The actual Claude workflows of a solo consultant, six months in
Share
Twitter is full of takes on what Claude can do for solo consultants. This article is about what Claude actually does, six months in, for one practice. Some of it matches the Twitter version. Most of it doesn't. The interesting part is the gap.
The frame: there are workflows you'll read about, try once, and abandon — they sound great in a 10-tweet thread but don't survive contact with real client work. There are workflows you'll discover yourself, accidentally, while trying to solve something specific. And there are workflows you'll run so many times the muscle memory makes you forget you're using AI at all. The third category is the only one that matters; everything else is performative.
Here's the breakdown.
This article assumes you've decided AI is part of your practice and you're past the wondering-what-it's-good-for stage. If you're earlier than that, start here. The frame for which tasks should go to AI at all is the prerequisite to picking which workflows to run.
Workflows that compound
These are the ones that became default. Multiple times a week, often daily. The compound is real because each run sharpens the prompt, and the time-savings stack.
1. The discovery dossier
Before any discovery call, three artifacts: company brief, decision-maker brief, fit flag. Same Claude tab, three prompts top to bottom. Three minutes total. Covered in detail in Article 3.
Why it compounds: discovery calls happen weekly. The prompts are the same every time. The time savings show up immediately and compound across every call you take for the rest of your career.
2. The meeting debrief
After every meaningful client meeting: paste the transcript, get back decisions made, action items, surprises, and a read on the room. Five minutes. Covered in Article 5.
Why it compounds: you have client meetings every week, sometimes daily. The debrief used to take 20 minutes by hand and was the kind of thing I'd skip when busy — which meant I'd lose the context by the time I got back to the engagement.
3. The deliverable section expander
Outline + bullets + AI prose pass + edit. The bullets are mine; the prose is Claude's; the edit is mine. Covered in Article 2.
Why it compounds: every consulting deliverable has sections. The compound effect is per-section, and the savings are large per section (90 minutes → 15 minutes). Across a memo with eight sections, the math becomes hours.
4. The proposal flow
Eight chained prompts that walk a deal from discovery debrief to signed proposal. Discovery debrief → ICP fit check → scope generator → pricing anchor → proposal draft → objection handler → follow-up sequence → deal debrief. Each prompt's output feeds the next.
This is enough work to be its own product — the Proposal-Closer Prompt Pack. Why it compounds: proposals are the highest-frequency Category 2 task, and the chain stops you from forgetting any step. Without it, the deal-debrief prompt — the one that makes the next proposal better — is the one that always gets skipped.
5. The case-study harvest
Within two weeks of an engagement wrapping: project debrief → public-facing case study → three social/newsletter pieces from the same source. The compound is the multiplier — one wrapped engagement produces four pieces of marketing content for ~30 minutes of work.
Why it compounds: every wrapped engagement is a marketing asset that decays at about 30% per month if you don't capture it. The harvest workflow turns the decay into compound interest.
Workflows I tried and dropped
These are the ones that sounded great and didn't survive real use. Worth naming them so you don't waste a week trying.
1. AI-as-strategy-partner
The pitch: paste your business situation into Claude and get strategic advice. The reality: Claude's advice is generic-good. It sounds right. It misses the texture that makes any strategic decision actually strategic — the political dynamic with the client, the constraint that's not in the brief, the thing your gut is telling you.
This is squarely Category 1 work. AI is bad at it and will sound confident anyway. Dropped after about three weeks of trying.
2. AI-drafted client emails
The pitch: Claude writes your client correspondence in your voice. The reality: client emails are short. The 70-90 second time savings of having Claude draft them are eaten by the editing pass to remove the AI flavor. For long emails — 5+ paragraphs — the savings are real. For typical client emails, it's faster to just write the email.
The exception: kickoff emails after a deal closes. Those have structure and are repeatable. Worth it. Day-to-day client correspondence: not worth it.
