jarmin already works with customer data, and ask each for a co-marketing slot or a partner-directory listing. Build one /partnerships page per partner on the Mercury template, with an offer for that partner's customers and a sample that uses their system. Track each page's requests separately.

    Partners lend trust and bring a pre-qualified audience. Each page also targets the "[platform] + AI analyst" searches buyers run.

    Risks and how to handle them

    • Partners slow to respond

      Start with platforms where a shared customer can make the introduction.

    Evidence

    • Your websiteMercury partner page: "25% off" the first month of the first AI employee on the Enterprise plan, "exclusively for Mercury customers", with free "white-glove onboarding and support"; it has a "Hire your first Jarmin (AI) employee" form asking for work email, name, company, phone, company headcount, department, responsibilities and how the visitor heard about Jarmin; Jarmin will "assess every request to determine whether Jarmin is a good fit".Source
    • Your websiteFinancial Analyst, ML Engineer, Account Executive and Compliance Analyst role pages: detailed tasks and named integrations (Excel, PowerPoint and data rooms; warehouses, feature stores and model registries; Salesforce, HubSpot, Gong and Outreach; Socure, Persona, Alloy, Middesk, World-Check, NICE Actimize, Verafin and Unit21); every page says customer data stays in the customer's environment and is not used to train shared models; none has a case study, metric, price or FAQ.Source
    • Your websiteLegal industry page: for in-house legal teams and law firms; six roles (Paralegal, Legal Assistant, Contracts Specialist, Legal Operations Analyst, Compliance Analyst, eDiscovery / Litigation Support); tools named include Ironclad, SpotDraft, iManage, NetDocuments, Westlaw, Lexis, SimpleLegal and Brightflag; promises paralegal work "overnight" and contract turnaround "the same day, not the same week"; no case studies, testimonials, results or FAQ.Source
  • 13

    Jarmin launched on YC as an "autonomous ML engineer that ships production AI/ML features", with use cases in semiconductor defect detection, biotech microbe analysis and recommendation systems. The manifesto is still entirely about ML talent. Yet the site's industry pages cover only fintech, investment banking and legal, so ML buyers land on a homepage now led by paralegals and analysts, and the manifesto and homepage tell two different stories.

    Medium impactLow effort First result in 1 month Co Founder & Head of ML

    First action

    Draft the manufacturing-quality page from the launch post.

    How to measure it

    ML role requests

    Now
    –
    Target
    +30%
    By
    2 quarters

    Steps

    1. 1

      Pick three use cases

      From the launch post and current deals.

      Co-Founder & Head of ML 1 day

    2. 2

      Write the pages

      Problem, workflow, systems and controls.

      Co-Founder & Head of ML 1 week

    3. 3

      Add samples

      Public-dataset work from idea 1.

      Co-Founder & Head of ML 1 week

    4. 4

      Update the manifesto

      Tie ML talent to the role thesis.

      Co-Founder & CEO 2 days

    5. 5

      Publish

      Pages, Markdown and sitemap.

      Member of Technical Staff (Product) 1 day

    6. 6

      Promote

      Share with the launch-post audience.

      Co-Founder & CEO 1 week

    First result
    1 month
    Cost
    Staff time only
    Owner
    Co Founder & Head of ML

    Smallest test

    Publish one ML use-case page.

    Why it works

    Add industry pages for the ML use cases from the launch post (manufacturing quality, life sciences, marketplaces and recommendations), each linking to the ML Engineer role and a sample from idea 1. Rewrite the manifesto so the ML-talent argument becomes one example of the broader "human-level roles" thesis.

    It keeps the founders' strongest credibility (ML) working for the business while the new roles build proof, and it gives the original launch audience somewhere to land.

    Risks and how to handle them

    • Diluted focus

      Keep ML as one industry among four, not the headline.

    Evidence

    • Launch postYC launch post ("You can now hire 24/7 Machine Learning Engineer employees"): an "autonomous ML engineer that ships production AI/ML features"; hiring ML engineers takes 3-6+ months at $200K+ a year; example uses include semiconductor defect detection, biotech microbe analysis and recommendation systems; no customer results. Jarmin's manifesto is still about ML talent ("two brutal constraints: compute and talent").Source
    • Your websiteSitemap (28 URLs, checked November 9, 2026, no lastmod dates): 5 role pages, 3 industry pages (fintech, investment banking, legal), 1 partner page (/partnerships/mercury), 7 careers pages, a newsletter page and a manifesto; no blog, case study, FAQ, comparison or customer pages.Source
    • ResearchAutomated homepage scan (November 2026): it counted six "notable customers" (Y Combinator, Meta, Apple, AWS, Lockheed Martin, JPMorganChase), which are in fact the investor and the founders' former employers; 0 testimonials and no case studies found; the headline changed from "24/7 ML Eng employees" to "AI Employees, Human-level roles".Source
  • 14

    Jarmin's buyers have three alternatives: hire a person (3-6+ months at $200K+ for an ML engineer, according to Jarmin's launch post), outsource, or buy software their own staff operate, such as Harvey and Rogo. Bretton sells AI agents plus human investigators. The site has no comparison pages, so buyers make that comparison without Jarmin in the room.

