Using AI as an Agency Owner
Where AI actually earns its keep in an agency owner's real week, three prompts built for scoping, pricing, and client-tier decisions, and which documented consulting frameworks are worth loading alongside it.
An agency owner's week is a run of small, high-stakes judgment calls that rarely feel like they deserve a framework: is this new request in scope or a change order, does this project actually clear margin at the rate we quoted, is this client relationship fine or quietly souring before the renewal call. Generic AI use treats each of these as a one-off question. The more useful use treats them as the same few recurring decisions, run the same structured way every time.
What does an agency owner's actual week look like, AI included?
Most weeks carry the same five decisions back to front: whether a new client request is genuinely in scope or should trigger a change order, whether a proposed project actually clears the agency's target margin once real hours are counted, whether a client relationship that's gone quiet is a problem or just a quiet month, whether to take a marginal-fit client because cash flow is tight this quarter, and how to price a request when the client asks for "just your day rate." None of these are exotic. All of them get decided under time pressure, often on gut feel, because there's rarely a spare hour to run the numbers properly, and the cost of getting any one of them wrong rarely shows up until months later.
Where does generic AI actually help here, and where does it quietly fail an agency owner?
Unaided, it's genuinely useful for drafting scope language precisely, generating comparison structures for a proposal, and running a margin calculation fast once you supply real hours and real rates.
It's weak at anything requiring knowledge only you have: whether this particular client is a referral goldmine worth carrying at a loss for a quarter, whether your team actually has capacity next month regardless of what the calendar shows, whether a relationship that's gone quiet is genuinely at risk or just going through a slow patch. Feed it a request without that context and it will answer confidently anyway, since a lack of context has never once stopped a language model from producing a fluent-sounding response.
What are the three prompts worth running this week?
These run in Claude, ChatGPT, or Gemini as written.
"Here's the original scope of work: [paste it]. Here's the new request the client just made: [describe it]. Is this covered under the original scope, or does it warrant a change order? Be specific about which line it falls under or doesn't."
Why it works: most scope disputes happen because nobody checked the request against the actual written scope in the moment, and a specific, line-by-line answer is much harder to argue with later than a gut feeling.
"This project is quoted at [price]. My best estimate of hours required is [hours] across [roles and rates]. Given our target margin of [percentage], does this project actually clear it? Show the math."
Why it works: a project can look profitable on the invoice and still be underwater once real hours are counted honestly, and forcing the actual math surfaces that gap before the contract is signed, not after.
"Here's what I know about each of these clients: [revenue, margin, hassle level, referral value, relationship history]. Sort them into keep, grow, or reconsider, and tell me which specific input is driving each call."
Why it works: naming which specific input, margin, hassle, referral value, is actually driving each tier forces a real answer instead of a vague overall impression, and it surfaces disagreements between inputs (high revenue, low margin) that a gut read tends to smooth over.
What's the honest limit here, even with all three run?
Run all three and this week's scope, pricing, and client calls will be sharper than gut feel alone produces. What you won't have is a system that knows your actual bank balance, your team's real capacity next month regardless of what the calendar says, or which difficult client is quietly worth keeping for the referrals alone. The gap isn't this week's decisions; it's that no prompt substitutes for knowing your own business from the inside, the way a decade of running it teaches you to read a client's tone before they've said anything worrying at all.
Which documented frameworks actually fit an agency owner, and why?
David Maister's Trust Equation, Trustworthiness equals Credibility plus Reliability plus Intimacy, divided by Self-Orientation, developed jointly with Charles Green and Robert Galford in The Trusted Advisor, gives a structured way to diagnose why a client relationship feels off before it becomes a churn problem. Blair Enns's documented case in Win Without Pitching argues against giving away strategic thinking for free to win business, directly relevant to the "just send us a proposal" requests that eat unpaid hours. David C. Baker's positioning discipline in The Business of Expertise makes the case that a narrowly positioned agency commands better fees and better-fit clients than a generalist one chasing every brief.
Each addresses a different recurring failure mode: Maister for the relationship that's quietly eroding, Enns for the free-work trap, Baker for the agency that's underpriced because it's positioned too broadly to justify a premium. None of the three requires abandoning the others; a client-tier conversation might need Maister's diagnostic, while the proposal for a new prospect leans on Enns's refusal to pitch for free.
How do you actually load one of these into Claude, ChatGPT, or Gemini?
