ProgrammableMarketing/runtm.com (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. Programmable
  2. Playbooks
  3. Runtime
Runtime logo

growth playbook · Nov 10, 2026

runtm.com (opens in a new tab)

Runtime growth playbook

Diagnosis

Runtime is a 7-person YC Spring 2026 company that has repositioned from sandboxed coding agents to "the AI agent harness for payment teams". It ships fast: SSO, data masking, Teams bots and a printable per-agent evidence report all arrived in October. It has also built a large search footprint: 77 comparison pages, 110 guides (49 of them in Spanish or Portuguese) and an open-model directory.

The weak point is proof. The homepage has one real customer result (Rain, $750k saved) with no case study behind it, while the cost chart beside it is labelled "Figures are illustrative." No pricing or pilot terms are published, the LATAM guides send readers to an English-only demo page, and the blog and GitHub still describe the old coding-agent product.

This playbook fixes proof and paths first: a Rain case study, real numbers, a published pilot offer and a Spanish/Portuguese demo route. It then uses the two open hires (Founding AI Ops and Forward Deployed AI Engineer) to turn each deployment into reusable proof, templates and customer-built second agents. The bigger bets are a self-serve sandbox on synthetic payments data, outcome-based pricing and a public payments-ops benchmark.

Growth Score

6/10

Biggest Opportunity

Turn the deployments Runtime already runs for payments companies "that move hundreds of billions of dollars a year" into public, measured proof, and route the large SEO and LATAM content footprint to it.

Biggest Risk

Risk and compliance buyers read carefully. Illustrative numbers, past-employer logos and coding-agent leftovers can make a 7-person company look less proven than Taktile, Unit21 and Sardine, which all show named customers and measured results.

01

Start here

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

  1. 1

    Replace the homepage's "illustrative" cost chart with Rain's real numbers, and give the $750k result its own case study page

    Send Rain a one-page draft of the case study with the $750k breakdown and ask for approval.

    High impact First result in 2 weeks
  2. 2

    Turn "By request; free pilot" into a published pilot offer with scope, timeline and success measures

    Write the pilot terms on one page: scope, weeks, data handling, success measure and exit.

    High impact First result in 1 month
  3. 3

    Finish the LATAM path: Spanish and Portuguese calls to action and demo pages for the 49 local-language guides

    Translate the "Book a demo" block on the three Mexico guides and point it to a Spanish demo page.

    High impact First result in 1 month

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.

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

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 · 9Fill-ins · 7Reconsider · 4
Effort →

Showing 23 of 23 ideas

This week 6

Quick wins with a clear first action.

  • 01

    The homepage has one real customer result: Rain, "Running at the speed of a $1.95B stablecoin company", with "$750k" in "Vendor and salary spend saved". It has no link and no story behind it, and the sitemap has no customer or case study pages. Right beside it, the headline proof ("10×" more runs, "−93%" cost per run, $0.42 to $0.03 a run) carries the line "Figures are illustrative." A risk or compliance buyer who reads that line discounts the whole page. Rivals Taktile and Unit21 lead with named, measured case studies.

    High impactLow effort First result in 2 weeks CEO

    First action

    Send Rain a one-page draft of the case study with the $750k breakdown and ask for approval.

    How to measure it

    Homepage-to-demo conversion

    Now
    –
    Target
    +25%
    By
    60 days

    Steps

    1. 1

      Get approval

      Rain's sign-off on the name, numbers and quotes.

      CEO (co-founder) 1 week

    2. 2

      Pull the data

      Real runs per week, cost per run and steps per case from Rain's run records, with dates.

      Forward Deployed AI Engineer 2 days

    3. 3

      Write the story

      Problem, workflows, rollout, controls and the $750k math, with a customer quote.

      CEO (co-founder) 3 days

    4. 4

      Publish /customers

      An index page plus /customers/rain, added to the sitemap.

      Founding engineer 1 day

    5. 5

      Swap the chart

      Real figures with a measurement period instead of "Figures are illustrative."

      Founding engineer 1 day

    6. 6

      Link it everywhere

      The homepage Rain block, comparison pages and guides that cite Rain.

      Founding engineer 2 days

    7. 7

      Measure

      Case study views, then demo bookings from those sessions.

      Founding AI Ops monthly

    First result
    2 weeks
    Cost
    Staff time only
    Owner
    CEO (co founder)

    Smallest test

    Link the Rain block to a short, approved one-page story this week.

    Why it works

    Ask Rain to approve a full case study: the workflows it runs on Runtime (payment support, fraud and transaction monitoring, compliance, cybersecurity, payment ops), the before and after, how the $750k was calculated, and its real cost per run and runs per week over time. Publish it at /customers/rain. Replace the illustrative chart with Rain's actual curve, or with an anonymised aggregate across live customers, and state the measurement period. Link the Rain block, the comparison pages and the guides that already mention Rain to the new page. Add a /customers index so later stories have somewhere to go.

    Payments buyers are trained to distrust unverified numbers. One named, measured story backs up the claims every other page makes, and it matches what rivals show, at almost no cost, because the result already exists.

    Ready to use

    CopyRain case study outline

    1. Rain in one line: the card issuer and stablecoin company and its scale
    2. The queues before Runtime: support, monitoring, compliance, payment ops
    3. Why in its own cloud: data, approvals, audit trail
    4. Rollout: first SOP, first agent, weeks to production
    5. Results: $750k saved (how it was calculated), runs per week, cost per run over time
    6. Controls: approvals, masking, evidence reports
    7. Quote from an ops or risk leader
    8. What's next, then Book a demo

    Risks and how to handle them

    • Rain limits what can be named

      Publish the workflows and ratios, and keep exact volumes private if needed.

    • Real numbers look less dramatic than the illustration

      Show the trend and the measurement period; credible beats dramatic for this buyer.

    Evidence

    • Your websiteHomepage (checked November 9, 2026): one customer result, Rain, "Running at the speed of a $1.95B stablecoin company" with "$750k" "Vendor and salary spend saved", and no link to a case study; a cost block ("10×" more runs, "−93%" cost per run, "$0.42 / run" to "$0.03 / run", "−29%" total weekly spend) labelled "Figures are illustrative."; four customer logos (Rain, Nixtla, Mono, LineLeap); "Book a demo" is the only call to action and no pricing is shown.Source
    • Your websiteSitemap (214 URLs, checked November 9, 2026): 77 comparison pages, 110 guides (16 in Portuguese for Brazil and 33 in Spanish for Chile, Colombia, Mexico and LATAM), 17 open-source model pages, 11 solution pages, 4 blog posts and 2 monthly release Briefs; no customer, case study, pricing, about or careers pages.Source
    • Your websiteHomepage Rain block: "Rain runs Runtime for" technical payment support, fraud and risk transaction monitoring, compliance, cybersecurity and payment operations, and "Also" engineering investigations, PR reviews, QA testing and reporting.Source
    • Competitor researchRivals' own homepages: Taktile shows Allianz, Mercury and Monzo, "75% less time to onboard" for Faire and go-live in about 4 weeks for Zilch; Unit21 shows a dozen customer logos and case studies (Nexo: 57% of alert reviews automated) under "Most systems generate alerts. Unit21 resolves them."; Sardine cites a Forrester Wave Leader placement. None of the three publishes prices.Source
  • 02

    Runtime has written 33 Spanish guides (Chile, Colombia, Mexico, LATAM) and 16 Portuguese guides for Brazil, built around SPEI, CLABE, INE, UIF and PLD. A Mexican fraud lead who reads one in Spanish then meets an English "Book a demo" button and an English-only demo page, with no hreflang telling search engines which page serves which market. The CEO's previous company served Latin America's largest asset and wealth managers, so this market is an advantage the funnel doesn't yet use.

    High impactMedium effort First result in 1 month CEO

    First action

    Translate the "Book a demo" block on the three Mexico guides and point it to a Spanish demo page.

    How to measure it

    LATAM demo bookings

    Now
    –
    Target
    8 a month
    By
    90 days

    Steps

    1. 1

      Translate CTAs

      Navigation, buttons and closing blocks on the 49 guides.

      Founding engineer 2 days

    2. 2

      Build demo pages

      /es/demo and /pt/demo with a language choice for the call.

      Founding engineer 2 days

    3. 3

      Add hreflang

      es-MX, es-CO, es-CL, pt-BR and the html lang attribute.

      Founding engineer 1 day

    4. 4

      Add local proof

      A LATAM-relevant customer line or the founders' LATAM background.

      CEO (co-founder) 2 days

    5. 5

      Route bookings

      To a Spanish- or Portuguese-speaking founder or engineer.

      Founding AI Ops 1 day

    6. 6

      Track

      LATAM demos and pilots as their own funnel.