    Medium impactMedium effort First result in 2 months Co Founder & CEO

    First action

    Write the paralegal comparison table.

    How to measure it

    Comparison-page requests

    Now
    –
    Target
    15 a month
    By
    2 quarters

    Steps

    1. 1

      Pick the three

      Most common alternatives in sales calls.

      Founding Account Executive 1 day

    2. 2

      Gather facts

      Hiring costs and rivals' public positioning.

      Co-Founder & CEO 3 days

    3. 3

      Write

      Tables plus "when to choose the other option".

      Co-Founder & CEO 1 week

    4. 4

      Legal check

      Claims about named rivals.

      Co-Founder & CEO 2 days

    5. 5

      Publish

      Linked from role pages.

      Member of Technical Staff (Product) 1 day

    6. 6

      Refresh

      Quarterly with dates.

      Co-Founder & CEO quarterly

    First result
    2 months
    Cost
    Staff time only
    Owner
    Co Founder & CEO

    Smallest test

    Publish the paralegal comparison.

    Why it works

    Write three category pages: "AI paralegal vs hiring a paralegal", "AI compliance analyst vs outsourced alert review" and "AI financial analyst vs an AI research tool". Each compares cost, ramp time, who owns the output, review and data controls, and when the other option is the better choice. Keep them factual and dated.

    Comparison searches signal high intent, and a fair page that admits trade-offs earns trust with careful buyers.

    Risks and how to handle them

    • Outdated rival claims

      Compare categories, date the pages and cite sources.

    Evidence

    • Launch postYC launch post ("You can now hire 24/7 Machine Learning Engineer employees"): an "autonomous ML engineer that ships production AI/ML features"; hiring ML engineers takes 3-6+ months at $200K+ a year; example uses include semiconductor defect detection, biotech microbe analysis and recommendation systems; no customer results. Jarmin's manifesto is still about ML talent ("two brutal constraints: compute and talent").Source
    • Competitor researchRivals' own homepages: Harvey shows 3,500+ legal organizations, 80+ AmLaw 100 firms, customer logos and case studies; Rogo shows 450+ financial institutions, 60,000 bankers and investors, testimonials from named banks and three case studies.Source
    • Competitor researchRivals' own homepages: Bretton ("Your AI compliance team") pairs AI agents with experienced investigators and shows outcomes such as an "87%" cut in review time at an FDIC-insured bank, with named bank and fintech testimonials; Artisan ("Hire Ava, the autonomous AI BDR") sells AI employees for go-to-market, with a demo request as the next step.Source
    • Your websiteSitemap (28 URLs, checked November 9, 2026, no lastmod dates): 5 role pages, 3 industry pages (fintech, investment banking, legal), 1 partner page (/partnerships/mercury), 7 careers pages, a newsletter page and a manifesto; no blog, case study, FAQ, comparison or customer pages.Source
  • 15

    The homepage promises that Jarmin "earns more responsibility over time", and the plan includes a dedicated CSM and "Complimentary AI advisory sessions". Nothing describes how that responsibility grows, or how a customer goes from one AI employee to a second.

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

    First action

    Write the three levels for the Compliance Analyst role.

    How to measure it

    Customers with a second role or level

    Now
    –
    Target
    30%
    By
    2 quarters

    Steps

    1. 1

      Define levels

      Three per role, with promotion evidence.

      Co-Founder & CEO 1 week

    2. 2

      Show it in product

      Current level and progress to the next.

      Member of Technical Staff (Product) 3 weeks

    3. 3

      Run reviews

      Monthly advisory session as promotion review.

      Co-Founder & CEO monthly

    4. 4

      Map adjacent roles

      The natural second hire for each role.

      Founding Account Executive 2 days

    5. 5

      Offer

      Second role at level three.

      Founding Account Executive ongoing

    6. 6

      Track

      Promotions and expansion revenue.

      Member of Technical Staff (Evals) monthly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Co Founder & CEO

    Smallest test

    Run one promotion review with the longest-standing customer.

    Why it works

    Define three levels for each role (drafts for review, owns routine items, owns the queue with spot checks), each with the evidence needed to move up. Use the advisory session as a monthly promotion review built on the first-week report data, and offer the adjacent role (Paralegal to Contracts Specialist, Compliance Analyst to KYC Operations) once a role reaches level three.