In Claude, each installs as a Skill: go to Settings → Skills → Add skill and upload the .zip. Claude reads the skill's description and auto-invokes it when your question matches, or you can force it with /david-maister-framework at the start of a message. Claude Code users get the same thing via ~/.claude/skills (every project) or a project's .claude/skills folder (that project only).
In ChatGPT, the plain .md file goes into a Custom GPT: Create a GPT → Configure → Instructions, paste the framework's content in. Standard Custom Instructions (the "About me" fields) cap out at 1,500 characters, too short for a full framework, so use a Custom GPT specifically, not Custom Instructions.
In Gemini, paste the .md content directly into the system prompt if the interface you're using supports one, or lead your first message with it: "Here is a thinking framework to apply throughout this conversation. Read it carefully, then answer using this framework's approach," followed by the file's contents.
Where should an agency owner actually start?
There's no bundle built around this specific mix of consulting-specific frameworks, the existing 10-persona stacks skew toward generalist founder thinking rather than fee structuring and positioning, so the Consultant & Advisor category itself, Maister, Enns, and Baker alongside Peter Block and Alan Weiss, is the better starting point than any current bundle. For the client-tier call specifically, Client Risk is a $149 tool built to run that judgment, including the retain-or-fire decision, across your whole client book.
All the copy-paste prompts from this guide, and the rest of the AI Practice series, live at /prompts too, free to copy without reading the guide first.
Frequently asked questions
Is this the same as the other Consultant & Advisor content on the site?
It's built to sit alongside it, not repeat it. The category's frameworks distill how Maister, Enns, and Baker think about trust, pricing, and positioning in general terms. This guide is the applied layer, three specific prompts for the specific decisions an agency owner actually faces this week, plus a pointer to which of those frameworks to load for which situation and why that one specifically.
Can AI actually tell me if a client relationship is at risk?
It can help you structure a read on one if you describe what's actually changed, slower replies, a new stakeholder cc'd, a renewal conversation being avoided. It has no independent visibility into the relationship itself, so the read is only as good as your own honest account of what's happening, not a signal it discovered on its own from watching the account the way you have.
Will using AI for pricing actually help me raise my rates?
It can help you catch a project that looks fine on revenue but is quietly under your target margin, which is a different and more immediately useful problem than rate-setting in the abstract. Whether you can actually charge more depends on your positioning and the market you're actually competing in, neither of which a margin calculation alone can fix for you.
What's the single biggest mistake agency owners make using AI here?
Asking it to decide whether to take a marginal client instead of asking it to lay out the actual tradeoff. A direct yes-or-no answer sounds confident regardless of whether it has the full picture, cash flow pressure, team capacity, referral value, none of which it knows unless you say so explicitly. Structuring the tradeoff is the more useful ask, and it's the one that actually respects what the tool can and can't know.
Does this apply if I'm a one-person agency, not a team?
Yes, arguably more directly. A one-person shop has no second opinion in the room when a pricing or client-fit decision comes up, which is exactly when a structured prompt against a documented framework is most useful, a stand-in for the colleague you don't have down the hall to sanity-check the call with before you commit to it and it's too late to undo.
Should the whole team use these prompts, or just the owner?
Whoever actually owns the client relationship and the pricing conversation should be the one running them, since the prompts depend on specifics, actual margin, actual history with this client, that only the person closest to the account reliably has. A junior account manager guessing at those numbers produces a confident-sounding but unreliable answer either way, regardless of how the prompt itself is worded.
What's the honest limit of using AI for agency decisions?
It has no access to your actual bank balance, your team's real capacity next month, or the referral value a difficult client might quietly be worth carrying at a loss. Use it to structure the tradeoff and catch a margin problem before you sign the contract; the judgment about your specific business, and the risk of being wrong, still has to be yours.
Written by Gareth Hoyle. Last updated 26 September 2026. Part of the authority.md guides library.
More guides.
Bringing AI Into a Creative Project Without Ruining It
What generic AI brainstorming actually gives a creative project and what it quietly flattens, four prompts that help today without taking over, and what changes when the AI applies Rick Rubin's documented listening practice instead of generic idea generation.
AI for Hiring Decisions: Where Claude and ChatGPT Actually Help
A grounded look at what generic AI use gets right and wrong when you're deciding on a hire, four prompts that genuinely sharpen the decision, and what changes when the AI applies a documented evidence-based hiring framework instead of generic interview advice.
Using AI to Prep for a Negotiation: What Actually Works
Generic AI negotiation advice is genuinely useful and genuinely limited. Here's what Claude and ChatGPT get right before a negotiation, four prompts that work today, and what changes once the AI has a documented framework instead of a generic one.