      Founding AI Ops monthly

    First result
    1 month
    Cost
    Translation review and staff time
    Owner
    CEO (co founder)

    Smallest test

    Spanish CTAs on the three Mexico guides for one month.

    Why it works

    Translate the call-to-action blocks on every Spanish and Portuguese guide, add /es/demo and /pt/demo pages that offer calls in Spanish or Portuguese, and add hreflang plus the correct html lang to localised guides and the open-model pages. Route those bookings to a Spanish- or Portuguese-speaking founder or forward-deployed engineer, and track LATAM demos as a separate funnel.

    The content already attracts readers; the gap is the last click. Buyers book faster in their own language, and few agent vendors serve LATAM payment teams in Spanish or Portuguese.

    Risks and how to handle them

    • Calls in Portuguese need a speaker

      Offer English or Spanish first in Brazil until a Portuguese speaker joins.

    Evidence

    • Your websiteMexico fraud guide (updated October 26, 2026): written in Spanish around SPEI, CoDi, CLABE, INE, UIF reports and PLD, but the navigation and "Book a demo" button stay in English and lead to the English /demo page; no hreflang is visible; the only customer cited is Rain.Source
    • Your websiteSitemap (214 URLs, checked November 9, 2026): 77 comparison pages, 110 guides (16 in Portuguese for Brazil and 33 in Spanish for Chile, Colombia, Mexico and LATAM), 17 open-source model pages, 11 solution pages, 4 blog posts and 2 monthly release Briefs; no customer, case study, pricing, about or careers pages.Source
    • Your websiteDemo page: "Book a demo with the Runtime team" and "Thirty minutes with the engineers who built it"; English only; the page text asks nothing about the workflow before the call; an email fallback for "Can't find a time that works?".Source
    • Y CombinatorCompany profile and job posts: Spring 2026 batch, 7 people, San Francisco; the CEO is a two-time YC founder and former BlackRock AI quant whose previous company, Mentum (YC S21, acquired), served Latin America's largest asset and wealth managers; the CTO built underwriting and reconciliation infrastructure at Modern Treasury, Finix and Hulu.Source
  • 03

    On its own chargeback comparison, Runtime lists rivals' prices (Chargeflow 25% of recoveries, Stripe Smart Disputes 30% on wins) and its own as "By request; free pilot". The free pilot is the strongest offer Runtime has, but no page explains it: the Enterprise page has no pilot terms, and Runtime Labs already promises to measure the first agent "against the cases you already resolved".

    High impactLow effort First result in 1 month CEO

    First action

    Write the pilot terms on one page: scope, weeks, data handling, success measure and exit.

    How to measure it

    Pilots started per month

    Now
    –
    Target
    4
    By
    90 days

    Steps

    1. 1

      Define scope

      One SOP, read-only, approvals on, customer's cloud.

      CEO (co-founder) 1 day

    2. 2

      Set the bar

      Accuracy against resolved cases and time saved per case.

      CTO (co-founder) 1 day

    3. 3

      Write the page

      Timeline, what the customer provides, what Runtime provides.

      CEO (co-founder) 2 days

    4. 4

      Add security answers

      Deployment option, masking, SOC 2 report access.

      CTO (co-founder) 1 day

    5. 5

      Link it

      Comparison tables, demo page, Enterprise page.

      Founding engineer 1 day

    6. 6

      Track

      Pilot starts and pilot-to-paid rate.

      Founding AI Ops monthly

    First result
    1 month
    Cost
    $0
    Owner
    CEO (co founder)

    Smallest test

    Link a one-page pilot PDF from the chargeback comparison this week.

    Why it works

    Publish /pilot: one SOP, a forward-deployed engineer, a fixed number of weeks, read-only to start, measured against the customer's own resolved cases, deployed in their cloud, and a clear next step if it meets the agreed bar. Link it from every comparison table cell that says "free pilot" and from the demo page.

    A written, bounded pilot lowers the perceived risk for a careful buyer and gives an internal champion something to forward to security and procurement.

    Risks and how to handle them

    • Too many small pilots for a 7-person team

      Publish clear entry criteria such as volume and workflow type.

    Evidence

    • Your website"Best AI Agents for Chargeback Management (2026)" (updated October 8, 2026): Runtime ranks itself first of nine ("We make Runtime, so weigh our pick accordingly"); Runtime's price is "By request; free pilot" while rivals' prices are listed (Chargeflow 25% of recovered chargebacks, Disputifier 20% capped at $250 a win, Stripe Smart Disputes 30% on wins); the close is "Bring one SOP"; no customer story is linked.Source
    • Your websiteEnterprise page: three deployments (Runtime cloud "for work that does not touch sensitive data", your cloud "recommended for PCI", self-hosted with a Helm chart); "No account executives": forward-deployed engineers map processes, build the first agents, set guardrails and train admins; proof is the Forbes feature, the founders' backgrounds, SOC 2 Type II and the public GitHub repository; no pricing or pilot terms.Source
    • Your websiteRuntime Labs: "An independent applied AI lab for fintech operations" working on evals, routing, reliability and memory, with "outcome-based measures" per workflow, a four-step phased-autonomy path, and the offer to build the first agent "and measure it against the cases you already resolved".Source
  • 04

    Runtime launched on Hacker News in May as "Sandboxed coding agents for everyone on a team". The site now sells "the AI agent harness for payment teams", but the public trail still tells the old story. The GitHub repository linked from the footer, and cited as proof on the Enterprise page, opens with "Open-source sandboxes where coding agents build and deploy". Two of the three dated blog posts are about coding sandboxes and never mention payments. A bank risk lead who checks the repo meets a different product.

    Medium impactLow effort First result in 2 weeks CTO

    First action

    Rewrite the first three lines of the README.

    How to measure it

    GitHub and blog referral demos

    Now
    –
    Target
    +50%
    By
    90 days

    Steps

    1. 1

      Rewrite README

      Sandbox layer, then the payments harness on top.

      CTO (co-founder) 1 day

    2. 2

      Add an example

      A reconciliation or support agent on synthetic data.

      Founding engineer 2 days

    3. 3

      Annotate old posts

      Editor's notes linking to the payments pages.

      CEO (co-founder) 1 hour

    4. 4

      Write two posts

      Payments-first stories from live deployments.

      CEO (co-founder) 1 week

    5. 5

      Review proof links

      Enterprise page cites the right asset.

      CEO (co-founder) 1 hour

    6. 6

      Track

      Referral traffic from GitHub and the blog.

      Founding AI Ops monthly

    First result
    2 weeks
    Cost
    $0
    Owner
    CTO (co founder)

    Smallest test

    Change the README header and watch repo-to-site clicks.

    Why it works

    Rewrite the README header to explain how the open-source sandbox layer underpins the payments harness: isolation, secrets proxy and audit. Add a payments example, and link from the README back to the security page and docs. Add an editor's note to the coding-agent posts explaining the evolution, and publish two payments-first posts so the blog index leads with the current product. Decide whether the Enterprise page should cite the repo or the security page as proof.

    Due diligence happens on GitHub and the blog. A consistent story stops the pivot from reading as a lack of focus, and it keeps the developer goodwill from the launch.

    Risks and how to handle them

    • Alienating the developer audience from the launch

      Keep the sandbox docs intact and explain how they fit.

    Evidence

    • GitHubRepository linked from the site footer and cited as proof on the Enterprise page: README headline "Open-source sandboxes where coding agents build and deploy", 303 stars, 39 forks, topics such as claude-code, cursor and sandbox; no payments content.Source
    • Your websiteBlog: four posts; "Running coding agents without burning down your machine" (February 5, 2026) says "Runtime spins up coding sessions without anyone touching their local machines" and never mentions payments; the newest post is "How to Build an AI Agent Harness" (October 8, 2026).Source
    • Hacker NewsLaunch HN (May 21, 2026, 103 points): "Runtime (YC P26) – Sandboxed coding agents for everyone on a team", priced as a flat platform fee plus compute with "no token markup" and a free tier; commenters found the demo video dizzying, reported 404s in the docs, were confused by the license and asked what it adds over free open-source sandboxes.Source
    • Your websiteEnterprise page: three deployments (Runtime cloud "for work that does not touch sensitive data", your cloud "recommended for PCI", self-hosted with a Helm chart); "No account executives": forward-deployed engineers map processes, build the first agents, set guardrails and train admins; proof is the Forbes feature, the founders' backgrounds, SOC 2 Type II and the public GitHub repository; no pricing or pilot terms.Source
  • 05

    The homepage has two logo strips in the same unlinked style. The real customer strip shows four logos (Rain, Nixtla, Mono, LineLeap). The engineer strip shows Modern Treasury, Finix, Marqeta, Adyen, BlackRock, Meta, Hulu, Disney and Y Combinator, with small print saying these are "companies like these" where an engineer "has built payment and risk systems". A skimming buyer can mistake those for customers, and a careful one may see it as overreach.