    A clear path turns trust into scope, and scope into revenue, without a new sales cycle.

    Risks and how to handle them

    • Promotion before it is earned

      Tie levels to eval thresholds.

    Builds on

    Evidence

    • Your websiteHomepage: H1 "AI Employees for human-level roles" and the promise that Jarmin "learns the job and your standards, works in the tools you already use, and earns more responsibility over time"; seven roles listed (ML Engineer, Financial Analyst, Account Executive, Compliance Analyst, Paralegal, Data Analyst, Project Manager) plus "Build your own", with Data Analyst and Project Manager marked "book a call to set up" and no page of their own; the only proof is "Backed by Y Combinator. Built by engineers from Meta, Apple, AWS, Lockheed Martin, and JPMorganChase."; no customer names, testimonials, results or sample work; the only call to action is booking a call on the CEO's calendar.Source
    • Your websitePricing page: "Pricing is not published." One Enterprise Plan for a Jarmin AI Employee: customized for the company and role, connected to its systems and data, "a complete record of every action", single sign-on and role-based access control, integration support, a dedicated Customer Success Manager and "Complimentary AI advisory sessions"; the next step is a call or an email to sales.Source
    • HiringMember of Technical Staff (Product): owns "the surfaces customers use to hire, direct, and trust their AI Employees", including interfaces for "reviewing, approving, and correcting agent work" and tools that make complex agent output understandable.Source
  • 16

    Every Jarmin role keeps a human in the loop: the attorney who signs, the banker who checks the model, the BSA officer who decides on a filing. Those reviewers see the quality of the work every day, and they know peers at other firms facing the same volume. The Founding Account Executive is meant to build pipeline through referrals, among other channels, but there is no referral path today.

    Medium impactLow effort First result in 1 month Founding Account Executive

    First action

    Ask the reviewer at the happiest current customer for one introduction.

    How to measure it

    Meetings from reviewer referrals

    Now
    –
    Target
    5 a month
    By
    2 quarters

    Steps

    1. 1

      Pick the moment

      After a strong first-week report.

      Founding Account Executive 1 day

    2. 2

      Get buyer approval

      Agree the ask with the buyer first.

      Founding Account Executive ongoing

    3. 3

      Write the note

      Forwardable, with a sample link.

      Founding Account Executive 1 day

    4. 4

      Set the credit

      Referral credit for the customer.

      Co-Founder & CEO 1 day

    5. 5

      Ask

      Two introductions per reviewer.

      Founding Account Executive ongoing

    6. 6

      Track

      Referrals and closes by role.

      Founding Account Executive monthly

    First result
    1 month
    Cost
    Referral credits
    Owner
    Founding Account Executive

    Smallest test

    Ask three reviewers this month.

    Why it works

    Once a customer's first-week report shows good acceptance, ask the reviewer (with the buyer's agreement) for two introductions to peers. Give a referral credit to the customer and a simple forwardable note, built around a sample from idea 1, for the reviewer to send. Track referrals by role.

    Professionals trust peers who have checked the work themselves more than any vendor claim, and regulated work spreads through professional networks.

    Risks and how to handle them

    • Confidentiality concerns

      The note shares only public sample work, never client output.

    Builds on

    Evidence

    • HiringFounding Account Executive, the first sales hire: works with the founders to find target companies and buyers, builds pipeline from scratch through outbound email, calls, events and referrals, runs founder-led deals through close, tests messaging and positioning, sets up prospecting, CRM and reporting, and turns what works into a repeatable playbook.Source
    • Your websiteParalegal role page: drafts pleadings, motions and agreements, writes cited research memos after a citator check, builds 50-state surveys, reviews contracts against the playbook, drafts privilege logs and assembles Bates-ordered productions; works with Westlaw, Lexis, Relativity and Everlaw; "Each claim includes its source (case, clause, document, and page)"; it "supports the lawyer, it is not one". No case studies, metrics, testimonials, pricing or FAQ.Source
    • HiringMember of Technical Staff (Product): owns "the surfaces customers use to hire, direct, and trust their AI Employees", including interfaces for "reviewing, approving, and correcting agent work" and tools that make complex agent output understandable.Source
  • 17

    Every role page promises that data stays in the customer's environment and is never used to train shared models, and the Contact page sends vendor reviews to the Trust Center. The Trust Center shows SOC 2 Type 2 as compliant, with 25 policies and 35 controls listed, but no documents available publicly. Bank, law-firm and fintech buyers run long security reviews, and rivals such as Rogo list several certifications on their homepage.

    High impactMedium effort First result in 1 month Co Founder & CTO

    First action

    Complete one standard security questionnaire and upload it to the Trust Center.

    How to measure it

    Security review time

    Now
    –
    Target
    -30%
    By
    2 quarters

    Steps

    1. 1

      Fill the questionnaire

      One standard format, answered once.

