Pillar Guide · AI Infrastructure
AI Workflows for
Solo Consultants.
AI tools don't save time by default — they save time when each one is assigned to a specific workflow job and connected to the rest of the system. This is the complete AI infrastructure for a solo consulting practice: every tool mapped to its job, every prompt explained, and the Workflow Compression metric for each layer.
The core argument
AI is a workflow layer, not a tool category.
The consultants who get the most leverage from AI aren't the ones using the most tools — they're the ones who've assigned each AI tool a specific job in a specific workflow and connected those tools so their outputs feed into each other. Claude's proposal draft feeds into PandaDoc. Fathom's transcript feeds into Claude's prep prompt. Perplexity's research feeds into the discovery call prep brief.
The alternative — using AI ad hoc, pasting things into ChatGPT occasionally, trying different tools without a system — produces inconsistent results and requires constant re-learning. The system approach produces compounding returns: the more you use it, the faster each workflow runs, because your prompts get better and your context gets richer.
This guide maps the AI infrastructure for each stage of the consulting workflow. It covers the right tool for each job, the specific prompts that work at consultant-level complexity, and the Context Reconstruction Cost reduction each system produces.
The stack
Four tools. Every AI job in the practice covered.
The solo consultant AI stack doesn't need to be large. Four tools — assigned to specific jobs, connected deliberately — cover every knowledge-work task in the practice. Adding more tools beyond these four typically increases Operational Surface Area without proportionate capability gain.
| Tool | Primary job | Secondary job | Cost |
|---|---|---|---|
| Claude Pro | Long-form writing: proposals, reports, strategy memos | Context reconstruction, deliverable drafting, email writing | $20/mo |
| Fathom | Discovery call recording, transcription, AI summary | All client call notes, action item extraction | $0 |
| Perplexity Pro | Prospect research, industry context, verifiable facts | Competitive landscape, market sizing, news monitoring | $20/mo |
| Notion AI (optional) | In-portal summarization, action item extraction | Meeting note formatting, status update drafting | $10/mo |
Total AI stack cost: $40/month (or $50/month with Notion AI). This is separate from but compatible with the automation stack (Make, HubSpot, Calendly). Together they form the complete consultant OS at $97/month total. See the complete OS guide →
Tool 1
Claude — The primary AI for client work.
Claude is the right AI for the high-complexity, long-form writing tasks that define consulting delivery. Proposals, strategy memos, client reports, engagement summaries, status updates — these are documents where voice, structure, and nuance matter, and where a mediocre first draft is worse than no draft at all. Claude's strength in these tasks comes from its ability to follow complex, structured prompts and maintain consistent quality across long outputs.
Where it fits: Claude handles five distinct jobs in the consulting workflow — (1) proposal drafting from discovery call notes, (2) context reconstruction before client calls, (3) deliverable drafting from structured briefs, (4) client communication rewrites ("make this email clearer and more direct"), and (5) strategic analysis frameworks when you need to think through a client problem before engaging.
Why Claude over ChatGPT for consulting work: Two specific advantages. First, Claude's extended context window (200K tokens on Pro) handles full engagement histories — you can paste an entire project's worth of meeting notes, the original proposal, and the current deliverable draft in one context, and Claude maintains coherence across all of it. ChatGPT's context is shorter and less reliable at the edges. Second, Claude follows structured prompts more reliably — when you specify eight required sections, it produces eight sections without improvising extras or skipping ones it finds difficult.
When to use ChatGPT instead: Tasks that benefit from web browsing and real-time information (use Perplexity for this instead), image generation (DALL-E integration), or code generation tasks where GPT-4's coding capability is preferred. For pure text work at consulting complexity, Claude is the better choice.
Five core Claude prompts for consultants.
Tool 2
Fathom — The call intelligence layer.
Fathom solves the Context Reconstruction Cost problem at the source. Every client call — discovery, kickoff, check-in, delivery review — is automatically transcribed and summarized. The AI summary identifies action items, decisions made, and key discussion points. You stop taking notes during calls and start listening fully. Post-call, you have a searchable record rather than reconstructed memory.
What it does: Fathom joins every Zoom call automatically once configured. It records, transcribes in real time, and generates an AI summary within 2–3 minutes of the call ending. The summary includes: a brief overview, key discussion points, decisions made, and action items with implied owners. The full transcript is searchable — you can find any moment from any call in seconds.
The workflow integration: After every client call, paste the Fathom AI summary into the relevant section of the client's Notion portal. This is the weekly 10-minute update habit — Fathom produces the content, you paste it in. The portal becomes a running record of every decision, every action item, and every discussion point without any manual writing. Context Reconstruction Cost drops to near-zero because the record is complete.
Discovery call specific use: After a discovery call, use Prompt 5 above (Claude debrief prompt) with the Fathom summary as input. Claude extracts the six proposal-ready fields in structured format. This replaces the 20–30 minutes most consultants spend writing up discovery notes from memory — and produces more accurate, complete input for the proposal because it's working from a verbatim transcript, not reconstructed notes.
