Generative AI for Professional Services: Drafting, Knowledge, and Research

A law firm deploys a commercial AI tool for document drafting. Within weeks, several issues surface. An associate uses the AI to draft a motion and the AI generates case citations that don't exist — fabricated citations that would sanction the associate and embarrass the firm if filed. A partner uses AI to draft a client memo and the AI surfaces information from matters the partner technically had access to but shouldn't have been using — crossing an ethical wall. An accounting firm's auditor uses AI for workpaper drafting and the AI mixes information across clients in ways that violate the confidentiality and potentially the independence rules PCAOB examines. A consultant uses AI for a client deliverable and the AI uses content from a prior engagement at a competitor that creates a confidentiality breach. Each is a consequence of generic AI that doesn't know firm-specific ethical walls, independence rules, or confidentiality obligations — and each carries liability that no productivity gain can offset.

Professional services generative AI done right is grounded in firm precedent with strict access controls and refuses to fabricate. Document drafting grounded in firm precedent with conflicts wall enforcement — the AI retrieves from approved firm content the user has access to, cites specific sources, and refuses to generate citations it can't verify. Client deliverable drafting with firm-specific voice and the privilege discipline the firm applies. Research acceleration with current law/regulation retrieval and cited sources. Refusal patterns preventing hallucination — citations must resolve to actual sources, statistics must resolve to actual data, quotes must resolve to actual text. With independence-aware design for accounting firms and ethical wall enforcement for law firms. Done with this discipline, generative AI accelerates drafting and research. Done casually, it creates malpractice or independence exposure.