AI strategy for a $150M professional services firm

You're a $150M partner-owned professional services firm with 380 staff across two offices. The work is document-heavy, judgment-heavy, and billable by the hour or by project. Partners are watching three things at once. Public coverage of "AI replaces lawyers / accountants / recruiters" is loud. Clients are starting to ask whether you use AI in their matters and whether the rate stays the same if you do. And competing firms (a few aggressively, most cautiously) are running pilots that look smart in pitch meetings but haven't yet shown up in margin or win-rate numbers. The board wants a position by next quarter. The right position depends on what AI is actually good at today versus what it's still bad at, and on partner economics, which is a different conversation than tooling.

Pick option A. Fund document-and-workflow AI now. Defer client-facing AI by a year. Treat partner economics as a separate conversation.

The honest read on AI in professional services in 2026: it's mature enough to compress associate-level document work (research, summarization, draft generation, intake processing) by 30 to 50%, and it's not yet mature enough to put on a client-facing surface where a hallucinated answer becomes a fee-earner's signature. Vendors will tell you both surfaces are ready. They aren't.

The work to fund this year is internal. Pick two or three workflows that consume the most associate or analyst hours (legal research, document review, financial statement analysis, candidate screening, depending on your service line), pilot a vendor in each, measure realization, and roll the winners firm-wide. That's a $400K to $900K capital commitment for the year and a real margin lift in year two. Defer the client-facing AI program until the hallucination rate on commercial models drops or the audit / liability framework matures, whichever comes first.

On partner economics: the question "does AI replace partners" is not a tooling question. It's a question about what gets billed, who keeps the margin, and how new associates train when AI is doing the work that used to teach them. That's a partnership-level discussion that deserves a separate facilitated session, not an answer in an AI strategy memo. Don't conflate them.

Year-one investment, three to four pilots: $400K to $900K. Includes vendor licensing ($150K to $400K depending on tool count and seat count), implementation and integration ($100K to $200K), partner-led change management ($50K to $150K), and 0.5 FTE program owner inside IT or operations ($100K to $150K loaded).

Expected year-one realization: $700K to $1.6M. The lift comes from associate or analyst hours redirected from low-margin work to billable engagement work. At a $150M firm, a 30% productivity lift on the targeted workflows recovers roughly $1M to $2M of capacity in a typical service line. Conservative case captures a third of that as margin.

Year two onward: 1 to 3 percentage points of EBITDA margin. Cumulative, assuming the right pilots scale firm-wide.

What this is not: a path to revenue growth. AI does not bring you new clients in year one. The investment case is margin and talent retention, not top line.

1. Quarter one. Stand up an AI governance committee (a managing partner, a service-line lead, head of IT, head of risk). Set a pilot budget of $400K to $900K. Define disclosure policy for client engagement letters.
2. Quarter one. Pick two or three workflows. Run vendor evaluations. Sign one-year (not three-year) pilot contracts.
3. Quarter two. Pilots run with measurement (hours saved, error rate, fee earner satisfaction).
4. Quarter three. Pilot review. Kill the losing tool. Scale the winners firm-wide.
5. Quarter four. Retrospective. Update the AI policy. Set quarter-one budget for year two.
6. Year two. Reopen the client-facing AI question with one year of internal data behind you.

• This week: pull the timekeeper data for the last two quarters and identify the three workflows that consume the most associate or analyst hours. The pilots target those, not whatever the vendors are selling.
• Next two weeks: schedule a 90-minute partnership session on AI disclosure to clients. The position you take there gates which pilots are viable.
• Next month: name a program owner (not a partner, not an IT director, someone closer to the work). The firms that fund AI but don't name an owner end the year with three vendor invoices and no measurable change.