Limitation: Fathom's free tier covers Zoom only. If you use Google Meet or Microsoft Teams regularly, Fathom charges $19–$29/month for those platforms. Otter.ai is the alternative — it covers more platforms and has its own AI summary capability, though the summary quality is slightly lower than Fathom's at the time of writing.
Tool 3
Perplexity — Research with citations.
The most important distinction between Perplexity and Claude for research tasks: Perplexity cites its sources. For consulting work — where you may need to present market data, competitive intelligence, or industry benchmarks to clients — cited research is meaningfully more valuable than AI-generated content you can't verify. Claude is excellent for writing and analysis; Perplexity is better for any research task where the answer's accuracy matters and you'd need to verify it anyway.
Pre-call prospect research: Before a discovery call, a 10-minute Perplexity research session produces: a company overview with recent news, the prospect's background from LinkedIn, the industry context for their problem, and 2–3 relevant case studies or data points you can reference naturally in the conversation. This is the input for Prompt 1 in Claude (context reconstruction) — Perplexity provides the external context; Fathom provides the call history; Claude synthesizes them into a pre-call brief.
The research prompt that works: "Tell me about [Company Name]. Include: what they do, their business model, recent news or changes, their typical challenges in [their industry], and any public information about [Prospect Name]. Cite your sources. I'm preparing for a discovery call and need accurate, current information."
Competitive landscape research: When a client asks "who else is doing this?" or "what are our competitors doing?", Perplexity produces cited answers faster than any manual research process. The follow-up prompt: "What do [Competitor names] do differently from [Client] in [specific area]? Cite your sources and note the date of the information." The date note matters — Perplexity's training data has a cutoff, and for fast-moving industries, you want to know whether the information is current.
Free vs Pro: Perplexity's free tier uses its own AI model and limits Pro searches. The $20/month Pro plan unlocks Claude and GPT-4 as search backends — meaning you get the research capabilities of Perplexity's source retrieval combined with the reasoning quality of frontier models. At consulting use volume (10–20 research sessions per week), Pro pays for itself quickly in recovered research time.
The system
Context management — the hidden ROI of AI in consulting.
The highest-leverage application of AI in solo consulting isn't writing — it's context management. A consultant with three active clients at different project stages carries significant cognitive overhead: remembering where each project stands, what was decided last week, what each client cares most about, and what needs to happen before the next call. This overhead is real work that's currently done by memory and email-searching — both unreliable and time-consuming.
The system that eliminates this: Notion client portals (structured, persistent context) + Fathom meeting notes (automatic call documentation) + Claude context reconstruction prompts (rapid re-immersion). Together, they reduce Context Reconstruction Cost from 30–45 minutes per client per week to under 5 minutes. At three active clients, that's 75–120 minutes per week recovered — from a 30-minute afternoon setup and a 10-minute weekly update habit.
Notion: persistent structured context
The client portal contains the full engagement history in structured form — decisions, deliverables, meeting notes, next steps. This is what you give Claude when you need context reconstruction. The structure makes it scannable in 2 minutes and parseable by AI in seconds.
Fathom: automatic call documentation
Every call becomes a searchable, summarized record automatically. No note-taking during calls. No "I think we decided X" — you know exactly what was decided, when, and by whom. Paste the summary into the Notion portal. The record builds itself.
Claude: rapid re-immersion
Before any client interaction, paste the last 3 Notion meeting notes + current scope into Claude with the context reconstruction prompt. Get a 4-minute briefing covering current status, priorities, open items, and one thing to watch. Arrive fully oriented at every call.
Prompt library
Additional prompts for the consulting workflow.
How it connects
The AI workflow from prospect to closed project.
What doesn't work
Four AI patterns that waste time instead of saving it.
Using AI without structured prompts
Ad hoc "write me a proposal for a strategy consulting engagement" produces generic output. Structured prompts with the six discovery fields, the client's exact language, and the specific outcome produce proposal-quality drafts. The prompt is the skill — the AI is the execution. Invest time in prompt development, not tool switching.
Treating AI output as final
Every AI output requires a pass for accuracy and voice. Claude will infer specifics you didn't provide, use your billing rate incorrectly, or write in a slightly more formal register than your natural voice. The 15–20 minute edit pass isn't overhead — it's where you add the judgment that the AI can't have.
Not giving AI enough context
The quality of AI output scales directly with the quality of context you provide. A proposal prompt with shallow discovery notes produces a shallow proposal. A prompt with verbatim client language, quantified problem costs, and specific outcome metrics produces a proposal that reads like you understood the client perfectly. The context is the work.
Using AI for everything
Not every task benefits from AI involvement. Strategic recommendations that require your judgment, nuanced client relationship decisions, and pricing negotiations are human tasks. Using AI on these introduces latency without value. The system works best when AI handles the structured, repeatable components and humans handle the judgment-intensive ones.
Get the AI prompt library
All five core Claude prompts plus 8 additional prompts for every recurring consultant workflow — formatted, tested, and ready to use. Free with your email.
- Context reconstruction prompt
- Discovery debrief + proposal setup prompt
- Status report from meeting notes
- Scope creep response template
- 8 additional workflow prompts
Free for subscribers
No spam. Unsubscribe any time.
Related resources