A Consultant's AI Workweek
Concrete, task-level use cases across a full engagement week — what AI drafts at every stage, and what you must own.
Knowing that AI can help is different from knowing exactly what to do with it on a Tuesday afternoon when you have 80 pages of interview notes and a deck due Thursday.
This lesson is specific. No theory. Each phase of a typical engagement week: what you hand to AI, what it gives back, and what you must own before anything moves forward.
Monday: Scope and Proposal
The engagement starts with the commercial anchor — scope, deliverables, timeline, and what success looks like. This is also where most consultants waste the most time on language that, at its core, follows the same structure every time.
What AI drafts. Give AI your notes from the scoping call — the problem as described, the constraints mentioned, the desired outcomes, and whatever the client said about timeline. Ask it to draft a scope-of-work section, a deliverables list, and a proposed timeline. It will give you something structurally complete and professionally worded.
What you must own. Defining what success actually looks like is not a drafting task. That requires you to read what was not said in the scoping call, identify where the stated problem and the real problem diverge, and make the strategic call about what is in and out of scope. The proposal language is AI's; the scope decision is yours. Your name is on the commercial terms.
Tuesday: Research and Data Room
Clients hand you documents — financial reports, market research, previous strategy work, operational data. Reading, summarizing, and cross-referencing that material is real work that used to justify real hours.
What AI drafts. Feed AI the documents (provided they contain no confidential client data you cannot share with a public tool — if they do, see the lesson on limits). Ask for a summary of each document, a list of key figures, and any contradictions across sources. For public or non-confidential documents, this takes minutes rather than days.
What you must own. Deciding what questions to ask the data is the work. AI surfaces what is there; you decide what matters. The insight that drives the hypothesis — "the distribution data contradicts the margin narrative, and that's the real problem" — is not in the summary. It is in the gap between what the data says and what you know about how this industry actually works.
Wednesday: Interview Synthesis and Hypothesis
You have a round of stakeholder interviews. The notes are a mess. The patterns are somewhere in there.
What AI drafts. Paste your interview notes — but only non-confidential material. Removing the company name is not enough on its own: absent a firm-vetted enterprise arrangement, you strip out the client-specific facts entirely and keep only the generic structure (see the lesson on limits). Ask AI to cluster themes, surface patterns across respondents, and produce a structured synthesis with emerging hypotheses listed. It will organize the material better and faster than most manual synthesis approaches.
What you must own. Deciding which patterns matter for this client is the judgment call. AI clusters what it sees in the text; you decide what is signal and what is noise given the context you carry from the client relationship. Forming the hypothesis — and stress-testing it against what you already know — is where the work is. Never let AI's synthesis become the hypothesis without your active scrutiny.
Thursday: Deck Draft
The deck is where most consultants spend the most time and where AI provides the most leverage.
What AI drafts. With a clear structure from Wednesday, ask AI to draft a slide-by-slide outline with body copy for each slide and suggested data visualization types. If you have already approved a hypothesis and narrative arc, the output will be structurally sound.
What you must own. Editing the deck is not optional review — it is the work. Cut every slide that does not earn its place. Rewrite the conclusion so it carries your recommendation, not a hedged summary of the analysis. The deck that lands in the exec meeting sounds like you: specific, directional, confident. That voice comes from the edit, not the draft.
Friday: Executive Summary and IP Capture
Two tasks. One faces the client. One is yours to keep.
What AI drafts for the exec summary. From the approved deck, AI can compress the narrative into a one-page or one-slide summary: situation, recommendation, and rationale. Give it the approved deck and ask for a tight summary. It will produce something that covers the ground.
What you own on the exec summary. The recommendation must be specific and owned. If the exec summary says "the company should consider options including..." it is not ready. The version you send says "the right move is X, because Y, and here is the first step." That language is yours.
What AI drafts for IP capture. Ask AI to help you distill the engagement into a reusable template: the framework you used, the discovery question structure that worked, the synthesis approach you refined. Strip every client-specific detail before asking. Give AI the non-confidential skeleton.
What you own on IP capture. Verifying that nothing client-specific, confidential, or NDA-covered made it into the template is your responsibility. The template encodes your method; it holds none of the client's facts.
The Pattern Across Every Phase
The matrix above makes the pattern visible: every consulting task has a scaffold (AI's job) and a judgment call (yours). The mistake is not using AI — it is skipping the judgment step and treating the draft as the deliverable.
A Note on Prompting
You do not need to become a prompting expert to use these approaches. The quality of what you get out scales directly with the quality of what you put in.
Three habits that make every AI interaction better:
Give context before asking. Tell AI what you are working on, who the audience is, and what the output is for before you ask for anything. A one-sentence setup cuts revision cycles.
Be specific about format. "Give me a slide outline" is vague. "Give me a 12-slide outline for a strategy presentation to a CFO, with a one-sentence message per slide and a suggested data type for each" is not.
React to structure before drafting content. Get the skeleton right before filling it in. Changing the narrative arc after the deck is written is expensive. Changing it at the outline stage costs nothing.
Continue at academy.jeremyknox.ai or the full collection at jeremyknox.ai.