      Co-Founder & CTO 1 week

    2. 2

      Draw the data flow

      Customer environment, access and logging.

      Co-Founder & CTO 3 days

    3. 3

      Write the terms

      No-training and no-retention clauses.

      Co-Founder & CEO 1 week

    4. 4

      Upload

      Request-access documents in the Trust Center.

      Co-Founder & CTO 1 day

    5. 5

      Link

      From each role page's data promise.

      Member of Technical Staff (Product) 1 day

    6. 6

      Track

      Review time per deal.

      Founding Account Executive monthly

    First result
    1 month
    Cost
    Staff time; legal review of terms
    Owner
    Co Founder & CTO

    Smallest test

    Send the pack to the next buyer in security review.

    Why it works

    Put a request-access pack in the Trust Center: the SOC 2 report under NDA, a completed standard security questionnaire, a data-flow diagram for an AI employee running in a customer environment, the subprocessor list, and contract language for no training and no retention. Link the pack from every role page next to the data promise.

    Shortening security review shortens the sales cycle, and turning the no-training promise into contract language makes it believable.

    Risks and how to handle them

    • Questionnaires vary by buyer

      Start with the most requested format and reuse answers.

    Evidence

    • Your websiteTrust Center (powered by Comp AI): SOC 2 Type 2 shown as "Compliant", with 25 policies and 35 controls listed; no documents available publicly.Source
    • Your websiteFinancial Analyst, ML Engineer, Account Executive and Compliance Analyst role pages: detailed tasks and named integrations (Excel, PowerPoint and data rooms; warehouses, feature stores and model registries; Salesforce, HubSpot, Gong and Outreach; Socure, Persona, Alloy, Middesk, World-Check, NICE Actimize, Verafin and Unit21); every page says customer data stays in the customer's environment and is not used to train shared models; none has a case study, metric, price or FAQ.Source
    • Your websiteContact page: no form; a book-a-call link plus six email addresses by purpose (sales handles "pricing, procurement, and role costs"); vendor reviews go to the Trust Center; no response times. The newsletter page says "We send infrequently and only when there is something worth reading" and shows no past issues.Source
    • Competitor researchRivals' own homepages: Harvey shows 3,500+ legal organizations, 80+ AmLaw 100 firms, customer logos and case studies; Rogo shows 450+ financial institutions, 60,000 bankers and investors, testimonials from named banks and three case studies.Source
  • 18

    Jarmin has a newsletter page that promises to send "only when there is something worth reading", and it shows no past issues. The sitemap has no blog, and the press page has no releases or coverage. A company that does real work in law, deals and compliance every day has no channel to show it.

    Medium impactLow effort First result in 2 months Co Founder & CEO

    First action

    Write the first note on the compliance alert queue.

    How to measure it

    Subscribers

    Now
    –
    Target
    500
    By
    2 quarters

    Steps

    1. 1

      Set the format

      One role, one job, 500 words.

      Co-Founder & CEO 1 day

    2. 2

      Clear the story

      Customer approval and anonymisation.

      Co-Founder & CEO 1 week

    3. 3

      Write

      Monthly field note.

      Co-Founder & CEO monthly

    4. 4

      Publish on site

      A /notes archive in the sitemap.

      Member of Technical Staff (Product) 2 days

    5. 5

      Distribute

      LinkedIn, X and outbound follow-ups.

      Founding Account Executive monthly

    6. 6

      Measure

      Subscribers and requests per issue.

      Founding Account Executive monthly

    First result
    2 months
    Cost
    Writing time
    Owner
    Co Founder & CEO

    Smallest test

    Send one note to the existing list.

    Why it works

    Send one short field note a month and publish it on the site: one role, one anonymised piece of work (the queue, what the AI employee did, what the reviewer corrected, what it learned), plus one public sample. Rotate through the industries and end each note with the intake form.

    Specific stories about the work build an audience and search footprint, and they give the Founding AE something to send each month.

    Risks and how to handle them

    • Customer data in stories

      Anonymise and get approval for each note.

    Evidence

    • Your websiteContact page: no form; a book-a-call link plus six email addresses by purpose (sales handles "pricing, procurement, and role costs"); vendor reviews go to the Trust Center; no response times. The newsletter page says "We send infrequently and only when there is something worth reading" and shows no past issues.Source
    • Your websiteSitemap (28 URLs, checked November 9, 2026, no lastmod dates): 5 role pages, 3 industry pages (fintech, investment banking, legal), 1 partner page (/partnerships/mercury), 7 careers pages, a newsletter page and a manifesto; no blog, case study, FAQ, comparison or customer pages.Source
    • Your websitePress and About pages: founded September 2025 in San Francisco, Y Combinator F25; three co-founders (CEO, CTO, Head of ML) who worked at Meta Superintelligence Labs, Apple, AWS, Lockheed Martin and JPMorganChase; no press releases, coverage, funding amounts or customer numbers.Source

Big bets 3

Larger plays that can change the trajectory.

