Six FTE worth of doc review, returned to partners as billable strategy.

Three practice groups ran their own document-review playbooks and none of them talked to each other. We unified the intake and review pipeline behind an AI-first stack, and gave the partners back the hours they actually wanted to bill.

What was actually broken.

The firm's commercial-litigation, M&A, and regulatory practices each ran their own document-review teams. Three different vendors, three different review platforms, three different QC standards. Junior associates spent 60-70% of their time on first-pass review, partners spent their weekends doing privilege checks the juniors should have caught, and the firm was writing off roughly $1.1M a year in unbillable review time. The managing partner had tried to mandate a single vendor twice and lost both battles to practice-group autonomy.

What we did, and what we deliberately did not do.

We did not try to force a single vendor. We built a thin AI-first review layer that sat in front of all three existing platforms and unified the output. Tagging, privilege screening, and responsiveness coding ran through a deterministic LLM pipeline trained on the firm's own redlines from the prior three years. Partners reviewed flagged-only at the end. We kept the practice-group playbooks intact — only the first-pass labor changed. We also rebuilt the intake form so the matter started with the right metadata, which alone cut downstream rework by a third.

“We assumed AI doc review would force us to standardize the practices. The opposite happened. Each group kept its playbook and got faster. The partners stopped doing first-pass on weekends. That alone paid for the engagement.”

What the numbers did, twelve months in.