    Medium impactLow effort First result in 1 week CEO

    First action

    Change the heading text on the engineer strip.

    How to measure it

    Homepage-to-demo conversion

    Now
    –
    Target
    No drop, with fewer "are they a customer?" questions on calls
    By
    30 days

    Steps

    1. 1

      Retitle

      Make "previously built at" explicit.

      CEO (co-founder) 1 hour

    2. 2

      Restyle

      Muted logos or text names.

      Founding engineer 2 hours

    3. 3

      Move

      Next to the team section.

      Founding engineer 1 hour

    4. 4

      Check other pages

      Guides and comparisons that repeat the engineer logos.

      Founding engineer 1 day

    5. 5

      Ask customers

      Permission to add more customer logos.

      Founding AI Ops 2 weeks

    6. 6

      Listen

      Note logo questions on demo calls.

      Forward Deployed AI Engineer 1 month

    First result
    1 week
    Cost
    $0
    Owner
    CEO (co founder)

    Smallest test

    Change the heading today.

    Why it works

    Retitle the engineer strip "Our engineers previously built payments and risk systems at", show the logos greyed out or as text, and move it next to the team section. Keep the customer strip on its own, and add customers as they approve.

    Compliance-minded buyers reward precision. A clear label keeps the credibility of the team's background without risking the trust lost when a reader feels misled.

    Risks and how to handle them

    • The page looks less impressive

      Pair it with the Rain case study from idea 1.

    Evidence

    • Your websiteHomepage engineer strip: "Forward-deployed AI engineers for your AI strategy" shows logos of Modern Treasury, Finix, Marqeta, Adyen, BlackRock, Meta, Hulu, Disney, Y Combinator and the University of Michigan as "companies like these" where an engineer "has built payment and risk systems", in the same unlinked logo style as the four-logo customer strip.Source
    • Your websiteHomepage (checked November 9, 2026): one customer result, Rain, "Running at the speed of a $1.95B stablecoin company" with "$750k" "Vendor and salary spend saved", and no link to a case study; a cost block ("10×" more runs, "−93%" cost per run, "$0.42 / run" to "$0.03 / run", "−29%" total weekly spend) labelled "Figures are illustrative."; four customer logos (Rain, Nixtla, Mono, LineLeap); "Book a demo" is the only call to action and no pricing is shown.Source
  • 06

    Real visitors already pass Core Web Vitals (field LCP 1.5 s), so this is not urgent. But in Google's mobile lab test the largest element, the animated H1, takes 8.6 s to appear, with 777 ms of blocking time and 593 KB of JavaScript, for a mobile score of 52. Buyers on phones and slower networks, and LATAM readers in particular, see the headline last. PageSpeed also flags low colour contrast and small touch targets.

    Low impactLow effort First result in 1 week Founding engineer

    First action

    Serve a static first headline word and defer the animation.

    How to measure it

    Mobile lab LCP

    Now
    –
    Target
    Under 3 s
    By
    30 days

    Steps

    1. 1

      Static headline

      First state in server HTML.

      Founding engineer half a day

    2. 2

      Defer animation

      Start the rotation after load.

      Founding engineer 2 hours

    3. 3

      Trim JavaScript

      Unused and legacy bundles.

      Founding engineer 1 day

    4. 4

      Cache headers

      Longer lifetimes for static assets.

      Founding engineer 1 hour

    5. 5

      Accessibility

      Contrast and touch-target sizes.

      Founding engineer half a day

    6. 6

      Re-test

      PageSpeed mobile and desktop.

      Founding engineer 1 hour

    First result
    1 week
    Cost
    $0
    Owner
    Founding engineer

    Smallest test

    Static first headline only, then re-run PageSpeed.

    Why it works

    Render the first headline state as static server HTML and start the rotation only after load. Defer non-essential scripts, set longer cache lifetimes, and fix contrast and touch-target sizes.

    The headline is the message; it should be the first thing painted, not the last. These are cheap fixes on a Next.js site.

    Risks and how to handle them

    • Effort spent where visitors already pass

      Cap it at three days; field data stays the guide.

    Evidence

    • Google PageSpeedHomepage, November 9, 2026: mobile score 52 (desktop 78); in the lab test the largest element, the animated H1, appears at 8.6 s on mobile, with 777 ms of blocking time and 593 KB of JavaScript; real visitors pass Core Web Vitals (field LCP 1.5 s, CLS 0); two accessibility issues: low colour contrast and small touch targets.Source

This quarter 12

The core plays for the next three months.

  • 07

    The job posts say customers are "fast-growing fintech unicorns and payments companies that move hundreds of billions of dollars a year", using Runtime to trace stuck payments, work alerts and clear reconciliation breaks. Publicly there is one result. Many fintechs won't be named, but they will often allow anonymised numbers.

    High impactMedium effort First result in 6 weeks Forward Deployed AI Engineer

    First action

    Add "60-day result card" to the forward-deployed engineer handover checklist.

    How to measure it

    Workflows with published results

    Now
    –
    Target
    6 of 11 solution pages
    By
    1 quarter

    Steps

    1. 1

      Template

      One card format and the fields it needs.

      Forward Deployed AI Engineer 1 day

    2. 2

      Permission

      An anonymisation clause in pilot terms.

      CEO (co-founder) 1 week

    3. 3

      Collect

      From run records at 60 days.

      Forward Deployed AI Engineer ongoing

    4. 4

      Publish

      On the solution pages.

      Founding engineer 1 day each

    5. 5

      Syndicate

      Guides and comparisons by workflow.

      Founding engineer 1 week

    6. 6

      Measure

      Demos from pages with and without cards.

      Founding AI Ops monthly

    First result
    6 weeks
    Cost
    Staff time only
    Owner
    Forward Deployed AI Engineer

    Smallest test

    One card on the payment operations page.

    Why it works

    For each solution page (payment ops, fraud and risk review, KYB/KYC, compliance, support, reconciliation), publish one result card: the type of company, the queue, cases per week, the share handled with approval, time per case before and after, and cost per run. Each deploying engineer collects these at the 60-day mark as part of handover. Show the matching card on each solution, guide and comparison page.

    Buyers want proof for their own workflow, not the platform in general. Anonymised specifics beat named but vague logos.

    Risks and how to handle them

    • Customers can be identified from the details

      Use ranges and company types, and get each card approved.

    Builds on

    Evidence

    • HiringFounding AI Ops (San Francisco, $120K–$165K, the first operations hire on a 7-person team): "Own the customer experience end to end, from onboarding and support", "Build and run internal agents on Runtime for company operations" and run "metrics, customer health, renewals"; customers are "fast-growing fintech unicorns and payments companies that move hundreds of billions of dollars a year"; "We dogfood Runtime heavily".Source
    • HiringForward Deployed AI Engineer (San Francisco or remote, $100K–$200K): embeds with customers to map "payment investigations, underwriting, reconciliation, and fraud reviews" and build agents, connects processors, ledgers, banks and warehouses, and must "Train customer teams to build and run their own agents"; should "care as much about reliability, access controls, and audit trails as you do about speed".Source
    • Your websiteSitemap (214 URLs, checked November 9, 2026): 77 comparison pages, 110 guides (16 in Portuguese for Brazil and 33 in Spanish for Chile, Colombia, Mexico and LATAM), 17 open-source model pages, 11 solution pages, 4 blog posts and 2 monthly release Briefs; no customer, case study, pricing, about or careers pages.Source
  • 08

    October's release added Compliance (beta) with risk scoring, an audit log and "a one-page evidence report you can print" for each agent. That one page is exactly what a compliance officer or partner bank asks for before approving automation, but it lives behind the login. Runtime already writes guides on AI agent compliance for the EU, UK and US.

    High impactLow effort First result in 1 month CTO

    First action

    Export the report for an internal demo agent and review it for anything sensitive.

    How to measure it

    Evidence-report downloads to demo

    Now
    –
    Target
    15% of downloaders book
    By
    1 quarter

    Steps

    1. 1

      Build a demo agent

      Synthetic data, realistic SOP.

      Forward Deployed AI Engineer 2 days

    2. 2

      Export

      The evidence report as PDF.

      CTO (co-founder) 1 hour

    3. 3

      Annotate

      What each section proves.

      CTO (co-founder) 1 day

    4. 4

      Publish

      Security, compliance solution page and compliance guides.

      Founding engineer 1 day

    5. 5

      Use in pilots

      Attach to every pilot proposal.

      CEO (co-founder) ongoing

    6. 6

      Track

      Downloads and resulting demos.

      Founding AI Ops monthly

    First result
    1 month
    Cost
    $0
    Owner
    CTO (co founder)

    Smallest test

    Send the sample to three prospects' compliance contacts.

