# Follow up every demo without the spreadsheet: the AI build for sales teams Source: https://getstartedwithai.net/tutorials/ai-demo-followup-build Site: Get Started With AI Tagline: Build AI automations that actually run your business. Audience: solo professionals, freelancers, consultants and AI implementers in training Contact: hello@getstartedwithai.co.uk This Markdown companion is generated from the prerendered page for AI crawlers. # Follow up every demo without the spreadsheet: the AI build for sales teams Sales reps lose deals in the gap between the demo and the follow-up. The follow-up email lands hours later (or the next morning), generic, missing what was actually said on the call. The build below removes that gap. Fathom or tl;dv records the call, Claude reads the transcript plus the prospect file, and a personalised follow-up email lands in the rep inbox 10 minutes after the call ends. The rep reviews, edits if needed, and sends. Demo-to-meeting-two conversion lifts 15 to 25 percent because the prospect gets a relevant follow-up while the call is still in their head. ## The tooling stack - Fathom Free or tl;dv Pro from about £15 a month annual: records the call and produces the transcript. - Claude Pro at £16 a month: reads the transcript and drafts the follow-up. - Zapier or Make: pipes the transcript from the recording tool into Claude and the draft into the rep inbox. - HubSpot, Pipedrive or Salesforce as the source of the prospect file. ## Step 1: Build the Demo Follow-up Project (45 minutes, one-off) 1. Open Claude Pro, Projects, New Project. Name it Demo Follow-up. 2. Custom Instructions: paste the role, the rep voice rules, the typical follow-up shape (recap of one specific moment, recap of decision criteria, recap of next step), the rules on what not to invent. 3. Project Knowledge: upload 10 best previous follow-up emails that closed, the rate sheet, the case study one-pagers. 4. Test with a known transcript. Confirm the output reads as the rep work. ## Step 2: Wire the recording tool (15 minutes) 1. In Fathom or tl;dv, configure the bot to join all sales calls. 2. Set up a Zapier trigger: when a call ends, send the transcript to Claude. 3. Test with one real call. 4. Confirm the transcript lands cleanly in the Claude project. ## Step 3: Wire the prospect file (15 minutes) 1. In Zapier, add a step that pulls the relevant prospect file from HubSpot or Pipedrive. 2. Pull these fields: company name and role of the contact, plus deal stage, last touchpoint and the last note on the record. 3. Pass the prospect file alongside the transcript into Claude. 4. Test with one real prospect. ## Step 4: Configure the draft prompt (15 minutes) The prompt is what makes the difference between a generic follow-up and one the prospect notices. 1. Prompt template: "Draft a follow-up email for [prospect]. Reference one specific moment from the transcript. Reference the decision criteria they named. Suggest a clear next step. Keep it under 150 words." 2. Test against 5 historic transcripts. 3. Iterate the prompt until the drafts read as the rep would write. 4. Lock the prompt in the Project Custom Instructions. ## Step 5: Pipe the draft into the rep inbox (10 minutes) 1. Configure the final Zapier step: send the Claude output as a draft email to the rep. 2. The draft sits in Drafts, not Sent. The rep approves before sending. 3. Add the prospect email as the recipient. 4. Add a one-line subject suggestion. 5. Schedule the rep review for within 30 minutes of the call ending. ## Step 6: Track the lift - Track demo-to-meeting-two conversion before the build for 4 weeks. - Track the same metric after the build for 4 weeks. - Aim for a 15 to 25 percent lift. - If the lift is below 10 percent, revisit the prompt and the project knowledge. - If above 25 percent, do not get smug. Repeat for 12 weeks before banking the result. ## What you ship A working demo follow-up pipeline that drafts a personalised email within 10 minutes of every call ending, references one specific moment from the call and one named decision criterion, and lands in the rep inbox for review. Demo-to-meeting-two conversion lifts 15 to 25 percent. Rep admin time drops about 30 minutes per opportunity. ## What to build next Once demo follow-up runs reliably, the matching build is the AI sales agent: a layer that handles inbound qualification and books the demo in the first place. Pair this with the AI sales agent build and the AI cold email outbound build. ## Related builds More tutorials in Marketing & Sales. Each one runs in 60 to 90 minutes and ships a finished asset. ### Send 50 cold emails a week that book meetings: the AI build for outbound 30 to 60 minutes. Solo Pro / Implementer. [Open tutorial](/tutorials/ai-cold-email-outbound-build) ### Publish a weekly newsletter in 90 minutes: the AI build for solo creators 60 to 90 minutes. Solo Pro. [Open tutorial](/tutorials/ai-newsletter-publish-weekly-build) ### Run a LinkedIn presence without burning Sundays: the AI build for solo professionals 60 to 90 minutes. Solo Pro. [Open tutorial](/tutorials/ai-linkedin-presence-build) ### Synthesise customer interviews into product moves: the AI build for founders 90 minutes. Implementer. [Open tutorial](/tutorials/ai-customer-interview-synth-build) ## Questions people ask before they start ### Will the AI invent details from the call? Not if the prompt is right. The prompt locks the draft to the transcript and the prospect file. Always review before sending; AI sometimes paraphrases in a way that loses precision. Most reps catch one edit per email in the first month, fewer after that. ### How fast does the email need to land? Within 30 minutes of the call ending is the sweet spot. The build targets 10 minutes. Above 60 minutes, the lift on conversion drops because the prospect has moved on. ### Does the rep still need to write follow-ups? No, the rep edits and sends. About 80 percent of drafts go out as written or with one small change. The other 20 percent need a real edit because the call had unusual dynamics. ### Will prospects notice the email is AI-drafted? Only if the rep does not edit. With voice calibration and the named-moment recap, the email reads as a thoughtful human follow-up. Most prospects reply faster because the email is more relevant than the typical follow-up. ### Is this safe with prospect data? On the Pro tier of Claude or ChatGPT, yes. Inputs from these tiers are not used for training. For regulated B2B sales (financial services, healthcare), use enterprise tier with a processor agreement.