Guide for consultants · AI
Write a proposal from meeting notes: a reviewed draft in 30 minutes
After a good first meeting, the proposal should reach the client quickly. In practice it often sits for days because the notes are incomplete, the useful wording is buried in an old proposal, and the draft has to be written alongside day-to-day client work. AI can often speed up the first draft considerably. That only works if your notes contain the right information and you know what to check yourself afterward. This guide shows a routine that produces a reviewed draft in about 30 minutes: ten minutes to complete your notes, ten minutes for the draft, ten minutes for the review. You set up the building blocks from Step 2 once and reuse them. You add price and terms yourself at the end.
What a proposal draft needs to contain
A proposal for consulting or professional services answers seven questions. If one of them is missing from your notes, the AI cannot answer it either. It will fill the gap with generic wording or make something up.
1. Situation: What is going on at the client, in their own words?
2. Goal: How will the client know after the project that it was worth it?
3. Scope: What exactly will you do, and what will you explicitly not do?
4. Approach: In which steps or phases will you work, and what does each step deliver?
5. Client input: What does the client need to provide, who is the contact, what access do you need?
6. Assumptions and open points: What have you assumed, and what is still unclear?
7. Timeline and pricing model: By when, and how will you bill? You set the amount yourself.
Step 1: Complete your notes right after the meeting
The most important step does not happen in the AI. It happens in the ten minutes after the meeting. Go through the seven questions and add bullet points to your notes. Capture in particular:
• The client's own words about the problem. A proposal that picks up the client's language reads as more tailored than any template wording.
• Commitments you made in the meeting, such as “We'll send you a proposal with two options by Friday.”
• What the client does not want. Exclusions matter as much as deliverables.
• Names, roles and dates that should appear in the proposal.
If you recorded and transcribed the meeting with everyone's consent, the transcript does not replace your notes. It is usually too long and full of side topics. First have the AI extract the answers to the seven questions, check them, and only then ask for the draft.
Step 2: Prepare your building blocks (once)
AI writes better proposals when it can see what your proposals usually look like. Pick one or two earlier proposals for similar work, plus your standard text on approach, client input or billing. Remove client names, amounts and confidential details from the old proposals, or replace them with placeholders. That way the AI picks up structure and tone, but none of another client's content.
Which AI is suitable depends on what is in your notes. For notes that include names and details about a client, the work usually belongs in a business or enterprise plan with a data processing agreement, not in a personal account. Also check whether your agreements with the client allow it. More on this below.
Step 3: Have the AI write the draft
Give the AI a clear task, your completed notes and the building blocks. An example instruction:
Write a proposal draft for a consulting engagement. Use only the information in the meeting notes below. Take structure and tone from the two sample proposals, but none of their content. Use these sections: situation, goal, scope with exclusions, approach in steps, client input, assumptions and open points, timeline. Use a professional, direct tone and pick up the client's own words where they fit. Do not invent numbers, dates, deliverables or commitments. Where information is missing, write [OPEN: what is missing]. Leave out the price. At the end, list every point you are unsure about.
The instruction to mark gaps with [OPEN] is the most important part. It makes visible what the AI did not know instead of covering it with smooth sentences.
Step 4: Review the draft in ten minutes
Do not read the draft for style first. Read it for content. Always check these points yourself:
• Every [OPEN] marker is resolved or deliberately turned into a question for the client.
• Commitments and dates match what you said in the meeting.
• The scope has not grown. AI tends to add extra services that sound sensible.
• The exclusions are in the text. Anything the proposal does not explicitly exclude, the client may later assume is included.
• Names, roles and company names are spelled correctly.
• Nothing comes from another client's project, including industry examples from the building blocks.
Then add price and terms and read the whole proposal once, the way the client will read it.
What the AI should not decide
A proposal usually commits your company. Three things therefore stay out of the AI's hands: the price, legal terms (such as liability, payment terms and your standard terms and conditions) and the decision whether you want to take on the work in this form at all. Whether you send a binding proposal or a non-binding first outline is also your call. The AI delivers a draft. Signing and sending the proposal is up to you.
Which AI is suitable for client notes
In personal ChatGPT accounts, such as Free, Plus or Pro, sharing conversations for model training is on by default. You can turn it off under “Settings”, “Data controls” and “Improve the model for everyone”. There is no data processing agreement for these accounts, though. For ChatGPT Business and ChatGPT Enterprise, OpenAI says it does not train on business data by default and offers a data processing addendum. Microsoft Copilot Chat with a work account is covered by Microsoft's enterprise data protection terms. Web searches that Copilot sends to Bing are the exception. Which option works for your company is best clarified with your data protection officer or privacy advisor. More in the guide Can our team put client data into ChatGPT?
Where this stops working
For one proposal a week, the routine above is enough. It gets harder when several people write proposals, the good building blocks sit in different folders and nobody remembers which proposal led to a contract with a similar client. Drafts are then written from memory again, and everyone works from a different template. The notes also come from different places: the calendar, email, a transcript or the CRM.
Where aenuma fits
aenuma sets up this routine in the tools your team already uses. That includes connecting the information it needs, such as meeting notes, earlier proposals and standard text, and testing the routine with representative examples from your day-to-day work. We agree in advance which steps a person approves. Sensitive information can be anonymized for the initial discussion.
If you first want to know whether this is worth it for your proposals, we start by identifying where the routine loses the most time and which requirements are missing.
Frequently asked questions
How long does a proposal with AI realistically take?
If your notes answer the seven questions and one or two suitable sample proposals are ready, a reviewed draft in about 30 minutes is realistic. If information is missing, the work only moves: the AI marks the gaps, and you still have to fill them. Large proposals with many work packages take longer, because pricing and alignment within the team remain your job.
Can I just paste my meeting notes into ChatGPT?
Not into a personal account if the notes include names or confidential details about a client. That usually requires a business or enterprise plan with a data processing agreement, a rule in your company about which data may go in, and confirmation that your agreements with the client allow it. See the section “Which AI is suitable for client notes” above.
How do I spot that the AI made something up?
Compare every number, date and deliverable in the draft with your notes. Anything not in your notes was invented or came from the sample proposals. That is why the instruction to mark missing information with [OPEN] and to list all uncertain points at the end helps.
Sources
Data Controls FAQ (OpenAI Help Center)
Enterprise privacy at OpenAI (OpenAI, updated January 8, 2026)
Angebotsanfrage: Beim Call for Proposal Zeit sparen (consulting-life.de, in German)
Enterprise data protection in Microsoft Copilot and Microsoft Copilot Chat (Microsoft Learn)
By Joshua Schütt, founder of aenuma. Last updated: October 2026.
Which proposals do you still write entirely by hand?
Tell us how a proposal comes together at your company today. We will tell you which part can sensibly be prepared and what you are better off doing yourself. Not sure where to start? The Time Sink Check shows which task makes a good first workflow.
Book a call