AI Layer · Sales · Brief 50
AI Proposal OS for Solo Consultants:
Using Claude and ChatGPT to Write Proposals That Close.
AI writes the proposal. You make the decisions that determine whether it wins. A well-configured AI workflow reduces first-draft time from 3–4 hours to 45–60 minutes — but only if the judgment work is done first. Five-stage system with exact prompts, the context document framework, and honest guidance on where AI fails. Updated May 2026.
Updated: May 2026 · Pricing verifiedThe judgment / drafting distinction
AI writes the proposal. You make the decisions that determine whether it wins.
There are two types of work in every proposal: judgment work — what to charge, what to promise, what to exclude, what risk to absorb — and drafting work — turning those decisions into buyer-ready prose. AI is a drafting tool. Every proposal failure that comes from AI — the generic language, the wrong framing, the price that feels off — traces back to conflating these two.
A well-configured AI proposal workflow reduces first-draft time from 3–4 hours to 45–60 minutes. Here is exactly where that time comes from: discovery synthesis (45–60 min → 10 min), proposal structure (30–45 min → 0 min), investment narrative (45–60 min → 15 min), objection sections (30–45 min → 10 min), polish (30 min → 15 min). Total: ~3.5 hours → ~50 minutes. But none of this works if the judgment layer was skipped.
The voice problem
Why AI proposals feel generic — and how to fix it.
AI proposals feel generic because the model has no access to your positioning, your client relationship context, your past wins, or your way of thinking about problems. The solution is a context document — a persistent, reusable file that front-loads the AI with everything it needs to write in your voice, not in a generic consulting voice.
What goes in the context document
- Your positioning statement and ICP description
- Your standard engagement structures and typical scope boundaries
- Your pricing philosophy (framing approach, not specific rates)
- Voice and tone guide with 2–3 past proposal excerpts you are proud of
- A "never use these phrases" list (buzzwords, hedges, generic consulting-speak)
- 2–3 brief past-win stories in your own words
Load this document into Claude Projects (as a Project document) or a ChatGPT Custom GPT (as knowledge base content). Every proposal session runs inside the project or custom GPT — the context is always present without re-pasting. See the AI Writing OS for the full AI stack configuration.
The five-stage workflow
Stage by stage, prompt by prompt.
Stage 1 — Discovery → Proposal Brief
Take your discovery call notes or transcript (from Fathom or Fireflies) and use AI to synthesize them into a structured brief. This is the input for every stage that follows. A weak brief produces weak output at every downstream stage.
Stage 1 prompt
You are synthesizing a discovery call transcript for a solo consultant preparing a client proposal.
From the transcript below, extract:
1. STATED PROBLEM: What the client said they need help with (their words, not yours).
2. UNDERLYING PAIN: The business consequence if this goes unsolved.
3. DESIRED OUTCOME: What "done" looks like in the client's words.
4. SUCCESS CRITERIA: How the client will measure whether this worked.
5. UNSTATED CONCERNS: Hesitation, skepticism, or risk signals expressed or implied.
6. DECISION CONTEXT: Who else is involved and what is the timeline pressure.
Format as a structured brief I can paste into a proposal drafting session.
TRANSCRIPT: [paste transcript or notes]
Human judgment re-entry: Review the brief against your own read of the call. AI will miss nuance — a hesitation, an unstated budget constraint, a political dynamic. Add these by hand before Stage 2. Time saved: ~45 minutes → ~10 minutes.
Stage 2 — Proposal Structure Generation
Use the completed Stage 1 brief to generate a full proposal outline. The best proposals are arguments, not catalogs of deliverables. Claude is strong at generating argument structure when given clear inputs. ChatGPT's structured output is better for deliverable-heavy SOW-style proposals.
Stage 2 prompt (add to system context)
Using the proposal brief below and your system context, generate a full proposal outline.
For each section include:
- The section header
- The argument this section is making (1 sentence)
- The key claims or evidence needed (bullet list)
- Any client-specific language from the brief to weave in
Brief: [paste Stage 1 output]
Human judgment re-entry: Review the outline and delete or restructure sections that don't fit this client. Editing a bad outline takes 5–10 minutes; generating prose from a bad outline wastes an hour. Time saved: ~35 minutes → 0 minutes (plus 5–10 min review).