  • 19

    The homepage offers "Build your own: describe the job and the work it owns", and the Mercury form already asks for responsibilities, department and headcount. Today each answer becomes a call with a founder. With a team of five, founder calls are the limit on how many roles Jarmin can scope.

    High impactHigh effort First result in 1 quarter Member of Technical Staff

    First action

    Run the last ten intake answers through a draft scoping prompt and compare with what the founders actually proposed.

    How to measure it

    Plan-to-deal conversion

    Now
    –
    Target
    10%
    By
    3 quarters

    Steps

    1. 1

      Back-test

      Ten past intakes against founder proposals.

      Member of Technical Staff (Agents) 2 weeks

    2. 2

      Define the plan format

      Tasks, escalations, systems, ramp and price.

      Co-Founder & CEO 1 week

    3. 3

      Build the tool

      Paste-in input, plan output.

      Member of Technical Staff (Agents) 6 weeks

    4. 4

      Design the page

      Accept, adjust or book a call.

      Member of Technical Staff (Product) 2 weeks

    5. 5

      Human check

      A founder approves quotes in the first month.

      Co-Founder & CEO 1 month

    6. 6

      Measure

      Plans, accept rate and deals.

      Member of Technical Staff (Evals) monthly

    First result
    1 quarter
    Cost
    Engineering quarter
    Owner
    Member of Technical Staff (Agents)

    Smallest test

    Email a generated role plan to the next five intake requests.

    Why it works

    Let a visitor paste a job description or a list of responsibilities. Jarmin's own agent returns a role plan within minutes: which tasks it would own, what it would escalate, which systems it needs, a ramp schedule, and a price range or quote. The buyer can accept, adjust or book a call, and every plan feeds the CRM.

    It shows the product working before the sale, scales scoping without more founder time, and creates a self-serve path for smaller buyers.

    Risks and how to handle them

    • Bad quotes

      Show ranges and keep founder approval until the back-test is accurate.

    Builds on

    Evidence

    • Your websiteHomepage: H1 "AI Employees for human-level roles" and the promise that Jarmin "learns the job and your standards, works in the tools you already use, and earns more responsibility over time"; seven roles listed (ML Engineer, Financial Analyst, Account Executive, Compliance Analyst, Paralegal, Data Analyst, Project Manager) plus "Build your own", with Data Analyst and Project Manager marked "book a call to set up" and no page of their own; the only proof is "Backed by Y Combinator. Built by engineers from Meta, Apple, AWS, Lockheed Martin, and JPMorganChase."; no customer names, testimonials, results or sample work; the only call to action is booking a call on the CEO's calendar.Source
    • Your websiteMercury partner page: "25% off" the first month of the first AI employee on the Enterprise plan, "exclusively for Mercury customers", with free "white-glove onboarding and support"; it has a "Hire your first Jarmin (AI) employee" form asking for work email, name, company, phone, company headcount, department, responsibilities and how the visitor heard about Jarmin; Jarmin will "assess every request to determine whether Jarmin is a good fit".Source
    • HiringCareers page: seven open roles, applied for through YC's Work at a Startup: Member of Technical Staff, and Member of Technical Staff for Agents, Agent Infrastructure, Evals, Models and Product (in person, San Francisco), plus a Founding Account Executive (San Francisco or remote US). YC lists the engineering roles at $200K–$300K with 0.25%–1.00% equity and the team size as 5.Source
    • Your websitePricing page: "Pricing is not published." One Enterprise Plan for a Jarmin AI Employee: customized for the company and role, connected to its systems and data, "a complete record of every action", single sign-on and role-based access control, integration support, a dedicated Customer Success Manager and "Complimentary AI advisory sessions"; the next step is a call or an email to sales.Source
  • 20

    Many of Jarmin's fintech and compliance roles work countable queues: alerts triaged, KYC files refreshed, disputes built, sanctions hits cleared. Today the only option is a per-employee Enterprise quote. Buyers of compliance operations already think in cost per case, and Bretton sells on measured review time.

    High impactHigh effort First result in 1 quarter Co Founder & CEO

    First action

    Calculate cost per alert for one current customer's queue.

    How to measure it

    Queue-role revenue growth

    Now
    –
    Target
    +40%
    By
    3 quarters

    Steps

    1. 1

      Measure unit costs

      Cost per item for each queue role.

      Co-Founder & CTO 2 weeks

    2. 2

      Set prices

      Per item, floor and tiers.

