The KYC copilot economics: build vs. buy vs. hybrid.

KYC is the highest-friction, highest-cost, highest-attrition process in retail finance. In a typical UK brokerage, KYC costs between £25 and £120 per completed account, with a 30–55% drop-off rate at the document-collection step. Improving either number improves the P&L visibly. Doing both changes the shape of the business.

This is why every board with an AI budget now has a KYC copilot on the roadmap. The mistake most of them make is deciding build, buy or hybrid before understanding what each choice actually costs to run at their volume, jurisdictions and risk appetite. This note lays out the honest economics.

The buy path

Buying a KYC-as-a-service vendor is the fastest path to a working system. Best-in-class vendors — the seven or eight names most firms already know — will be live in your environment within eight to ten weeks and give you a real reduction in operator time from week one.

The buy-path economics come in at £3–£8 per KYC check for verification, £0.50–£2 for ongoing screening, plus a platform fee. At 10,000 accounts per year that is £40k–£100k in variable cost. At 100,000 it is £400k–£1m. The vendor’s roadmap is not yours, the pricing lever is not yours, and the compliance envelope is theirs. For firms with narrow volume, narrow jurisdictions and vanilla risk appetite, the buy path is the correct answer.

The build path

Building a KYC copilot in-house costs, in our observed range, £250,000 to £600,000 for a production-grade first release and £150,000 to £300,000 annually to operate. That is the AI-copilot layer only — the underlying identity-verification providers (Onfido, Jumio, Veriff, Sumsub, etc.) are usually bought regardless.

The build-path advantage is that the compliance envelope is yours, the workflow is yours, the model choices are yours, and every improvement compounds. Firms with high volume, multi-jurisdiction complexity or a genuinely differentiated risk appetite recover the build cost in under twelve months and then run at a fraction of the per-check cost of the buy path.

The build-path failure mode is that in-house teams underestimate the compliance layer. They ship a good extraction model and a decent workflow, and then the CCO refuses to sign off because there is no policy layer above the model and no evals layer beneath it. That failure is not a build-path failure. It is an architecture failure. Fixing it is what this firm does, but it is cheaper to design in from day one.

Where the hybrid wins

For most mid-sized UK firms, the honest answer is hybrid. Buy the identity-verification substrate. Build the copilot on top. Here is what that shape includes.

• Identity verification and biometrics: bought. Well-served by two or three commodity vendors, no meaningful differentiation available.
• Sanctions and PEP screening: bought. Same reason. Data quality is the moat, and no in-house firm will match a specialist.
• Document extraction and structured output: built. Domain-shaped, and the shape you want is different from every other firm.
• Policy layer and rules engine: built. Non-negotiable — this is the compliance envelope.
• Analyst workflow and case management: built. Every firm’s risk appetite is different and this is where the workflow either helps or hurts your ops team.
• Evals and monitoring: built. The vendor cannot score outputs against your rubric because they do not have it.

The unit economics, plainly

For a UK firm processing 30,000 new accounts a year across two jurisdictions with a moderate risk appetite, we have modelled all three shapes side by side in three separate engagements. The results converge tightly.

Buy: total annual cost £220k–£310k, per-account cost £7.30–£10.30. Vendor pricing lever not yours. Time-to-first-decision median 4 minutes.

Build: total annual cost £400k in year one (including amortised build), £220k in years two onward, per-account cost £7–£13 year one falling to £6–£8 by year three. Full compliance envelope ownership. Time-to-first-decision median 90 seconds.

Hybrid: total annual cost £310k year one, £220k thereafter, per-account cost £7–£10 falling to £5.50 by year three. Vendor pricing lever on the substrate only, not the workflow. Time-to-first-decision median 70 seconds.

The hybrid wins on total cost from year two, wins on time-to-first-decision from day one, and wins on compliance posture at every point. For most firms in this shape, it is the honest recommendation.

When each shape is wrong

The buy path is wrong when the vendor cannot cover your jurisdictions, when your risk appetite differs materially from their default rules, or when your KYC volume is high enough that the per-check cost dominates. The build path is wrong when volume is too low to justify build cost, when the internal team does not have AI depth, or when the roadmap can be met by a vendor without differentiation for the business. Hybrid is wrong when the firm cannot resource ongoing maintenance of the built components — a hybrid that becomes an unmaintained monolith is worse than a clean buy.

The single most important step is deciding honestly which shape you are in before writing the first line of code or signing the first contract. Two weeks of diligence here saves twelve months of rework.