    Why it works

    Generate the evidence report for a demo agent on synthetic data, such as a reconciliation agent, and publish it as a downloadable sample on the security and compliance pages, inside the compliance guides and in pilot materials. Annotate each section: what is tested, who approved, which controls apply and how risk is scored.

    It lets the toughest internal stakeholder say yes early, and it turns a feature into proof without needing customer permission.

    Risks and how to handle them

    • The sample reveals internal detail

      Use synthetic data and review before publishing.

    Evidence

    • Your websiteThe Brief, October 2026: Compliance (beta) with risk scoring, an audit log and "a one-page evidence report you can print" for each agent; data masking; Okta and Duo SSO; Microsoft Teams bots and a Microsoft 365 connector; a new dashboard. No customers are named, and the Releases page has no email signup or RSS feed.Source
    • Your websiteGuides index: ops-agent guides for Adyen, Finix, Lithic, Marqeta, Stripe and Unit, plus guides on partner-bank requests, Reg E dispute investigations, AI agent compliance in the EU, UK and US, and "what is know your agent".Source
  • 09

    Runtime already publishes ops-agent guides for six payment platforms, and the CTO built payment infrastructure at Finix and Modern Treasury. Customers on the same processor share the same ops problems: the same dashboards, webhooks, dispute flows and reconciliation files. Today each guide ends in a generic demo request.

    Medium impactMedium effort First result in 1 quarter CTO

    First action

    Pick the two platforms most common among current customers and pilots.

    How to measure it

    Platform-template demos

    Now
    –
    Target
    10 a quarter
    By
    2 quarters

    Steps

    1. 1

      Rank platforms

      By customers and pipeline.

      CEO (co-founder) 1 day

    2. 2

      Build templates

      One agent per platform with the connector pre-wired.

      Forward Deployed AI Engineer 2 weeks each

    3. 3

      Update guides

      Platform-specific calls to action.

      Founding engineer 1 day

    4. 4

      Apply to directories

      Partner or app listings where programmes exist.

      CTO (co-founder) 1 month

    5. 5

      Co-market

      Offer a joint guide to each partner team.

      CEO (co-founder) 1 quarter

    6. 6

      Track

      Demos and pilots by platform.

      Founding AI Ops monthly

    First result
    1 quarter
    Cost
    Engineering time
    Owner
    CTO (co founder)

    Smallest test

    One template for the most common platform.

    Why it works

    For each platform, ship a ready-made agent template (for example, disputes or payout investigations) with the connector pre-wired. Apply to each platform's partner or app directory where one exists, and make the guide's call to action "Start with the <platform> template". Track demos by platform.

    Platform-specific templates shorten the pilot and give Runtime a reason to be listed where customers of those platforms look for tools.

    Risks and how to handle them

    • Partner programmes are slow or closed

      The templates still speed up pilots on their own.

    Evidence

    • Your websiteGuides index: ops-agent guides for Adyen, Finix, Lithic, Marqeta, Stripe and Unit, plus guides on partner-bank requests, Reg E dispute investigations, AI agent compliance in the EU, UK and US, and "what is know your agent".Source
    • Y CombinatorCompany profile and job posts: Spring 2026 batch, 7 people, San Francisco; the CEO is a two-time YC founder and former BlackRock AI quant whose previous company, Mentum (YC S21, acquired), served Latin America's largest asset and wealth managers; the CTO built underwriting and reconciliation infrastructure at Modern Treasury, Finix and Hulu.Source
  • 10

    The Founding AI Ops role is to "Build and run internal agents on Runtime for company operations": onboarding, support triage, sales ops, finance, recruiting and security/compliance. The post says "We dogfood Runtime heavily". Those internal agents are working examples of what a lean ops team can do, and they don't involve any customer's data.

    Medium impactMedium effort First result in 2 months Founding AI Ops

    First action

    Write up the first internal agent the hire ships, with its numbers.

    How to measure it

    Demos that viewed a template

    Now
    –
    Target
    20%
    By
    2 quarters

    Steps

    1. 1

      Pick the first

      Support triage or onboarding.

      Founding AI Ops 1 week

    2. 2

      Measure it

      Cases, cost per run, time saved.

      Founding AI Ops 1 month

    3. 3

      Write the template

      SOP, tools, approvals, results.

      Founding AI Ops 2 days

    4. 4

      Publish

      A "Runtime on Runtime" page and docs.

      Founding engineer 1 day

    5. 5

      Repeat

      One template a month.

      Founding AI Ops monthly

    6. 6

      Track

      Template views among demo bookers.

      Founding AI Ops monthly

    First result
    2 months
    Cost
    Staff time only
    Owner
    Founding AI Ops

    Smallest test

    Publish one internal agent write-up.

    Why it works

    Each month, publish one internal agent as a template with its SOP, tools, approval points, cost per run and what it replaced, starting with support triage and customer onboarding. Add a "How Runtime runs on Runtime" page and include the templates in the docs.

    It is proof the company controls fully, it doubles as onboarding material for customers, and it shows the lean-team thesis in practice.

    Risks and how to handle them

    • Internal workflows look too simple for banks

      Pair each one with the matching payments workflow guide.

    Evidence

    • HiringFounding AI Ops (San Francisco, $120K–$165K, the first operations hire on a 7-person team): "Own the customer experience end to end, from onboarding and support", "Build and run internal agents on Runtime for company operations" and run "metrics, customer health, renewals"; customers are "fast-growing fintech unicorns and payments companies that move hundreds of billions of dollars a year"; "We dogfood Runtime heavily".Source
    • Your websiteDocs: "Runtime is where operations teams create and run AI agents"; the Quickstart promises "Ten minutes from a blank org to a graded run"; a one-line CLI install and a hosted app; no signup, free tier or pricing information.Source
  • 11

    The forward-deployed engineer's job includes "Train customer teams to build and run their own agents" so the product keeps working after they leave. With 7 people, each engineer can only embed with a few customers. Expansion depends on customers building agent number two without Runtime in the room.

    High impactMedium effort First result in 2 months Forward Deployed AI Engineer

    First action

    Define "certified builder": the tasks a customer must complete alone.

    How to measure it

    Self-built second agents

    Now
    –
    Target
    60% of customers within 60 days of the first
    By
    2 quarters

    Steps

    1. 1

      Define skills

      SOP to agent, tools, approvals, evals.

      Forward Deployed AI Engineer 2 days

    2. 2

      Build the course

      Two sessions using the customer's first agent.

      Forward Deployed AI Engineer 1 week

    3. 3

      Run it

      At each handover.

      Forward Deployed AI Engineer ongoing

    4. 4

      Office hours

      Weekly, shared Slack or Teams channel.

      Founding AI Ops weekly

    5. 5

      Track

      Second-agent rate and time to build it.

      Founding AI Ops monthly

    6. 6

      Publish

      Describe the certification on the Enterprise page.

      CEO (co-founder) 1 day

    First result
    2 months
    Cost
    Staff time only
    Owner
    Forward Deployed AI Engineer

    Smallest test

    Certify one customer team and watch for a second agent.

    Why it works

    Create a two-session builder certification for each customer's admins and ops leads, built on the docs quickstart and their first agent. Graduates build their second agent with office-hours support only. Track "second agent without an engineer" as the main activation metric, and publish the certification so prospects see the path to independence.

    Self-built agents expand usage without adding headcount, and buyers worried about vendor dependence see a clear handover.

    Risks and how to handle them

    • Customer-built agents are lower quality

      Require evals to pass before production.

    Evidence

    • HiringForward Deployed AI Engineer (San Francisco or remote, $100K–$200K): embeds with customers to map "payment investigations, underwriting, reconciliation, and fraud reviews" and build agents, connects processors, ledgers, banks and warehouses, and must "Train customer teams to build and run their own agents"; should "care as much about reliability, access controls, and audit trails as you do about speed".Source
    • Your websiteDocs: "Runtime is where operations teams create and run AI agents"; the Quickstart promises "Ten minutes from a blank org to a graded run"; a one-line CLI install and a hosted app; no signup, free tier or pricing information.Source
    • Your websiteEnterprise page: three deployments (Runtime cloud "for work that does not touch sensitive data", your cloud "recommended for PCI", self-hosted with a Helm chart); "No account executives": forward-deployed engineers map processes, build the first agents, set guardrails and train admins; proof is the Forbes feature, the founders' backgrounds, SOC 2 Type II and the public GitHub repository; no pricing or pilot terms.Source
  • 12

    Runtime already has a guide on partner-bank request agents and a printable evidence report for every agent. Every fintech on a sponsor bank answers that bank's requests for information, so the bank sees the same slow answers from dozens of programmes. One bank relationship could reach many fintechs that all share one painful workflow.