      Co-Founder & CEO 1 week

    3. 3

      Meter

      Count items in the action record.

      Member of Technical Staff (Agent Infrastructure) 1 month

    4. 4

      Report

      Cost per item against the human baseline.

      Member of Technical Staff (Product) 2 weeks

    5. 5

      Pilot

      Three customers on the new model.

      Founding Account Executive 1 quarter

    6. 6

      Decide

      Roll out or revert by margin.

      Co-Founder & CEO 1 day

    First result
    1 quarter
    Cost
    Engineering for metering
    Owner
    Co Founder & CEO

    Smallest test

    Quote one prospect both ways.

    Why it works

    For queue roles, offer a price per item worked (for example, per alert dispositioned or per KYC review), with a monthly minimum and volume tiers. Report cost per item next to the customer's human baseline in the weekly report. Keep per-employee pricing for project roles such as ML engineering and deal work.

    Usage pricing matches how operations leaders budget, makes it easy to start small, and grows with the customer's volume without a new sale.

    Risks and how to handle them

    • Volatile revenue

      Monthly floor and annual volume commitments.

    Builds on

    Evidence

    • Your websiteFintech page: for neobanks, payments and PSPs, lending and BNPL, crypto, and BaaS and sponsor-bank programs; six roles (Compliance Analyst, KYC and Onboarding Operations, Fraud and Disputes Operations, Customer Support, Underwriting and Credit Operations, Finance and Reconciliation Operations); "Verifiable and consistent, not a black box"; for sponsor-bank programs, first-line KYC and alert triage are "handed off as documented cases"; no metrics, case studies or customer names.Source
    • Your websitePricing page: "Pricing is not published." One Enterprise Plan for a Jarmin AI Employee: customized for the company and role, connected to its systems and data, "a complete record of every action", single sign-on and role-based access control, integration support, a dedicated Customer Success Manager and "Complimentary AI advisory sessions"; the next step is a call or an email to sales.Source
    • Competitor researchRivals' own homepages: Bretton ("Your AI compliance team") pairs AI agents with experienced investigators and shows outcomes such as an "87%" cut in review time at an FDIC-insured bank, with named bank and fintech testimonials; Artisan ("Hire Ava, the autonomous AI BDR") sells AI employees for go-to-market, with a demo request as the next step.Source
  • 21

    The fintech page already names BaaS and sponsor-bank programs, where Jarmin hands off first-line KYC and alert triage "as documented cases", and every compliance decision comes with a rationale and an audit trail. A sponsor bank oversees many fintech programs, and each one has to show that its compliance work holds up. One bank relationship could reach many programs at once.

    High impactHigh effort First result in 2 quarters Co Founder & CEO

    First action

    Ask a current fintech customer what their sponsor bank asks to see each month.

    How to measure it

    Bank-referred programs

    Now
    –
    Target
    5
    By
    3 quarters

    Steps

    1. 1

      Learn the review

      What sponsor banks check in fintech programs.

      Co-Founder & CEO 1 month

    2. 2

      Standardise cases

      One case format matching that review.

      Member of Technical Staff (Agents) 1 month

    3. 3

      Build the dashboard

      Alert ageing, quality and escalations.

      Member of Technical Staff (Product) 1 month

    4. 4

      Pilot

      One fintech and its bank.

      Founding Account Executive 1 quarter

    5. 5

      Offer the bank

      Preferred-vendor terms for other programs.

      Co-Founder & CEO 1 quarter

    6. 6

      Measure

      Programs referred and retained.

      Founding Account Executive quarterly

    First result
    2 quarters
    Cost
    Founder time and product work
    Owner
    Co Founder & CEO

    Smallest test

    One meeting with a current customer's sponsor-bank oversight team.

    Why it works

    Build a sponsor-bank pack: a standard case format that matches what bank oversight teams review, a program-level dashboard of alert ageing and quality, and a reference design for a fintech's first line. Pilot it with one fintech and its sponsor bank, then offer the bank a preferred-vendor arrangement for its other programs.

    Selling to the party that judges compliance quality turns one sale into a channel, and documented cases are exactly what Jarmin already produces.

    Risks and how to handle them

    • Long bank cycles

      Start through a fintech customer that already works with Jarmin.

    • Regulatory sensitivity

      Keep decisions and sign-off with named humans, as the pages already state.

