Artificial Intelligence for Professional Services: Matter, Engagement, and Knowledge

A Big Law firm's data team builds a matter analytics model predicting realization risk. The model scores well on historical data. The Managing Partner reviews and the questions surface that determine whether the model actually deploys: does the model's training data contain attorney work product that creates privilege implications, how does the model handle matters with ethical walls between practice groups, can it surface insights to partners without exposing the reasoning that might itself be protected, and does its feature set include any conflicts-sensitive client information. Meanwhile at the accounting firm, a similar engagement risk model raises independence concerns — the SEC and PCAOB independence rules prohibit certain services and data uses for attest clients, and a model that pools data across clients could create independence violations that threaten the attest practice. These aren't academic concerns. They determine whether AI can go into production at professional services firms.

Professional services AI that deploys is designed around privilege and independence from the start. Matter and engagement analytics with client-specific data isolation — information from Client A's matter never influences analysis for Client B. Ethical walls between practice groups enforced at the data layer, not just the access layer. Independence-aware design for accounting firms with attest engagements where the independence rules limit what data and services can be pooled. Work product protection where matter analytics doesn't expose the reasoning patterns that constitute attorney work product. Knowledge retrieval with strict access controls matching the firm's conflicts walls. Model risk management that the General Counsel and Chief Risk Officer can defend. Done this way, AI delivers in professional services. Done as generic enterprise ML, it can't pass GC review.