Most law firms don’t set out to build a compliance platform. It starts with a general-purpose AI tool pointed at conflicts, a script someone wrote to help with intake, a spreadsheet that tracks outside counsel guidelines. None of it feels like a platform decision. Until two years pass and a senior analyst leaves, the script breaks, and nobody is sure who owns fixing it.
General-purpose AI doesn’t know what a true conflict is
Conflicts clearance, matter intake, and client obligations management run on a data model that general-purpose AI doesn’t have. That model encodes what counts as a true conflict for your firm, why a matter was passed, which relationships carry institutional weight, and how your rules apply when a new matter touches three jurisdictions and a prior lateral hire. A general-purpose model has none of that context. Pointing it at your compliance data doesn’t give it the legal compliance model — it gives you confident-sounding output with no audit trail and nothing defensible underneath it. Encoding a firm’s vocabulary, obligations, and methods into a general agent is years of work from scratch. That work is already done.
Every change to regulations, client requirements, or AI models becomes your team’s problem
Assume the build goes well on day one. The maintenance problem starts immediately. Client guidelines change. New AI model versions ship. Bar rules update. Every one of those changes requires someone inside the firm to own the update, test the logic, validate the output, and document the rationale for when a regulator or client asks. That’s permanent overhead — staffed by people whose time is almost certainly better spent on higher-value work.
97 of the AmLaw 100 run on Intapp. That installed base represents two decades of exactly this work: building and maintaining the compliance data model, the legal-specific architecture, the workflow integrations. Firms choosing to build instead are spending time and engineering resources to get to where the category already is, while bearing the maintenance burden indefinitely after arrival.
The governance risk doesn’t show up in a project budget
Every firm leader eventually has the same conversation about a firm-wide compliance agent: “How do I know it will respect the ethical walls, not hallucinate, not surface the wrong data, and produce an audit trail I can defend to a court or a partnership vote?” General AI vendors can’t answer those questions, because the answers require an architecture the firm runs on — not a model with a wrapper around it.
General AI keeps the content of a document and strips the metadata around it. In a firm, the metadata is the compliance posture. Which partner is walled off? Which deal team has MNPI? Which documents are on litigation hold? Strip those tags and you haven’t built an assistant — you’ve built a breach. There’s also a second exposure in the same approach: to make a firm’s documents usable, general AI copies them into a new store outside the firm’s retention rules and access controls. A discovery request reaches it. A retention schedule does not. 33% of firms now use AI firmwide, up from 17% in 2025. In that environment, 76% of professionals still use AI tools their firm hasn’t sanctioned. A homegrown tool built on general AI doesn’t close that gap. It widens it in a different direction.
The opportunity cost is what closes the argument
Intapp Compliance with Celeste deploys agents that draft intake requests, construct conflicts search strategies, match corporate trees, and ingest outside counsel guidelines — absorbing the routine work that rising volume used to demand more people for. Every quarter a firm spends building is a quarter a competitor widens that gap.
The build decision doesn’t just cost time and engineering. It costs the compounding capacity advantage that starts the day a competitor goes live.
See how Compliance with Celeste helps your firm accept more business, faster — without adding headcount or compromising risk. Get a demo.