Give AI a bounded task, verified context, and a review checklist. Use the time you save to understand the buyer, not simply send more messages.
Start with one repeatable job
Choose a task you already do often: preparing for a discovery call, organizing meeting notes, or drafting a recap. Write down what a good result looks like before asking AI to help. For a recap, that might be the buyer’s stated goal, one unresolved question, and the next step you actually agreed to.
Run a small trial with anonymized examples. Record the time spent preparing, editing, and verifying the answer. If the model produces a fast draft that takes longer to correct, narrow the task or improve the source material. Your useful measure is time to an accurate, usable result.
Before the call: facts, hypotheses, questions
Supply the company’s public product page, relevant announcements, and your own verified notes. Ask for a brief divided into confirmed facts, plausible hypotheses, and questions to investigate. A job title is not proof of budget authority; a funding announcement is not proof of an active purchase.
Turn the brief into a short call plan. Ask about the current process, the cost of the problem in the buyer’s terms, the desired outcome, people affected, and how decisions happen. Rehearse likely concerns one at a time. Keep questions conversational rather than reading an AI-generated interrogation.
After the call: turn notes into commitments
Use approved notes or a transcript you are permitted to process. Ask AI to extract decisions, unresolved questions, owners, dates, and exact evidence. Leave a field blank when the conversation did not establish it. Review the output against the source before updating your CRM.
A meeting assistant can help collect summaries and action items. Fathom documents CRM integrations, including HubSpot and Salesforce. Confirm your plan, recording process, and field mappings before depending on automatic updates. The account owner still needs to verify what was promised.
Draft the next useful message
Ask for a short recap that reflects the buyer’s language, answers one question, and makes the next step easy. Remove invented urgency, exaggerated benefits, and the familiar “just checking in” filler. Verify every name, number, date, and product claim.
Keep a human review step before external messages. When a buyer declines, stop the sequence and record that preference. More activity is not progress if it erodes trust or creates inaccurate commitments.
Close the loop every week
Review a small sample of AI-assisted work. Count unsupported claims, missing commitments, editing time, and whether the intended next step occurred. Compare similar work before and after the change. A reply-rate increase alone does not prove the workflow improved qualified pipeline.
Use the prompt builder to create your first brief. Store successful prompts with a good example and a quality checklist in your team’s approved workspace. Revisit them when your offer or sales process changes.
A prompt to make this practical.
Using only these anonymized notes, extract: buyer goal, confirmed facts, open questions, agreed next action, owner, and date. Quote the supporting note for each commitment. Mark anything absent as unknown. Draft a concise recap for human review.
Use anonymized information in an approved tool. Verify the result before sending or updating records.
Customize your prompt →Try this today.
Pick one call this week. Create a brief before it, a verified recap after it, and a note on what AI improved or got wrong.
AI sales prompt builder ↗