Stage 3 — Investment Summary Drafting
The most judgment-sensitive section. Decide the fee, payment structure, what's included and excluded, and the risk reversal before touching the AI. The AI then takes those decisions as inputs and produces the narrative — the ROI anchor, the investment framing, the risk reversal language.
Stage 3 prompt
Write the investment summary section for this proposal.
Engagement: [brief description]
Fee: [amount and structure]
Included: [list]
Excluded: [list]
Risk reversal: [e.g., "30-day checkpoint exit clause"]
ROI anchor: [e.g., "current process costs approximately $3,200/month in lost productivity"]
Write in [Name]'s voice (see system context). Lead with the business case, present the investment, close with the risk reversal. Do not use phrases like "investment in your future" or "partnership." Keep it direct and confident.
Where AI produces generic language: Target specifically: (1) any hedging language — "this may help," "we believe" — and replace with direct statements; (2) any generic ROI language and replace with the client's specific numbers; (3) any phrase that could have been written for any client. Time saved: ~50 minutes → ~15 minutes.
Stage 4 — Objection-Handling Sections
The best proposals handle objections before the follow-up call. A client who reads a proposal that already addresses their concern does not need to raise it. Use AI to surface the strongest objections — not the obvious ones, the actual ones — and write pre-emptive language for each.
Stage 4 prompt — steelman the objection
Based on the proposal brief below, identify the 3–5 strongest objections this specific client is likely to have.
For each objection:
1. State it as the client would think it — the real concern, not a softened version.
2. Identify whether it is a price, timing, risk, fit, or priority objection.
3. Write a pre-emptive paragraph I can weave into the relevant proposal section — without appearing defensive.
Approach each objection by making the strongest possible case for why the client would be right to have it, then resolve it on the merits.
Brief: [paste Stage 1 output]
Important: Do not insert objection-handling as a standalone "Objections" section — that signals defensiveness. Weave pre-emptive paragraphs into the sections where the objection would naturally arise. Time saved: ~35 minutes → ~10 minutes.
Stage 5 — Version Control and Iteration
Build a system that improves with each engagement. After every proposal — won or lost — spend 10 minutes updating the context document: add client-specific language that landed well, add any objection that came up that wasn't anticipated, update the "never use" list, note what the client cited as their reason for moving forward.
A consultant who runs 20 proposals through this system has a context document calibrated to their specific market, buyer language, and objection landscape. The AI is increasingly useful as the context document becomes richer — not because the model improves, but because your inputs do.
Claude vs ChatGPT
Which tool fits your proposal type.
For the full AI stack comparison, see the AI Writing OS. For proposals specifically:
Strategy / advisory proposals → Claude
Selling thinking, not deliverables. Claude generates argument structure and nuanced reasoning more naturally. Prose for conceptual work reads more like a peer thinking through a problem.
Implementation / deliverable proposals → ChatGPT
Selling output and timelines. ChatGPT's structured formatting and list output is better suited to SOW-style proposals with phases, milestones, and explicit deliverable lists.
Fractional / retainer proposals → Claude for narrative, ChatGPT for terms
The narrative case for the fractional model requires Claude's reasoning. The rate, hours, and scope terms are better formatted by ChatGPT's structured output.
What AI will not fix
The honest close.
If win rate is low, the problem is usually upstream: a poor discovery call, a misaligned prospect, or a pricing structure that doesn't match market expectations. AI-assisted drafting will not fix a discovery problem — it will accelerate the production of a well-written proposal that still doesn't close.
This workflow assumes a solid discovery output. If the discovery is weak, fix that first — see the Discovery Call OS. For the tools that format and send the final proposal, see the Proposal OS. For pricing framework before you write a word, see the Pricing OS.
Diagnostic — what's your actual pain?
Takes too long → Your highest-leverage investment is Stages 1 and 2 (discovery synthesis + structure). This is where the hours are.
Win rate is low → Invest in Stage 3 (investment narrative) and Stage 4 (objection handling). The problem is likely in how you're framing value, not how long it takes to write.
Proposals feel generic → Your problem is the context document. You haven't loaded enough of your own voice, positioning, and past-win language into the system.
Get the Solo Consultant OS Blueprint
Five-layer OS architecture, tool selection by practice stage, and automation wiring — free for subscribers.
- Five-layer OS framework
- Tool selection by practice stage
- Make automation scenarios
- Weekly OS Review template
Free for subscribers
No spam. Unsubscribe any time.
Related reading