Your sales team writes proposals from scratch every time, and they look nothing alike.
Proposals are the last mile of a sales cycle—and they're still handwritten. Your team spends 2–4 hours per deal rebuilding the same sections: executive summary, scope, pricing table, timeline, terms. Each rep has their own style. Clients see it. You lose deals to competitors who look more organized. Here's the workflow: after a discovery call, your rep (or a sales ops person) pastes notes into a prompt. An LLM generates a first draft—sections, tone, structure all consistent. The rep then spends 30 minutes editing for client specifics, not writing from zero. It goes to a manager for approval before send. Failure modes are real: the AI might hallucinate pricing, miss custom terms, or sound too generic. That's why the human gate matters. The math: if you close 20 deals a month and each proposal saves 90 minutes of writing, you recover 30 hours. At $100/hour loaded cost, that's $3k/month in time. The system costs $200–500/month in LLM tokens and orchestration. Payback is immediate. The hard part isn't the AI—it's the approval gate. You need a sales manager or ops person to catch hallucinations before they reach the client. That person becomes your quality filter. Train them on what to watch: pricing consistency, scope creep, missing custom terms, tone mismatches. Without that gate, you ship broken proposals and lose trust. Implementation is straightforward: connect your discovery notes (Slack message, email, or CRM field) to a prompt that structures them into a proposal template. Use Claude or GPT-4 for the generation step. Route the draft to a human reviewer in Slack or your CRM. Once approved, auto-populate your proposal tool (PandaDoc, Proposify, or custom PDF) and send. The biggest win: consistency. Every proposal has the same sections in the same order. Clients perceive professionalism. Your team stops reinventing the wheel. And you get data: which sections take the longest to edit, which clients ask for the same changes, where your discovery notes are too thin.
Trigger: Sales rep completes discovery call and logs notes in CRM or Slack.
1. Create a prompt template that takes discovery notes as input and outputs a structured proposal (executive summary, problem statement, solution, scope, timeline, pricing, terms, next steps). 2. Connect your CRM or Slack to an LLM API (OpenAI, Anthropic, or a no-code tool like Zapier + ChatGPT). 3. Set up a human approval gate: route the draft to a sales manager or ops person via Slack thread or CRM task. 4. Once approved, auto-populate your proposal tool (PandaDoc, Proposify, or Google Docs template) and send. 5. Track edit time and approval time to measure impact. 6. Iterate on the prompt based on what the human reviewer changes most often.