Most firms only earn when the client calls. AI is making that the worst position to hold.

Episodic delivery

The project

What AI does

AI compresses the delivery time on every project. Same scope, fewer hours, smaller invoice. Selling overhead and ramp-up cost stay fixed. Each new engagement requires re-scoping, re-staffing, and re-learning the client’s situation. The math gets harder every quarter, because per-engagement revenue keeps shrinking while the cost of starting one stays put.

What’s working

Move one offering off project-mode, not the whole book. Pick the work clients buy from you again and again, the kind that has gone almost routine, not your most impressive engagement. Routine is what you can systematize and price as a continuous service. Try to convert three offerings at once and you get three half-built ones. Genuinely one-off work stays episodic: M&A, litigation, audits with hard cutoffs.

On-call delivery

The standby

What AI does

On-call is dressed-up reactive work. The retainer fee creates the appearance of continuous engagement; the underlying delivery still triggers when the client asks. As clients use AI for the routine questions they used to call you for, the retainer feels softer to them every quarter. They start asking what they’re paying for. The retainer that earns its keep names what the firm produces, not what the client can call about.

What’s working

Convert the standby fee to a named output the client opens and uses: a quarterly strategy memo, a scenario analysis, a board-ready briefing. Pick one that lands on a fixed date so the value is visible every cycle, not inferred. A retainer the client never sees a deliverable from is the one they cancel when they ask what they pay for. The fee can stay. What’s billed has to change. If the line item describes hours of access, AI will commoditize it.

Continuous delivery

The watch

What AI does

AI handles the monitoring at near-zero marginal cost. Anomalies, regulatory changes, transaction patterns. The system flags them when they happen, not at the next quarterly meeting. The firm’s role shifts from delivering work to interpreting signals. The economics flip with it: continuous monitoring revenue compounds while episodic revenue resets to zero.

What’s working

Pick one client-relevant signal AI can watch reliably. Build a monthly retainer around the visibility, and write the boundary into the agreement before it starts: what the retainer covers, what triggers a separate engagement, how fast you respond. The thing that kills these retainers is not weak monitoring. It is the slow pile of off-scope “quick questions” that turns a margin into free work. Define the edge of the service first. Then the recurring revenue holds.

Proactive delivery

The first call

What AI does

Proactive engagement inverts the buying dynamic. There’s no RFP, no comparison shopping, no proposal process. The firm has the context, the data, and the recommendation. The client didn’t go to market. The competition isn’t in the room. AI makes this affordable for firms of any size: the monitoring that used to require a dedicated team now runs continuously at near-zero cost.

What’s working

Build the workflow so AI surfaces candidates and your people decide which one earns a phone call. Set the bar high: call only when being wrong would actually cost the client something. The first few calls set what your name means when it shows up on their phone. Call with noise and they learn to ignore you, the way every team buried in false alerts learns to ignore the alerts. A finding that changes what the client does protects the next call. That judgment gate is the offering, not the monitoring underneath it.