    Builds on

    Evidence

    • Your websiteFintech page: for neobanks, payments and PSPs, lending and BNPL, crypto, and BaaS and sponsor-bank programs; six roles (Compliance Analyst, KYC and Onboarding Operations, Fraud and Disputes Operations, Customer Support, Underwriting and Credit Operations, Finance and Reconciliation Operations); "Verifiable and consistent, not a black box"; for sponsor-bank programs, first-line KYC and alert triage are "handed off as documented cases"; no metrics, case studies or customer names.Source
    • Your websiteFinancial Analyst, ML Engineer, Account Executive and Compliance Analyst role pages: detailed tasks and named integrations (Excel, PowerPoint and data rooms; warehouses, feature stores and model registries; Salesforce, HubSpot, Gong and Outreach; Socure, Persona, Alloy, Middesk, World-Check, NICE Actimize, Verafin and Unit21); every page says customer data stays in the customer's environment and is not used to train shared models; none has a case study, metric, price or FAQ.Source
    • Competitor researchRivals' own homepages: Bretton ("Your AI compliance team") pairs AI agents with experienced investigators and shows outcomes such as an "87%" cut in review time at an FDIC-insured bank, with named bank and fintech testimonials; Artisan ("Hire Ava, the autonomous AI BDR") sells AI employees for go-to-market, with a demo request as the next step.Source

Wildcards 2

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

  • 22

    Jarmin sells an AI Account Executive that keeps the CRM current, preps calls, drafts follow-ups and builds sequences. It is also about to hire its first human AE to build pipeline from scratch and set up CRM and reporting. Few vendors can show their own product doing the job they sell, inside their own company.

    Medium impactMedium effort First result in 1 quarter Founding Account Executive

    First action

    Set up the AI Account Executive in Jarmin's own CRM before the new AE starts.

    How to measure it

    Hours handed back to the AE per week

    Now
    –
    Target
    10
    By
    1 quarter

    Steps

    1. 1

      Set up

      AI Account Executive in Jarmin's CRM and inbox.

      Co-Founder & CTO 1 week

    2. 2

      Set the baseline

      Hours per week on admin, before and after.

      Founding Account Executive 1 week

    3. 3

      Run

      Daily use for a quarter.

      Founding Account Executive 1 quarter

    4. 4

      Log

      Tasks, corrections and hours saved.

      Member of Technical Staff (Evals) weekly

    5. 5

      Write it up

      Case study with real numbers.

      Co-Founder & CEO 1 week

    6. 6

      Use it

      On the Account Executive page and in outbound.

      Founding Account Executive ongoing

    First result
    1 quarter
    Cost
    Internal usage
    Owner
    Founding Account Executive

    Smallest test

    Use it for call prep for the first month.

    Why it works

    From day one, pair the Founding AE with Jarmin's own AI Account Executive. Track what it handles (CRM updates, briefs, follow-ups, sequence steps), the hours it saves and the pipeline it touches. After a quarter, publish the results as Jarmin's first case study, with real numbers from its own sales team.

    It produces the first public proof without waiting for customer approval, and buyers can see the founders believe in the product.

    Risks and how to handle them

    • Seen as a vanity case

      Publish the corrections and limits as well as the wins.

    Evidence

    • HiringFounding Account Executive, the first sales hire: works with the founders to find target companies and buyers, builds pipeline from scratch through outbound email, calls, events and referrals, runs founder-led deals through close, tests messaging and positioning, sets up prospecting, CRM and reporting, and turns what works into a repeatable playbook.Source
    • Your websiteFinancial Analyst, ML Engineer, Account Executive and Compliance Analyst role pages: detailed tasks and named integrations (Excel, PowerPoint and data rooms; warehouses, feature stores and model registries; Salesforce, HubSpot, Gong and Outreach; Socure, Persona, Alloy, Middesk, World-Check, NICE Actimize, Verafin and Unit21); every page says customer data stays in the customer's environment and is not used to train shared models; none has a case study, metric, price or FAQ.Source
    • Your websiteHomepage: H1 "AI Employees for human-level roles" and the promise that Jarmin "learns the job and your standards, works in the tools you already use, and earns more responsibility over time"; seven roles listed (ML Engineer, Financial Analyst, Account Executive, Compliance Analyst, Paralegal, Data Analyst, Project Manager) plus "Build your own", with Data Analyst and Project Manager marked "book a call to set up" and no page of their own; the only proof is "Backed by Y Combinator. Built by engineers from Meta, Apple, AWS, Lockheed Martin, and JPMorganChase."; no customer names, testimonials, results or sample work; the only call to action is booking a call on the CEO's calendar.Source
  • 23

    Jarmin's site is fast: PageSpeed scores 99 on mobile and real visitors pass Core Web Vitals, with LCP at 970 ms. Two accessibility checks fail, though (low colour contrast and no main landmark), and fonts make up 345 KB of a 1,095 KB page. Enterprise and public-sector buyers sometimes run accessibility checks on vendors, and both fixes are small.

    Low impactLow effort First result in 1 week Member of Technical Staff

    First action

    Add the main landmark to the shared layout.

    How to measure it

    PageSpeed accessibility

    Now
    –
    Target
    100
    By
    2 weeks

    Steps

    1. 1

      Add the landmark

      Main element in the layout.

