The problem, as you feel it
The discovery call goes well. The prospect is interested. You say you will send a proposal by Friday. Then Friday is a week of real work, the proposal gets written Sunday night from the last proposal with the names changed, and it goes out Monday with a chart that still says the old client’s numbers on slide seven.
Meanwhile a competitor sent theirs the same afternoon as the call. It was specific. It quoted the prospect’s own words back to them. It won.
Every business that sells with a document (agencies, contractors, consultants, event companies, architects, IT providers, anyone who quotes custom work) has this problem, and the standard fix, a better template, does not fix it. The template is not the slow part. The slow part is turning what you learned on the call into a document that reflects it.
What an AI proposal system does
The pattern is the same every time: capture what was said, structure it, write into your template, have a person check it, send it while the conversation is still warm.
- The call becomes notes automatically. Record the call (with consent), and the transcript is summarised into the structure a proposal needs: their situation, what they said the problem costs them, what they want, their timeline, their budget signals, who decides, the objections they raised.
- The proposal is written into your template, section by section, using their language for their problem and your language for your solution. Pricing is assembled from your rate card and the scope discussed, never invented.
- The deck version is built from the same content, in your brand, with the charts and visuals your standard pitch uses, populated with their specifics.
- A person reviews it. You read a document that is 90 percent done and fix the 10 percent, rather than staring at a blank page.
- It goes out the same day, as a link that tells you when they open it and which sections they linger on, with the PDF attached for the traditionalists.
Take a business like this one
The business in this section is a composite of the kind of company I talk to, not a named client. The numbers are the shape of the problem, not a case study.
Picture a nine-person marketing agency. Two partners sell, the rest deliver. They run around six discovery calls a month and send four proposals, because two calls a month never get a proposal at all: the partner ran out of time and the prospect went quiet. Their proposals take four to six hours each, go out five to nine days after the call, and win about one in four.
What I set up for an agency shaped like this:
- Call capture. Calls are recorded with consent and transcribed. Within ten minutes of hanging up, the partner has a structured brief in the shared drive.
- A proposal template that is actually a system. The agency’s existing proposal, rebuilt so each section knows what it needs from the brief: the “your situation” section pulls the prospect’s own words, the scope section maps what they asked for to the agency’s service menu, pricing comes from the rate card with the partner’s discretion range built in.
- The AI writing pass. The brief plus the template produce a complete first draft in the agency’s voice, with any section it could not fill honestly left marked for the partner rather than padded.
- The deck, automatically, from the same content, in the agency’s brand, with the case studies chosen to match the prospect’s industry.
- Review and send. The partner spends thirty to forty-five minutes making it theirs, then sends a tracked link the same afternoon.
In an agency shaped like this you should expect every discovery call to get a proposal, proposals to go out the same day or the next morning instead of the next week, and the four to six hours to become under an hour. Win rates on proposals that arrive within a day of the call and quote the prospect’s own words are consistently better than the ones that arrive a week later; how much better depends on the agency, but two extra proposals a month sent on time is a different business.
What it costs to run
Transcription and AI writing cost a dollar or two per proposal. Document tracking is a small monthly tool. The setup is the work: rebuilding your proposal and deck into templates that know their own structure, capturing your voice, encoding your rate card and scoping rules, and tuning the output on your last ten real proposals until a partner cannot tell which ones were drafted by the system.
What I need from you
Your last ten proposals, including the ones that lost. Your rate card and how you actually scope, including the unwritten rules. Your deck. Two hours across a couple of sessions to review generated drafts and tell me where they sound wrong. Then the habit of hitting record at the start of discovery calls.