Happy to answer questions about:
- The 17 stagegates
- How risk scoring works (deterministic, not black box)
- Regulatory requirements (FCA/MLR 2017)
- What works vs what doesn't
Built this to test if foundation models can commoditize compliance middleware.
Thanks for asking! Honestly, I'm testing a thesis more than building a product.
The thesis: If foundation models can reason through structured workflows, a lot of "middleware" categories (compliance, GRC, back-office automation) are just orchestrating free public data + applying published formulas. The "platform" becomes commoditized.
For KYC specifically:
- Data is free (OFAC, UN, EU sanctions lists, Companies House, etc.)
- Formulas are published (MLR 2017, FinCEN CDD rules)
- Workflows are well-documented
So what are teams paying £60K/year for? Orchestration + audit trail.
If Claude can orchestrate and markdown can audit.. the economics shift dramatically.
Goals:
1. Prove open source can compete with commercial platforms (for standard workflows)
2. Make compliance accessible to smaller teams who can't afford £60K licenses
3. Test if this pattern applies to other regulated categories (legal, accounting, HR compliance)
Not building a company or raising money. Just want to see if expertise-as-code can disrupt vertical SaaS in regulated industries.
What do you think? Does this pattern apply to other categories you've seen?
Author here.
Quick context on the pilot:
30-day test with 5 compliance analysts at UK fintech doing standard individual onboarding (Know Your Customer checks).
Before: 95 minutes per case (manual searches across OFAC, UN, EU sanctions lists, Companies House, adverse media, PEP databases)
After: 27 minutes per case (Claude orchestrates the searches, analyst reviews at 17 mandatory checkpoints)
Key architectural decision: NO auto-approvals. Every decision requires explicit analyst approval + notes.
Legal team spent 3 weeks reviewing before approving pilot. Main concern was audit trail - solved with immutable markdown logs.
Demo slides show the full workflow: https://github.com/vyayasan/kyc-analyst/blob/main/docs/demo-...
Happy to answer questions about: - The 17 stagegates - How risk scoring works (deterministic, not black box) - Regulatory requirements (FCA/MLR 2017) - What works vs what doesn't
Built this to test if foundation models can commoditize compliance middleware.
This looks interesting, what's your goal with this project?
Thanks for asking! Honestly, I'm testing a thesis more than building a product.
The thesis: If foundation models can reason through structured workflows, a lot of "middleware" categories (compliance, GRC, back-office automation) are just orchestrating free public data + applying published formulas. The "platform" becomes commoditized.
For KYC specifically: - Data is free (OFAC, UN, EU sanctions lists, Companies House, etc.) - Formulas are published (MLR 2017, FinCEN CDD rules) - Workflows are well-documented
So what are teams paying £60K/year for? Orchestration + audit trail.
If Claude can orchestrate and markdown can audit.. the economics shift dramatically.
Goals: 1. Prove open source can compete with commercial platforms (for standard workflows) 2. Make compliance accessible to smaller teams who can't afford £60K licenses 3. Test if this pattern applies to other regulated categories (legal, accounting, HR compliance)
Not building a company or raising money. Just want to see if expertise-as-code can disrupt vertical SaaS in regulated industries.
What do you think? Does this pattern apply to other categories you've seen?