3. AI-generated thought-leadership content
The pitch: turn one engagement into ten LinkedIn posts. The reality: what came back was readable, generic, and worse than what I'd write tired on a Friday. The output had the unmistakable AI texture — three-part lists, summary paragraphs, hedged claims. It would have damaged the brand to publish.
The fix isn't more prompting. The fix is the workflow in Article 7 — bullets first, prose pass after, edit ruthlessly. With that workflow, the LinkedIn posts work. Without it, they don't.
(Yes, that's a recursive reference to this article. The workflow was good enough to write the article that argues for the workflow.)
4. Claude as a coding pair
Different category. Some consultants do code work; many don't. For the engagements I take, Claude as a coding pair is irrelevant — but Claude Code, the CLI tool that runs the Digital Kreative storefront automation, is genuinely useful in a different way (it's the thing that drafts the publishing scripts and the GraphQL queries that hit the Shopify Admin API). That's a different workflow than "AI as a coding partner inside a consulting engagement." Mentioning it for completeness, not as a workflow recommendation.
The setup that makes the rest work
Two operational details that aren't workflows on their own but enable everything else:
Use Claude Projects, one per client
Each active client gets its own Claude Project with custom instructions: their industry context, the engagement scope, voice notes I picked up from their writing, and any constraint that doesn't show up in the brief. Every conversation about that client lives in their Project. The model has continuous context across the engagement, which means the prompts get sharper week to week.
This sounds simple and is the single biggest setup decision. Without it, every conversation starts from scratch. With it, by week three of an engagement Claude knows the client well enough to flag scope creep before I notice it.
Save your prompts as files, not as bookmarks
Every prompt in the workflows above lives in a prompts/ folder on disk, organized by category (prompts/proposals/, prompts/meetings/, etc.). The folder is in version control. The advantage: each prompt is editable, history-tracked, and shareable across machines.
The alternative — bookmarking prompts in Claude's UI or pasting from a Notion database — is fine but slower. The folder approach makes prompts feel like code rather than content, which is the right mental model for them.
What this looks like in a week
A real Tuesday tagged by category, six months in:
| Time | Task | AI involvement |
|---|---|---|
| 8:30 | Read prospect's site + LinkedIn | Discovery dossier prompt — 3 min |
| 9:00 | Discovery call | None (Category 1) |
| 9:35 | Debrief notes from the call | Debrief prompt — 5 min |
| 10:00 | Strategic decision: do I propose? | None (Category 1) |
| 10:15 | Draft proposal | Section-expander on each section — 30 min total |
| 11:00 | Active-client meeting | None during; debrief prompt after — 5 min |
| 12:30 | Lunch | None |
| 13:30 | Draft strategy memo for active client | Outline + bullets myself; section-expander on each |
| 15:00 | Follow-up email to morning's prospect | Follow-up prompt — 2 min |
| 15:15 | Newsletter post for the week | Section-expander from bullets |
| 16:30 | Quarterly retrospective prep (it's Friday in two days) | Pull pipeline data — 30 min |
Time spent on AI tabs: roughly 45 minutes across the day. Time saved compared to the no-AI version: roughly 3-4 hours. This isn't theoretical math; the workflows above produce that savings concretely on this kind of day.
The honest caveat
Six months is a real sample. It's also one practice. Your mileage will vary on a few axes: what kind of consulting you do (delivery-heavy vs. strategy-heavy), how much of your work is Category 2 vs. 1, how comfortable you are with editing AI prose. The workflows above will work for most solo consultants doing project-based engagements; they'll work less well for retainer-based advisory work where most of the value is the conversation, not the deliverable.
The frame matters more than the specific workflows. Pick one, run it for two weeks, decide whether it compounds for your practice. If yes, keep it. If no, drop it.
Going deeper
The five workflows that compounded for me are covered in detail in The Solo Consultant's AI Playbook — the chapter structure maps roughly onto the workflow categories above. The Playbook is the operating model with the prompts; this article is the personal-experience version.
The proposal-flow workflow has its own deeper version in the Proposal-Closer Prompt Pack — eight prompts, the Claude skill, and the Acme Widgets worked example.
— Digital Kreative