    High impactMedium effort First result in 1 quarter CEO

    First action

    Ask two customers which sponsor banks they answer to and whether they would introduce Runtime.

    How to measure it

    Bank-referred pilots

    Now
    –
    Target
    3
    By
    2 quarters

    Steps

    1. 1

      Map banks

      Sponsor banks behind current customers.

      CEO (co-founder) 1 week

    2. 2

      Package the agent

      RFI agent plus evidence report.

      Forward Deployed AI Engineer 2 weeks

    3. 3

      Get introductions

      Through customers' bank contacts.

      CEO (co-founder) 1 month

    4. 4

      Bank pitch

      Consistency, turnaround and audit trail.

      CEO (co-founder) 1 month

    5. 5

      Group pilot

      Several of the bank's fintechs at once.

      Forward Deployed AI Engineer 1 quarter

    6. 6

      Track

      Pilots and conversions from bank channels.

      Founding AI Ops monthly

    First result
    1 quarter
    Cost
    Founder time
    Owner
    CEO (co founder)

    Smallest test

    One bank conversation, introduced by one customer.

    Why it works

    Package the RFI agent with a standard evidence report from idea 8. Pitch sponsor banks on faster, more consistent answers with a full audit trail, and offer their fintech programmes a group pilot. Start with banks that current customers already work with, introduced by those customers.

    Banks are a distribution channel with authority: a fintech is far more likely to adopt a tool its partner bank prefers, and the bank gains consistency across programmes.

    Risks and how to handle them

    • Banks move slowly

      Keep direct fintech sales running in parallel.

    Builds on

    Evidence

    • Your websiteGuides index: ops-agent guides for Adyen, Finix, Lithic, Marqeta, Stripe and Unit, plus guides on partner-bank requests, Reg E dispute investigations, AI agent compliance in the EU, UK and US, and "what is know your agent".Source
    • Your websiteOpen-source models page (updated October 29, 2026): 17 open-weight models ranked for agent work, with Spanish and Portuguese versions and picks for reconciliation, chargebacks, KYC/KYB, AML, underwriting, monitoring and support; "Benchmark scores are reported by each model provider" and "are not independently verified by Runtime".Source
    • Your websiteThe Brief, October 2026: Compliance (beta) with risk scoring, an audit log and "a one-page evidence report you can print" for each agent; data masking; Okta and Duo SSO; Microsoft Teams bots and a Microsoft 365 connector; a new dashboard. No customers are named, and the Releases page has no email signup or RSS feed.Source
  • 13

    Runtime publishes a monthly Brief of releases and over 200 guides, comparisons and model pages, but the Releases page has no email signup or RSS feed. A risk lead who reads a guide today and isn't ready for a demo has no way to stay in touch.

    Medium impactLow effort First result in 1 month CEO

    First action

    Add a signup block to the ten most-read guides.

    How to measure it

    Subscribers

    Now
    –
    Target
    1,000
    By
    2 quarters

    Steps

    1. 1

      Pick a tool

      Simple email platform with double opt-in.

      Founding AI Ops 1 day

    2. 2

      Add signup

      Guides, comparisons, Releases.

      Founding engineer 1 day

    3. 3

      Write the format

      Insight, template or result, releases.

      CEO (co-founder) 1 day

    4. 4

      Localise

      Spanish and Portuguese editions.

      CEO (co-founder) monthly

    5. 5

      Send

      First of each month.

      Founding AI Ops monthly

    6. 6

      Measure

      Subscribers, opens and demo replies.

      Founding AI Ops monthly

    First result
    1 month
    Cost
    Under $100 a month
    Owner
    CEO (co founder)

    Smallest test

    Signup on ten guides for one month.

    Why it works

    Add a one-field signup to the guides, comparisons and Releases page. Send The Brief monthly with one ops insight, one new template or result, and the releases. Send Spanish and Portuguese editions to LATAM readers. Use the list to invite readers to pilots and webinars.

    Enterprise cycles are long. A list turns one-time search visitors into a warm audience for the day they are ready.

    Risks and how to handle them

    • Low engagement

      Lead with an operational insight, not release notes.

    Evidence

    • Your websiteThe Brief, October 2026: Compliance (beta) with risk scoring, an audit log and "a one-page evidence report you can print" for each agent; data masking; Okta and Duo SSO; Microsoft Teams bots and a Microsoft 365 connector; a new dashboard. No customers are named, and the Releases page has no email signup or RSS feed.Source
    • Your websiteSitemap (214 URLs, checked November 9, 2026): 77 comparison pages, 110 guides (16 in Portuguese for Brazil and 33 in Spanish for Chile, Colombia, Mexico and LATAM), 17 open-source model pages, 11 solution pages, 4 blog posts and 2 monthly release Briefs; no customer, case study, pricing, about or careers pages.Source
  • 14

    At launch Runtime described pricing as "a flat platform fee plus compute" with "no token markup" and a free tier. Today there is no pricing page, and comparisons say "By request" next to rivals' published percentages. Taktile, Unit21 and Sardine don't publish prices either, so explaining the structure alone would set Runtime apart.

    Medium impactLow effort First result in 1 month CEO

    First action

    Decide which parts of the launch pricing still hold.

    How to measure it

    Pricing-page demos

    Now
    –
    Target
    10% of pricing visitors
    By
    1 quarter

    Steps

    1. 1

      Confirm the model

      Fee, compute, tokens and deployments.

      CEO (co-founder) 2 days

    2. 2

      Write the page

      Structure, examples, FAQ.

      CEO (co-founder) 2 days

    3. 3

      Cost-per-run explainer

      How routing and scripts lower cost.

      CTO (co-founder) 2 days

    4. 4

      Publish

      Navigation link and quote form.

      Founding engineer 1 day

    5. 5

      Update comparisons

      Link "By request" to /pricing.

      Founding engineer 1 day

    6. 6

      Measure

      Visits and quote requests.

      Founding AI Ops monthly

    First result
    1 month
    Cost
    $0
    Owner
    CEO (co founder)

    Smallest test

    A pricing FAQ section on the Enterprise page.

    Why it works

    Add /pricing that explains the model: the platform fee and what it covers, compute passed through at cost, no markup on model tokens, what changes for Runtime cloud, your cloud and self-hosted, and what the pilot costs (nothing). Add a cost-per-run explainer linked to the routing story, and a "get a quote" form.

    Buyers want to know how cost will grow before they invest time. "No token markup" is a strong line that is currently invisible.

    Risks and how to handle them

    • Locks in a model too early

      Publish the structure, not the numbers.

    Builds on

    Evidence

    • Hacker NewsLaunch HN (May 21, 2026, 103 points): "Runtime (YC P26) – Sandboxed coding agents for everyone on a team", priced as a flat platform fee plus compute with "no token markup" and a free tier; commenters found the demo video dizzying, reported 404s in the docs, were confused by the license and asked what it adds over free open-source sandboxes.Source
    • Your website"Best AI Agents for Chargeback Management (2026)" (updated October 8, 2026): Runtime ranks itself first of nine ("We make Runtime, so weigh our pick accordingly"); Runtime's price is "By request; free pilot" while rivals' prices are listed (Chargeflow 25% of recovered chargebacks, Disputifier 20% capped at $250 a win, Stripe Smart Disputes 30% on wins); the close is "Bring one SOP"; no customer story is linked.Source
    • Competitor researchRivals' own homepages: Taktile shows Allianz, Mercury and Monzo, "75% less time to onboard" for Faire and go-live in about 4 weeks for Zilch; Unit21 shows a dozen customer logos and case studies (Nexo: 57% of alert reviews automated) under "Most systems generate alerts. Unit21 resolves them."; Sardine cites a Forrester Wave Leader placement. None of the three publishes prices.Source
  • 15

    Every page ends with "Bring one SOP", and the demo is "Thirty minutes with the engineers who built it". The demo page asks nothing about the workflow beforehand, so the first call starts cold. Ops leaders already have SOPs as documents, and Runtime's first step is "Describe the work in plain English and attach the SOP."

    High impactMedium effort First result in 1 month Forward Deployed AI Engineer

    First action

    Write the spec template from an existing pilot.

    How to measure it

    Teardowns that become pilots

    Now
    –
    Target
    30%
    By
    1 quarter

    Steps

    1. 1

      Spec template

      Steps, tools, approvals, measures.

      Forward Deployed AI Engineer 1 day

    2. 2

      Upload form

      Redaction guidance and an NDA option.

      Founding engineer 2 days

    3. 3

      Use Runtime

      An internal agent drafts the spec for engineer review.

      Founding AI Ops 1 week

    4. 4

      Promote

      Calls to action on guides and comparisons.

      Founding engineer 1 day

    5. 5

      Run the call

      Walk through the spec.

