Every compliance leader has lived through the migration. Twelve weeks of vendor onboarding, mapping fields between the new case-management system and the old, rewriting SOPs to match the new tool's vocabulary, retraining analysts on a new UI two months before the audit. The new system is better. The migration was not. By the time the team is fluent, half the original justification has eroded.
The agentic AI compliance category has a fork in the road. One side asks you to migrate. The other side does not. This is the load-bearing positioning question — and it deserves more attention than it usually gets in vendor decks.
The migration tax
Every replacement compliance system has the same tax structure. There's the obvious cost — license, integration fees, training. And there's the invisible cost: the operational drag of running two systems in parallel during cutover, the friction of losing institutional knowledge encoded in the old SOPs, the audit risk during the transition window when neither system is fully canonical.
Cumulative migration tax for a typical mid-market VASP swapping case-management vendors is six to nine months of effective compliance throughput loss. That's a quarter or more of cases that move slower because the team is half-on the new system. For a regime tightening — AUSTRAC AML/CTF reform, MiCA enforcement, GENIUS Act issuer requirements — the tax is not theoretical. It is regulatory exposure.
The opposite approach
An AI compliance agent that works inside your existing stack pays no migration tax. The agent reads from Chainalysis or TRM where you already operate. It pulls Sumsub records you already have. It writes case dispositions back to Salesforce, Zendesk, Unit21, Hummingbird, or your in-house case management. It posts handoffs to the Slack or Teams channel your team is already in. Your team's vocabulary, SOPs, and case taxonomy stay where they are. The agent layers on top.
This is not a smaller version of the migration story. It is a categorically different story. The pitch is not 'move to our system'. The pitch is 'we run inside your system'. The compliance team learns no new UI. The MLRO reviews evidence in the place she already reviews evidence. The audit trail extends, rather than replaces, the one already in production.
Competitive map
The agentic AI compliance category sorts cleanly along this axis.
Data providers — Chainalysis, TRM, Elliptic
Strong at tracing and risk signals. Stop at data. Don't orchestrate across tools. Don't draft SARs. Don't hold the evidence ledger for an examiner. Cogentic reads from them — does not replace them. They sit in the stack, and the agent reaches into them.
Tradfi AML platforms — Bretton, Unit21, Hummingbird
Strong AI orchestration for tradfi banks. Built around their case-management surface — adopting them is a migration. Onchain blind: no wallet attribution, no entity graph across wallets and legal entities, no tokenised-securities transfer logic. Wrong shape for crypto.
Legacy compliance suites — Actimize, SAS, Oracle FCCM, Fenergo
18-month deployments. Built for tradfi. Slow to ship AI agents with audit rails. Onchain blind. Migration tax is at its highest here — these systems take quarters to install before they produce value.
Travel Rule messaging — Notabene, Sygna, 21 Analytics
Protocol routing, not workflow. Cogentic ingests them as counterparty data sources rather than treating them as standalone surfaces. They're plumbing — the agent reasons over the data they expose.
Cogentic — the orchestration layer
AI agents inside the compliance stack you already run. Native case view available if you want one; bidirectional case-management write-back if you keep yours. Reporter-of-record model. SR 11-7 documentation by default.
The honest caveat
Working inside someone else's stack is harder engineering than greenfield. Real adapters to Chainalysis Reactor, TRM Forensics, Sumsub, Fireblocks, AUSTRAC goAML take time and customer ground-truthing. Second movers in this category can copy the architecture but not skip the integration work. That's the moat — and it's why most vendors in this category default to migration: the alternative is harder to ship.
Cogentic chose the harder path because the migration tax is the single biggest reason mid-market VASPs delay compliance investment. Removing that tax is the difference between a deal that closes in Q3 2026 ahead of AUSTRAC and a deal that closes in Q1 2027 after the reform has bitten.
Where this lands
The compliance leaders in our pilot conversations consistently tell us the same thing: their team's velocity isn't constrained by alert quality. It's constrained by the time it takes to gather context across five tools per case and the cognitive cost of switching between them. An agent that pulls context from those five tools and presents it inline solves the constraint without forcing the team to change tools.
If the agentic AI compliance category converges on this approach, the next two years of the market move toward stack-native agents. If it converges on the migration approach, the market repeats the last decade of tradfi compliance — slow, expensive, and underwhelming. The bet on stack-native is the bet on a faster, cheaper compliance market.