AI client work for consultancies: a practical playbook
5 minutes read
Consultancies feel the generic-AI-draft problem acutely, because the output goes straight to a paying client. A practical playbook for making AI work you can bill for.
Consultancies feel the generic-AI-draft problem acutely, because there is no internal buffer. The output of the work is the thing the client pays for. A vague, unsourced draft that a product team might quietly fix internally is, for a consultancy, the deliverable itself. That raises the bar on what "AI helped with this" is allowed to mean.
This playbook is about clearing that bar: using AI on real client work in a way you can review, defend, and put your name on.
Why consultancy work raises the bar
Client-facing AI work has to survive a client's scrutiny, not just an internal one. A few things make it harder than general AI use.
A research summary, a strategy, a first draft of a report goes to a client paying for the firm's judgement, so there is no room for a confident claim nobody can support. Every client is different, which means a draft from generic knowledge is generic by default. And consultancies handle other firms' material: pasting it into a public chat tool to get a faster draft is a client-trust problem waiting to surface.
Consultancies that use AI well move faster on the parts of delivery that used to eat senior hours. The problem is that the way most people use AI, a chat window and a generic prompt, is the wrong shape for client work.
The playbook
The goal is a first draft a principal can review quickly and stand behind. These moves get you there.
| Move | What it fixes | Result |
|---|---|---|
| Work from approved sources per client | Ungrounded claims, client-data risk | Drafts built on trusted material under agreed access rules |
| Keep evidence visible for review | Slow, full-rewrite review | Each claim shows its source; review becomes a check |
| Keep the senior reviewer accountable | AI output going out unchecked | A person signs off on every deliverable |
| Reuse what works across clients | Rebuilding the same process each time | One good project becomes a repeatable method |
Work from approved sources, per client
Decide, for each client, what the AI may draw on: the external sources you trust, your firm's prior work and methods, and the client material the engagement permits. This keeps the draft sourced and keeps client data inside the rules you agreed to. It is the single biggest difference between AI you can bill for and AI you have to hide. The connected knowledge guide covers how a firm's own stack and data feed this.
Keep the evidence visible for review
The account lead reviewing a draft should see where each claim came from and which parts are assumptions the model made. That is what turns review from a rebuild into a check. Without visible evidence, the reviewer has to re-source everything, and the time AI saved on drafting comes straight back out at review.
Keep the senior reviewer accountable
No consultancy should let an AI draft reach a client unreviewed, and no tool should imply otherwise. The value is a faster route to a draft the reviewer can trust, not the removal of the reviewer. The person who signs off is the person who stands behind the work. The complete guide to reliable AI client work makes this case in full.
Reuse what works across clients
Consultancies run the same shapes of work repeatedly: a research brief, a positioning analysis, a first draft of a report. Once a method produces a reliable result, save it as a reusable workflow so the next client starts from a proven process rather than a blank window. This is how one good project becomes consistent quality across the team. See reusable AI workflows for how that works.
What about cost per client?
Cost matters for consultancies because AI usage attaches to client work, and you need to know what each engagement actually costs. Hebno makes AI usage visible by team and client, with no surprise markup on the models behind the work. That is real and useful, and it is deliberately not the headline. Controlling cost on work you cannot deliver solves nothing. Reliable delivery comes first; cost visibility rides along with it. Pricing is a tailored proposal — request a demo to discuss what that looks like for your team.
Getting started
You do not need an engineer or an IT project to start. Pick one recurring kind of client work, define the approved sources for it, and run a real brief through the reliable-work sequence: sources, draft with visible evidence, expert review, sign-off. If the reviewed draft is faster to a deliverable you can stand behind, expand from there.
