The three categories of consulting work — and which ones AI should never touch
7 min readFriday, 4:15 PM. A proposal is due before the prospect cools off, a client deliverable keeps slipping, and Tuesday's meeting transcript is still unread. Claude gets a paragraph of instructions, two sentences of the answer make it into the doc, and the tab closes.
Time saved: maybe six minutes. Time spent: ten, counting the context switch.
That's the most common way AI goes wrong for a solo consultant, and it's worth naming first, because no prompt improves until it's fixed. The problem is using AI like a slightly faster Google. As long as that's the frame, the real time savings stay out of reach.
A better frame starts by accepting that not all consulting work is the same kind of work. Some of it is yours. Some of it is partly yours. Some of it shouldn't have been on your plate in the first place. AI's job is different in each case, and mixing them up is what burns the hours the productivity threads promised to give back.
The three categories
A consulting week breaks into roughly three kinds of work. They want different tools, and using the same tool on all three is how you end up saving six minutes on a Friday afternoon.
Category 1 — Work that requires your judgment
Diagnosing a client's actual problem underneath the one they described. Deciding what goes in scope and what to push back on. Reading a room. Naming the political dynamic the client is dancing around. Writing the one sentence in the deliverable they'll quote back to their CEO.
How much AI can do: almost none of it.
This is what your rate pays for. AI is bad at it and sounds confident anyway, which is a dangerous combination. The model doesn't have the context. It wasn't in the meetings, didn't see the body language shift when someone said "we're aligned on the priorities," and doesn't know which stakeholder actually decides. It will produce something fluent and unhelpful.
If you find yourself prompting AI to do Category 1 work ("diagnose the real problem based on these notes"), you're using it to launder a hypothesis you should sit with longer. Sit with it longer.
Category 2 — Work that has structure but needs your voice
Writing the proposal. Drafting the kickoff email. Producing the strategy memo. Turning a finished engagement into a case study. Writing the section of a deliverable that pulls your findings together. Drafting the follow-up that revives a stalled deal. Writing the newsletter post that brings in the next one.
How much AI can do: most of the first draft.
This is where AI works as a first-draft engine. You don't hand off the writing; you hand off the blank page. AI gets you to a rough draft, and you take it the rest of the way, which is the part the client is paying for.
With a clear brief and a sample of your writing, a section that would take you an hour and a half from scratch can take fifteen minutes to draft and edit. Across a week, that adds up to real hours, without lowering the quality of anything the client sees.
The risk here is a lazy brief, which produces AI-flavored prose a client can sense even if they can't say why. Edit hard, especially the opening and closing paragraphs, where AI writing gives itself away most.
Category 3 — Work that has structure and doesn't need your voice
Pre-call research on a prospect. Turning meeting notes into action items. A follow-up email that thanks the right person for the right thing. Four social posts from a finished case study. The boilerplate sections of a deliverable. Pulling the key facts out of a 90-minute call transcript.
How much AI can do: nearly all of it.
This is where AI works as an executor. You give it inputs, accept the output with a light edit, and move on.
This work tends to happen in fifteen- and twenty-minute pieces, which is why it's easy to miss how much of the week it takes. It's also the cheapest time to win back, because it needs structure rather than taste, and structure is what AI is good at. The shift worth making is to hand Category 3 work to AI by default rather than occasionally. AI does it unless there's a specific reason it shouldn't.
The three-question screen
Before you give any task to Claude or ChatGPT, ask three questions. If any answer is no, do the task yourself.
1. Can I describe the output I want in 100 words?
If you can't, you don't know yet what you want, and AI will helpfully invent something. Spend two more minutes on the task, write the brief in your head or on paper, and then come back. Most prompts that produce generic output do so because the brief was generic.
2. Can I tell at a glance whether the output is good?
If it's in your area of expertise and you'd spot a problem, yes. If it isn't ("draft the indemnity clause for this engagement"), no, because confident mistakes in your blind spot are how reputations get damaged. AI will produce a clause that reads correctly to a layperson and isn't enforceable, and you won't find out until it matters.
3. Is editing the output faster than writing it from scratch?
Sometimes it isn't. If editing the draft would take 80% as long as writing your own, write your own.
It's a thirty-second check. Run it before you open the tab.
What this looks like in a consulting week
Here's an illustrative Tuesday, tagged by category. Some of these are the same task at different points in an engagement.
| Task | Time | Category | What you do |
|---|---|---|---|
| Read a prospect's website and LinkedIn before a discovery call | 12 min | 3 | AI produces a one-page dossier |
| The discovery call itself | 30 min | 1 | You. Phone in your pocket. |
| Decide whether to propose for this deal | 5 min | 1 | You. A quick judgment call. |
| Write the proposal | 90 min | 2 | AI drafts; you edit the opening, the pricing rationale and "why me" |
| Draft the kickoff email after they accept | 8 min | 3 | AI writes; you tweak two phrases |
| Draft the strategy memo for an active client | 4 hours | 2 | You outline; AI expands sections; you edit |
| Decide what to recommend in the strategy memo | (inside the 4 hours) | 1 | You. Several walks. |
| Pull action items from yesterday's client meeting | 7 min | 3 | AI works from the transcript; you check it |
| Write a newsletter post about a wrapped engagement | 30 min | 2 | AI drafts from your project debrief; you sharpen the opening |
| Quarterly pipeline retrospective | 90 min | Mostly 1 | You, with AI finding patterns in the data, not drawing the conclusions |
A week run this way moves most of the Category 3 work off your plate, cuts the drafting time on Category 2 work, and keeps Category 1 work entirely yours.
A note on Claude vs. ChatGPT
Both work. We default to Claude because it commits to a verdict when asked and pads less on structured business writing. ChatGPT can be corrected with the right prompt; Claude needs less correcting. If you already pay for ChatGPT, the workflows in this article work there too, because the prompts don't depend on the tool. We've compared the two in more detail here.
What to do this week
Pick the next three tasks on your calendar that have a clear structure. Tag each one as Category 2 (needs your voice) or Category 3 (doesn't). Run the three-question screen on the Category 3 ones, and give the first one that passes to Claude with a 100-word brief. Time how long the edit takes, and compare it with how long writing it yourself would have taken.
If the edit was faster, that task belongs to Claude from now on.
Going deeper
This article is the headline frame. The longer version, eight workflow chapters that apply this model to specific parts of a consulting week, with the prompts and worked examples, is The Solo Consultant's AI Playbook. Chapter 1 is this article expanded; Chapters 2 through 8 are the workflows, each dropping you into a specific Category 2 or Category 3 task with the prompt that runs it.
If proposals are where this framework matters most in your week, the Proposal-Closer Prompt Pack goes deep on that one workflow.
These category calls are worth making once rather than every time you open a tab, and that's what the Claude Workspace Kit does: standing instructions and context files kept in a folder (your ICP, offers, rate floor and voice), plus written standards for what finished work looks like, so each task arrives with its constraints already attached.
Either way, what decides whether AI saves you time is less the prompt than whether you decided which category the task belongs to before you opened the tab.
— Digital Kreative