      Member of Technical Staff (Product) 1 hour

    2. 2

      Fix contrast

      Failing text colours.

      Member of Technical Staff (Product) 2 hours

    3. 3

      Subset fonts

      Only the weights in use.

      Member of Technical Staff (Product) 2 hours

    4. 4

      Self-host

      Serve fonts from the site's own host.

      Member of Technical Staff (Product) 1 hour

    5. 5

      Re-test

      PageSpeed mobile and desktop.

      Member of Technical Staff (Product) 1 hour

    6. 6

      Check templates

      Role and industry pages.

      Member of Technical Staff (Product) 1 hour

    First result
    1 week
    Cost
    $0
    Owner
    Member of Technical Staff (Product)

    Smallest test

    Fix the landmark and re-run PageSpeed.

    Why it works

    Raise contrast on the failing text, wrap page content in a main landmark, and self-host a subset of the font files with only the weights in use. Re-run PageSpeed to confirm accessibility moves from 93 toward 100 and page weight drops.

    Small fixes keep a polished site polished, and accessibility is part of looking enterprise-ready.

    Risks and how to handle them

    • Brand colours change

      Darken only the failing text.

    Evidence

    • PageSpeedGoogle PageSpeed (mobile 99, desktop 100): real visitors pass Core Web Vitals (LCP 970 ms at p75, 43% faster than before; INP 34 ms; CLS 0); fonts are 345 KB of a 1,095 KB page and third-party code 530 KB; accessibility 93, with two failures: low colour contrast and no main landmark.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 23 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 #6 (opens programmable.info in a new tab)
  • Build this list in Info for idea #14 (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.

  • InStride growth strategyA public benefit corporation that runs employer-sponsored education and skills programs: upfront tuition, Capability Accelerators, career…21 growth ideas
  • Zapier growth strategyA no-code AI orchestration platform: Zap workflows, Zapier MCP for AI chat tools, an SDK, Agents, Chatbots, Tables, Forms and Canvas…22 growth ideas
  • VTEX growth strategyA composable, AI-native commerce platform for B2C and B2B, with order management, marketplace, customer service and retail media (VTEX Ads).22 growth ideas
  • Babylist growth strategyA universal baby registry that grew into a full ecosystem: the Babylist Shop, Babylist Health (insurance-covered breast pumps and care)…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
  • Volka growth strategyA Limassol game studio of about 100 people that makes Taonga: the Island Farm (a browser game on its own portal and Facebook since 2016)…23 growth ideas
  • Hostinger growth strategyWeb hosting, domains, an AI website builder, business email, VPS and GPU hosting, AI Router and agent hosting, with Hostinger Agent as a…22 growth ideas
  • Conduct growth strategyAn AI operating system for enterprise software that reads SAP custom code, configuration and dependencies so IT teams can understand…21 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. Jarmin did not commission, review or endorse it, and Programmable is not affiliated with Jarmin. 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 Jarmin runs its business.

Something wrong, or do you represent Jarmin? 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 307 ideas

    Show sample work, open the intake form site-wide, publish a starting price, fix the proof line and llms.txt, give the Founding AE three outbound plays, and make the accessibility fixes.

    Programmable can ship 1 of these in days.

  2. Days 31 to 6012 ideas

    Role pages and FAQs, the quality scorecard, the first-week report, fixed-scope pilots, partner pages, the ML engineering page, comparisons, the responsibility ladder, reviewer referrals, the security pack and the newsletter.

  3. Days 61 to 904 ideas

    Commit to the big bets: self-serve role scoping, per-item pricing for queue roles and the sponsor-bank channel, plus the in-house AI Account Executive test.

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 work where every number and claim ties back to a source. The most cited problem: hiring specialists takes months and costs $200K+ a year.

  • Ramp-up in days instead of a months-long hire
  • Volume work (alerts, contracts, spreading, privilege logs) crowds out judgment work
  • In their words: “human-level roles”

Market

Competitive landscape

4 competitors compared: Harvey, Rogo, Bretton and Artisan.

  • Gap to exploit: One vendor for delegated roles across legal, deal and compliance work
  • Harvey: Software the lawyers operate themselves, not a delegated role
  • Rogo: A tool for bankers to query, not an analyst that owns the deliverable

Hiring signals

Jarmin, a team of five, is hiring its first salesperson and five specialist engineers.

  • Founding Account Executive: First sales hire; outbound, referrals and a repeatable playbook. Ideas 6, 16, 22.
  • Member of Technical Staff (Product): Surfaces to hire, direct and trust AI employees. Ideas 10, 15, 19.
  • Member of Technical Staff (Evals): Measures AI employee performance on the job. Ideas 9, 11.

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