      Forward Deployed AI Engineer ongoing

    6. 6

      Track

      Teardowns, calls and pilots.

      Founding AI Ops monthly

    First result
    1 month
    Cost
    Engineer time per teardown
    Owner
    Forward Deployed AI Engineer

    Smallest test

    Offer teardowns to the next ten demo bookers.

    Why it works

    Add an "Send us one SOP" option: the prospect uploads a redacted procedure and within 48 hours receives a one-page agent spec: steps the agent takes, tools it needs, approval points, expected share automated and how it would be measured. The demo call then walks through their own spec.

    It gives value before the call, qualifies serious prospects, and makes the demo about the buyer's real work instead of a generic tour.

    Risks and how to handle them

    • Prospects share sensitive data

      Ask for redacted SOPs and use masking on intake.

    • Too many low-fit requests

      Qualify on company type and volume first.

    Evidence

    • Your websiteDemo page: "Book a demo with the Runtime team" and "Thirty minutes with the engineers who built it"; English only; the page text asks nothing about the workflow before the call; an email fallback for "Can't find a time that works?".Source
    • Your website"Best AI Agents for Chargeback Management (2026)" (updated October 8, 2026): Runtime ranks itself first of nine ("We make Runtime, so weigh our pick accordingly"); Runtime's price is "By request; free pilot" while rivals' prices are listed (Chargeflow 25% of recovered chargebacks, Disputifier 20% capped at $250 a win, Stripe Smart Disputes 30% on wins); the close is "Bring one SOP"; no customer story is linked.Source
    • HiringForward Deployed AI Engineer (San Francisco or remote, $100K–$200K): embeds with customers to map "payment investigations, underwriting, reconciliation, and fraud reviews" and build agents, connects processors, ledgers, banks and warehouses, and must "Train customer teams to build and run their own agents"; should "care as much about reliability, access controls, and audit trails as you do about speed".Source
  • 16

    The demo page promises "Thirty minutes with the engineers who built it", but its text asks nothing about the workflow, the company type or where agents must run, even though the Enterprise page offers three different deployments. Each call with an engineer starts without context.

    Medium impactLow effort First result in 2 weeks Founding engineer

    First action

    List every page that links to /demo and the workflow each one is about.

    How to measure it

    Demo-to-pilot conversion

    Now
    –
    Target
    +20%
    By
    1 quarter

    Steps

    1. 1

      Audit tags

      Every page linking to /demo and its workflow.

      Founding engineer half a day

    2. 2

      Add questions

      Workflow, company type, deployment.

      Founding engineer 1 day

    3. 3

      Pre-fill

      From the referring page.

      Founding engineer 1 day

    4. 4

      Route

      Language, compliance and self-serve paths.

      Founding AI Ops 1 day

    5. 5

      Brief the engineer

      Answers in the calendar invite.

      Founding AI Ops 1 day

    6. 6

      Measure

      Booking rate and demo-to-pilot rate.

      Founding AI Ops monthly

    First result
    2 weeks
    Cost
    $0
    Owner
    Founding engineer

    Smallest test

    One pre-filled workflow question.

    Why it works

    Add three short questions: which workflow (pre-filled from the page the visitor came from), company type (PSP, issuer, bank, lender, other), and where agents must run (Runtime cloud, your cloud, self-hosted). Route by answer: LATAM to idea 2's route, compliance-heavy to the evidence report, early-stage to the self-serve path.

    Small forms that pre-fill from context barely hurt conversion, and they let each engineer-led call start with the right example.

    Risks and how to handle them

    • Fewer bookings

      Make the questions optional and test for two weeks.

    Evidence

    • Your websiteDemo page: "Book a demo with the Runtime team" and "Thirty minutes with the engineers who built it"; English only; the page text asks nothing about the workflow before the call; an email fallback for "Can't find a time that works?".Source
    • Your websiteEnterprise page: three deployments (Runtime cloud "for work that does not touch sensitive data", your cloud "recommended for PCI", self-hosted with a Helm chart); "No account executives": forward-deployed engineers map processes, build the first agents, set guardrails and train admins; proof is the Forbes feature, the founders' backgrounds, SOC 2 Type II and the public GitHub repository; no pricing or pilot terms.Source
  • 17

    Of 77 comparison pages, a number compare consumer or general AI agents with each other (grok-bot-vs-town, instinct-vs-meta-muse, instinct-vs-openai-dots, dust-vs-notion-agents, openai-dots-vs-grok-bot). They sit next to valuable payments pages such as sardine-vs-unit21, middesk-vs-sumsub and simetrik-vs-modern-treasury. Off-topic pages pull in readers who will never buy and blur what the site is about for search engines and AI assistants.

    Medium impactLow effort First result in 1 month CEO

    First action

    Export the last 90 days of traffic and demos by comparison page.

    How to measure it

    Comparison-page demo rate

    Now
    –
    Target
    2x
    By
    1 quarter

    Steps

    1. 1

      Export

      Traffic and demos per page.

      Founding AI Ops 1 day

    2. 2

      Classify

      Payments, adjacent, off-topic.

      CEO (co-founder) 1 day

    3. 3

      Prune

      Noindex or redirect off-topic pages with no demos.

      Founding engineer 1 day

    4. 4

      Strengthen

      Result cards and the pilot offer on payments pages.

      Founding engineer 1 week

    5. 5

      Date and source

      Methodology and rival-pricing sources.

      CEO (co-founder) 1 week

    6. 6

      Review

      Quarterly.

      CEO (co-founder) quarterly

    First result
    1 month
    Cost
    $0
    Owner
    CEO (co founder)

    Smallest test

    Noindex five off-topic pages and watch the rest.

    Why it works

    Sort every comparison by search traffic and demo bookings. Noindex or redirect off-topic pages that bring no demos, and put the effort into payments comparisons: add a real result card, the pilot offer and a dated methodology to each.

    A tighter topical footprint helps the pages that matter rank and be cited, and the effort goes where buyers are.

    Risks and how to handle them

    • Losing traffic that occasionally converts

      Prune only pages with no demos in 90 days.

    Evidence

    • Your websiteComparison hub: pages outside payments such as grok-bot-vs-town, instinct-vs-meta-muse, instinct-vs-openai-dots, dust-vs-notion-agents and openai-dots-vs-grok-bot sit next to payments pages such as sardine-vs-unit21, middesk-vs-sumsub and simetrik-vs-modern-treasury.Source
    • Your websiteSitemap (214 URLs, checked November 9, 2026): 77 comparison pages, 110 guides (16 in Portuguese for Brazil and 33 in Spanish for Chile, Colombia, Mexico and LATAM), 17 open-source model pages, 11 solution pages, 4 blog posts and 2 monthly release Briefs; no customer, case study, pricing, about or careers pages.Source
  • 18

    The Founding AI Ops hire will run "metrics, customer health, renewals" and build internal agents on Runtime. Runtime already stores a record of every run, with steps, cost and grades, so usage trends per customer are available. On a team of 7, a quiet decline at one large account is the biggest revenue risk.

    High impactMedium effort First result in 1 month Founding AI Ops

    First action

    Define the five signals and pull them for every customer.

    How to measure it

    Gross revenue retention

    Now
    –
    Target
    95%+
    By
    4 quarters

    Steps

    1. 1

      Pick signals

      Runs, approvals, failures, new agents, active builders.

      Founding AI Ops 2 days

    2. 2

      Build the agent

      Reads run records and scores health.

      Founding AI Ops 1 week

    3. 3

      Set actions

      Next step for each drop pattern.

      Founding AI Ops 2 days

    4. 4

      Weekly review

      With founders and engineers.

      Founding AI Ops weekly

    5. 5

      Renewal prep

      Value summary 90 days before renewal.

      Founding AI Ops ongoing

    6. 6

      Publish

      As a template (idea 10).

      Founding AI Ops 1 day

    First result
    1 month
    Cost
    Staff time only
    Owner
    Founding AI Ops

    Smallest test

    A weekly health table for the top five customers.

    Why it works

    Build an internal agent that reads each customer's weekly runs, approval rates, failed or escalated runs and new agents built, scores health, and posts to the shared customer channel with a suggested next step: a new template, office hours or an engineer visit. Review flagged accounts weekly and 90 days before each renewal.

    Usage drops come before churn. Catching them early with the product's own data is cheap, and it shows customers the same discipline Runtime sells.

    Risks and how to handle them

    • False alarms on seasonal volume

      Compare against each customer's own baseline.

