AI as a sales copilot — where it helps and where it hurts
AI Sales 12 min
0/19
Business problem
Salespeople use AI in two extreme ways: either 'I don't use it because it will ruin the client relationship', or 'AI will write everything for me, I'll just send'. The first wastes time on work AI does 10x faster. The second sends clients content without context, with errors, and forgets verification — AI hallucinates, and the client senses it.
Proven method
A sales copilot works on 3 levels:
- BEFORE the conversation: company and person research, business context, buying signals, potential pain points, preparing questions and hypotheses,
- DURING: AI as a question assistant (only if it doesn't distract; otherwise notes after),
- AFTER the conversation: transcription, summary, extracting pain/needs/objections/buying signals, next steps, follow-up. HARD RULES:
- CLIENT DATA: don't feed confidential data to AI without consent — carry anonymized context or pick a tool with a proper policy,
- HALLUCINATIONS: AI can invent facts about a company (size, decision makers, events) — verify every fact before using it in a conversation,
- VERIFICATION: AI is an assistant with an error-prone memory, not a source of truth. The rule: content that goes out to the client always passes through a human.
Market example
- BAD
- Improvised general pitch lacking buyer context or verified data.
- GOOD
- Before a meeting with a performance agency: AI (from research) prepares a brief: the company has 12 people, opened a branch in Gdansk (trigger), the Head of Sales joined 3 months ago (a new role — openness to change), leads in spreadsheets (a pain hypothesis), competitor: HubSpot Starter. AI proposes 5 discovery questions. You verify the facts (you check the website, LinkedIn) and pick 3 questions. After the meeting AI generates a summary and follow-up — you edit and send. The client gets the impression you understand them — because you have context, not because 'AI wrote it'.
Ready-to-use prompt
You are a B2B sales consultant. Context: I'm a salesperson. I sell [PRODUCT]. Customers: [SEGMENT]. Task: Design my sales copilot: 1. Where AI helps me in the process (before/during/after the conversation) — specific tasks 2. Where AI should NOT decide (relationship, emotions, decisions) — and what I do then 3. Data rules: what I can give AI (context) and what I can't (sensitive data) + how to anonymize 4. Risks: hallucinations — how to verify facts AI gives about a company 5. My rule: what goes out to the client from AI, and what always passes through me Format: 5 sections.
Model rekomendowany: Claude 3.5 Sonnet / GPT-4o Copy and paste into ChatGPT, Claude or Perplexity
ALTERNATYWNE SCENARIUSZE
Variations & alternative scenarios
Solo salesperson version: a copilot in an AI notebook (one prompt per conversation).
Team version: shared prompts + a briefs database + a data policy.
Practical exercise
List 5 tasks in your sales process that you do by hand (research, notes, follow-up). Next to each write how AI can help. Write down 2 hard rules: what data does NOT go into AI and how you verify facts.
Zapisano w przeglądarce