    Evidence

    • HiringFounding AI Ops (San Francisco, $120K–$165K, the first operations hire on a 7-person team): "Own the customer experience end to end, from onboarding and support", "Build and run internal agents on Runtime for company operations" and run "metrics, customer health, renewals"; customers are "fast-growing fintech unicorns and payments companies that move hundreds of billions of dollars a year"; "We dogfood Runtime heavily".Source
    • Your websiteDocs: "Runtime is where operations teams create and run AI agents"; the Quickstart promises "Ten minutes from a blank org to a graded run"; a one-line CLI install and a hosted app; no signup, free tier or pricing information.Source

Big bets 3

Larger plays that can change the trajectory.

  • 19

    The docs promise "Ten minutes from a blank org to a graded run", and Runtime cloud is meant "for work that does not touch sensitive data". The only way in, though, is a demo with an engineer. The launch audience was used to a free tier, and a 7-person company can only run so many demos.

    High impactHigh effort First result in 2 quarters CTO

    First action

    Build the synthetic data set for one workflow.

    How to measure it

    Sandbox-to-pilot conversion

    Now
    –
    Target
    10%
    By
    3 quarters

    Steps

    1. 1

      Synthetic data

      Processor, ledger and queue for three workflows.

      Forward Deployed AI Engineer 1 month

    2. 2

      Sandbox orgs

      Isolated, time-limited, Runtime cloud only.

      CTO (co-founder) 1 month

    3. 3

      Guided quickstart

      Build and grade in ten minutes.

      Founding engineer 2 weeks

    4. 4

      Usage limits

      Caps on runs and compute.

      CTO (co-founder) 1 week

    5. 5

      Hand-off

      Active users routed to an engineer.

      Founding AI Ops 1 week

    6. 6

      Launch

      "Try it" next to "Book a demo".

      CEO (co-founder) 1 day

    7. 7

      Measure

      Activation, graded runs and pilots.

      Founding AI Ops monthly

    First result
    2 quarters
    Cost
    Engineering time plus compute
    Owner
    CTO (co founder)

    Smallest test

    Invite twenty recent demo bookers to one synthetic workflow.

    Why it works

    Open a free sandbox org preloaded with a synthetic processor, ledger and ticket queue: stuck payments, a reconciliation break, a KYB case. Visitors build and grade an agent on fake data in ten minutes. When they want to connect real systems, they book a pilot. Gate by work email, and send active sandbox users to an engineer.

    It lets ops leads and engineers prove value to themselves before a meeting, and it shows which workflows prospects try first.

    Risks and how to handle them

    • Compute costs from free users

      Tight caps and a work-email gate.

    • Distracts from enterprise deployments

      Build on the existing quickstart only.

    Evidence

    • Your websiteDocs: "Runtime is where operations teams create and run AI agents"; the Quickstart promises "Ten minutes from a blank org to a graded run"; a one-line CLI install and a hosted app; no signup, free tier or pricing information.Source
    • Your websiteEnterprise page: three deployments (Runtime cloud "for work that does not touch sensitive data", your cloud "recommended for PCI", self-hosted with a Helm chart); "No account executives": forward-deployed engineers map processes, build the first agents, set guardrails and train admins; proof is the Forbes feature, the founders' backgrounds, SOC 2 Type II and the public GitHub repository; no pricing or pilot terms.Source
    • Hacker NewsLaunch HN (May 21, 2026, 103 points): "Runtime (YC P26) – Sandboxed coding agents for everyone on a team", priced as a flat platform fee plus compute with "no token markup" and a free tier; commenters found the demo video dizzying, reported 404s in the docs, were confused by the license and asked what it adds over free open-source sandboxes.Source
  • 20

    Runtime Labs frames each workflow with "outcome-based measures". In disputes, rivals already charge for outcomes (Chargeflow 25% of recoveries, Stripe Smart Disputes 30% on wins). Runtime's routing story (cost per run falling from $0.42 to $0.03 in the illustration) means its own cost per case drops over time, so a per-case price would grow its margin as it improves.

    High impactHigh effort First result in 2 quarters CEO

    First action

    Model per-case economics on one live workflow.

    How to measure it

    Outcome-priced contracts

    Now
    –
    Target
    30% of new deals
    By
    3 quarters

    Steps

    1. 1

      Model economics

      Cost per case versus price on one workflow.

      CEO (co-founder) 2 weeks

    2. 2

      Define resolved

      Per workflow, with graders.

      CTO (co-founder) 2 weeks

    3. 3

      Meter it

      Resolved cases in run records and billing.

      CTO (co-founder) 1 month

    4. 4

      Offer it

      To two pilots as an option.

      CEO (co-founder) 1 quarter

    5. 5

      Compare

      Close rate and margin against the platform fee.

      Founding AI Ops quarterly

    6. 6

      Decide

      Roll out, adjust or drop.

      CEO (co-founder) 1 day

    First result
    2 quarters
    Cost
    Billing engineering
    Owner
    CEO (co founder)

    Smallest test

    Quote one prospect both ways.

    Why it works

    Offer an optional per-resolved-case price for well-defined workflows (reconciliation breaks, payment investigations, alert reviews), with a platform minimum. Define "resolved" with the customer using their own resolved-case history, and report it in the run records.

    Ops leaders compare against headcount per case. Pricing in that unit makes the ROI obvious and links Runtime's revenue to the value it delivers.

    Risks and how to handle them

    • Disputes over what counts as resolved

      Agree the definition and graders before go-live.

    Builds on

    Evidence

    • Your websiteRuntime Labs: "An independent applied AI lab for fintech operations" working on evals, routing, reliability and memory, with "outcome-based measures" per workflow, a four-step phased-autonomy path, and the offer to build the first agent "and measure it against the cases you already resolved".Source
    • Your website"Best AI Agents for Chargeback Management (2026)" (updated October 8, 2026): Runtime ranks itself first of nine ("We make Runtime, so weigh our pick accordingly"); Runtime's price is "By request; free pilot" while rivals' prices are listed (Chargeflow 25% of recovered chargebacks, Disputifier 20% capped at $250 a win, Stripe Smart Disputes 30% on wins); the close is "Bring one SOP"; no customer story is linked.Source
    • Your websiteHomepage (checked November 9, 2026): one customer result, Rain, "Running at the speed of a $1.95B stablecoin company" with "$750k" "Vendor and salary spend saved", and no link to a case study; a cost block ("10×" more runs, "−93%" cost per run, "$0.42 / run" to "$0.03 / run", "−29%" total weekly spend) labelled "Figures are illustrative."; four customer logos (Rain, Nixtla, Mono, LineLeap); "Book a demo" is the only call to action and no pricing is shown.Source
  • 21

    Runtime's open-model directory ranks 17 models for agent work and offers picks per payments task, but says the benchmark scores "are not independently verified by Runtime". Runtime Labs already works on evals and routing for fintech ops. Nobody publishes how agents perform on reconciliation breaks, chargebacks or KYB reviews, and Runtime has the harness to measure it.

    High impactHigh effort First result in 2 quarters CTO

    First action

    Write 50 synthetic reconciliation cases with expected answers.

    How to measure it

    Benchmark-sourced demos

    Now
    –
    Target
    15 a quarter
    By
    3 quarters

    Steps

    1. 1

      Design tasks

      Five workflows, synthetic and realistic.

      Forward Deployed AI Engineer 1 month

    2. 2

      Write graders

      Correctness, steps, policy adherence.

      CTO (co-founder) 2 weeks

    3. 3

      Run models

      All directory models in the harness.

      CTO (co-founder) 2 weeks

    4. 4

      Publish

      Results on the open-model pages, with method.

      Founding engineer 1 week

    5. 5

      Localise

      Spanish and Portuguese task sets.

      CEO (co-founder) 1 month

    6. 6

      Promote

      Launch post and newsletter.

      CEO (co-founder) 1 week

    7. 7

      Refresh

      Quarterly, as models change.

      CTO (co-founder) quarterly

    First result
    2 quarters
    Cost
    Compute plus engineering time
    Owner
    CTO (co founder)

    Smallest test

    Publish one workflow's results across five models.

    Why it works

    Build a public, synthetic payments-ops task set (reconciliation, disputes, KYB, alert triage, partner-bank requests) with graders. Run the listed models through the Runtime harness, and publish accuracy, steps and cost per case. Refresh it quarterly, add Spanish and Portuguese tasks, and let others submit runs.

    The directory becomes a source others cite, it backs up the routing and cost story with Runtime's own data, and it brings in exactly the technical buyers who choose platforms.

    Risks and how to handle them

    • Benchmark seen as biased

      Publish tasks, graders and code openly.

    Evidence

    • Your websiteOpen-source models page (updated October 29, 2026): 17 open-weight models ranked for agent work, with Spanish and Portuguese versions and picks for reconciliation, chargebacks, KYC/KYB, AML, underwriting, monitoring and support; "Benchmark scores are reported by each model provider" and "are not independently verified by Runtime".Source
    • Your websiteRuntime Labs: "An independent applied AI lab for fintech operations" working on evals, routing, reliability and memory, with "outcome-based measures" per workflow, a four-step phased-autonomy path, and the offer to build the first agent "and measure it against the cases you already resolved".Source

Wildcards 2

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

  • 22

    Runtime already has a guide on "what is know your agent", and its product now produces a printable evidence report for each agent, with risk scoring and an audit log. Payment teams know KYC and KYB; the same idea applied to the agents doing their work has no common template yet.

    Medium impactMedium effort First result in 1 quarter CEO

    First action

    Draft the template from the evidence report's sections.

    How to measure it

    KYA template downloads

    Now
    –
    Target
    300
    By
    2 quarters

    Steps

    1. 1

      Draft

      From evidence report sections.

      CTO (co-founder) 1 week

    2. 2

      Review

      Customer compliance leads.

      CEO (co-founder) 1 month

    3. 3

      Publish

      Free download plus the guide.

      Founding engineer 1 day

    4. 4

      Show auto-fill

      Runtime filling the template.

      CTO (co-founder) 1 week

    5. 5

      Localise

      Spanish and Portuguese versions.

      CEO (co-founder) 2 weeks

    6. 6

      Track

      Downloads and demos.

      Founding AI Ops monthly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    CEO (co founder)

    Smallest test

    Share the draft with five compliance leads.

    Why it works

    Release a free, editable Know Your Agent template: agent purpose, owner, data access, approvals, risk score, evals, change log and evidence. Show how Runtime fills it automatically. Invite compliance leads at customers and partner banks to comment, and publish version 1.0 with their input.

    Whoever writes the template that compliance teams adopt becomes the default reference, and Runtime's product produces it out of the box.

    Risks and how to handle them

    • Seen as marketing, not a standard

      Keep it vendor-neutral and openly licensed.

    Builds on

    Evidence

    • Your websiteGuides index: ops-agent guides for Adyen, Finix, Lithic, Marqeta, Stripe and Unit, plus guides on partner-bank requests, Reg E dispute investigations, AI agent compliance in the EU, UK and US, and "what is know your agent".Source
    • Your websiteThe Brief, October 2026: Compliance (beta) with risk scoring, an audit log and "a one-page evidence report you can print" for each agent; data masking; Okta and Duo SSO; Microsoft Teams bots and a Microsoft 365 connector; a new dashboard. No customers are named, and the Releases page has no email signup or RSS feed.Source
  • 23

    Rain uses Runtime for payment ops and compliance, and "Also" for engineering investigations, PR reviews, QA testing and reporting. Runtime launched to developers (103 points on Hacker News, 303 GitHub stars) and still supports coding agents. At many fintechs, the stuck-payment escalation lands on an on-call engineer before ops ever sees a pattern.

    Medium impactMedium effort First result in 1 quarter CTO

    First action

    Ask Rain's engineers which investigations save the most time.

    How to measure it

    Engineering-led expansions

    Now
    –
    Target
    3
    By
    3 quarters

    Steps

    1. 1

      Learn from Rain

      Top engineering investigations.

      Forward Deployed AI Engineer 1 week

    2. 2

      Build the template

      Alert-triggered investigator.

      CTO (co-founder) 1 month

    3. 3

      Write the guide

      For payments platform engineers.

      CTO (co-founder) 1 week

    4. 4

      Reach developers

      GitHub, Hacker News and engineering posts.

      CEO (co-founder) 1 month

    5. 5

      Expand

      Introduce ops teams after the first incidents.

      Forward Deployed AI Engineer 1 quarter

    6. 6

      Track

      Engineering accounts and ops expansions.

      Founding AI Ops monthly

    First result
    1 quarter
    Cost
    Engineering time
    Owner
    CTO (co founder)

    Smallest test

    One engineering team trial at a current customer.

    Why it works

    Package a "payment incident investigator" for engineering teams: triggered by an alert, it pulls logs, processor status and ledger entries, drafts the root cause and opens a ticket or PR for review. Market it to the developer audience Runtime already has, then expand from engineering into the ops teams that share the incident.

    Engineers adopt tools faster than ops buys platforms, and an investigation agent in engineering creates an internal champion for the ops rollout.

    Risks and how to handle them

    • Muddies the payments-ops positioning

      Frame it as payment incidents, not general coding.

    Builds on

    Evidence

    • Your websiteHomepage Rain block: "Rain runs Runtime for" technical payment support, fraud and risk transaction monitoring, compliance, cybersecurity and payment operations, and "Also" engineering investigations, PR reviews, QA testing and reporting.Source
    • Hacker NewsLaunch HN (May 21, 2026, 103 points): "Runtime (YC P26) – Sandboxed coding agents for everyone on a team", priced as a flat platform fee plus compute with "no token markup" and a free tier; commenters found the demo video dizzying, reported 404s in the docs, were confused by the license and asked what it adds over free open-source sandboxes.Source
    • GitHubRepository linked from the site footer and cited as proof on the Enterprise page: README headline "Open-source sandboxes where coding agents build and deploy", 303 stars, 39 forks, topics such as claude-code, cursor and sandbox; no payments content.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.

ProgrammableDigital

Show it, don't just describe it

These ideas need something to show: a demo, a video, an animation, graphics or a deck. Digital turns screenshots or the site into all of them, in the company's own brand.

Make a demo (opens programmable.digital 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.

  • StackAdapt growth strategyAn AI-powered programmatic advertising and marketing platform (DSP) covering CTV, display, native, video, audio, DOOH, in-game, email…21 growth ideas
  • Plasmidsaurus growth strategySequencing as a service: whole plasmid, RNA-Seq, amplicon, whole genome, microbiome, genotyping and custom sequencing, ordered online and…22 growth ideas
  • Viktor growth strategyAn AI employee that lives in Slack and Microsoft Teams, connects to 9,800+ tools and does the work: reports, dashboards, code, campaign…23 growth ideas
  • V Beauty Pure growth strategyA West Palm Beach, Florida nail brand in black matte bottles: acrylic systems, AcryGel, gel polish, rubber base, nail art, brushes and…22 growth ideas
  • O'Reilly Media growth strategyA technology and business learning platform: 75,000+ books and videos from O'Reilly and nearly 200 publishers, live expert-led events…21 growth ideas
  • Mobbin growth strategyA design reference library of 621,500+ screens, flows, animations and UI elements from real mobile apps, web apps and sites, with AI…21 growth ideas
  • Tinuiti growth strategyThe largest independent full-funnel media agency in the US: commerce and Amazon, retail media, search, social, TV, audio and display…21 growth ideas
  • BrainPOP growth strategyAnimated educational movies and learning activities for K-12 across subjects, with BrainPOP, BrainPOP Jr., BrainPOP Science, BrainPOP ELL…20 growth ideas

See every playbook and website review

Independent research. Programmable wrote this playbook from public sources, as they stood on November 10, 2026, and cites them throughout. Runtime did not commission, review or endorse it, and Programmable is not affiliated with Runtime. 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 Runtime runs its business.

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

30/60/90 roadmap

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

  1. Days 1 to 306 ideas

    Proof and paths: the Rain case study with real numbers, the LATAM demo route, a published pilot offer, clean-up of the coding-agent pages, clear logo labels and mobile hero speed.

  2. Days 31 to 6012 ideas

    Scale the proof and the motion: a workflow results library, a sample evidence report, processor and partner-bank channels, internal agents as templates, customer-built second agents, a newsletter, a pricing page, SOP teardowns, a better demo intake, comparison-page pruning and a health agent.

  3. Days 61 to 905 ideas

    Commit to the big bets and test the wildcards: a self-serve sandbox, outcome-based pricing, a public payments-ops benchmark, a Know Your Agent template and an engineering-team wedge.

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 agents that run in their own cloud, with approvals before money moves. The most cited problem: stuck payments, reconciliation breaks and alert queues that grow with volume.

  • Engineers who build the first agent with the team
  • Compliance and partner-bank scrutiny of any automation
  • In their words: “SOP”

Market

Competitive landscape

4 competitors compared: Taktile, Unit21, Sardine and Building in house.

  • Gap to exploit: One harness across payment ops, risk, compliance and support instead of a point tool per queue
  • Taktile: Centred on onboarding and underwriting decisions rather than broad payment ops
  • Unit21: Limited to financial crime, not payment operations or support

Hiring signals

Runtime is a 7-person team hiring two founding roles that show its model: forward-deployed engineers instead of account executives, and an AI Ops lead who will run customer success, renewals and internal operations on…

  • Forward Deployed AI Engineer: Embeds with payments customers and trains them to build their own agents. Ideas 7, 11, 15, 23.
  • Founding AI Ops: Customer health, renewals and internal agents on Runtime. Ideas 10, 13, 18.

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 first (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)
  • 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