04 · 40 minutes
AI for sales
Start with the work. Every stage of your sales lifecycle already has a human job — the question is what part of it AI should be trained to do.
The lifecycle
Stage by stage, human job and AI job
- Human job
- Decide which markets, routes and account types are worth our capacity.
- AI job
- Turn a loose idea of a good customer into a written, testable ideal customer profile.
- Quick win
- Have AI interview you about your best five accounts and draft the ICP from your answers.
- Workflow
- ICP document → territory list → research queue.
- Agent
- Monitors a defined market and adds newly qualifying companies each week.
- Requires human judgment
- Whether the route economics actually work.
The centerpiece
Research used to be a task. Now it can be an assignment.
I recently used AI to investigate roughly 100 of the best potential marketing-services prospects in San Francisco. The point isn't marketing services — it's that AI did significant market research while I did other work.
The old way
- Website
- Notes
- Qualification
- Repeat
- Repeat
- Repeat
Specify once
- Define ICP
- Define geography
- Define evidence
- Define disqualifiers
- Define scoring
- Define output
Then run it
- Assign research
- Go do something else
- Return to results
AI doesn't always need you sitting there watching it.
Eight questions the assignment must answer
- 01Who exactly are we looking for?
- 02Where?
- 03What makes them attractive?
- 04What makes them a bad fit?
- 05What evidence should AI find?
- 06How should opportunities be scored?
- 07What information will the salesperson need?
- 08What should the final deliverable look like?
Scenarios this room can run today
- —Find strong healthcare prospects in our service area.
- —Find hospitality companies that fit our route.
- —Find manufacturers likely to need uniform rental.
- —Find food-processing prospects.
- —Find strong local independent opportunities.
- —Find accounts with signals suggesting a provider-change opportunity.
Build it now
Prospect research assignment builder
1. Specify the assignment
Your answers save on this device, so you can walk away and come back to them later in the day.
2. Send the assignment
You are a senior B2B market researcher for a commercial laundry, linen and uniform services company. Your job is to produce a sales-ready prospect research deliverable — not a list of names. TARGET MARKET [target market] GEOGRAPHY [geography] TARGET INDUSTRIES [industries] IDEAL CUSTOMER PROFILE [what makes a company a strong fit] MINIMUM SIZE [minimum size threshold] DO NOT INCLUDE [exclusions and disqualifiers] QUALIFICATION CRITERIA [what must be true for this to be a real opportunity] EVIDENCE REQUIRED For every company, cite specific evidence you actually found, with links. [what evidence counts] If you cannot verify something, say "not verified" — never estimate and present it as fact. SCORING Score each company 1-10 on fit, need signal, accessibility and likelihood of switching, then give an overall score with one sentence of reasoning. HOW MANY Return 25 companies, ranked best first. OUTPUT FORMAT A table with these fields: Company, location, size estimate, fit score, why it fits, evidence (with links), likely need, best contact role, personalization hook, suggested opening line After the table, add: the three strongest opportunities and why, any patterns you noticed in this market, and what information a salesperson would still need to gather. RULES - Work the list systematically. Do not stop early. - No generic filler. Every line should be usable on a call. - Flag anything that looks like an existing customer or a bad fit and exclude it.
3. Walk away — then come back
AI doesn't always need you sitting there watching it.
What should come back
- Company
- Fit score
- Why it fits
- Evidence
- Likely need
- Personalization
- Possible opening
After the meeting
One meeting, ten useful outputs
- —Summary
- —Pain points
- —Buying signals
- —Objections
- —Missing information
- —CRM update
- —Follow-up draft
- —Tasks
- —Next steps
- —Proposal inputs
Before the call
The brief every rep should walk in with
- —Company summary
- —Recent developments
- —Potential needs
- —Relationship history
- —Key people
- —Likely objections
- —Questions to ask
- —Personalization opportunities
Coaching
What a reviewed call can tell a sales manager
- —Discovery score
- —What worked
- —What was missed
- —Question quality
- —Talk/listen balance
- —Objections handled
- —Coaching suggestions
- —Better questions
- —Next-call plan
Pipeline + retention
Using the information you already have
Pipeline intelligence
- — Stalled deals
- — Missing next steps
- — Weak follow-up
- — Risks
- — Priority
- — Manager coaching opportunities
Retention signals
- — Unresolved issues
- — Account risk
- — Sentiment
- — Service trends
- — Renewal preparation
- — Expansion opportunities
Add AI to the workflow
Don't build a separate universe called “AI”
- Website leadAI researchesCRM
- Sales meetingAI analyzesCRM + email + tasks
- Customer complaintAI evaluatesSuggested response + escalation
- Expert conversationAI structuresContent system
- CRM dataAI reviewsPipeline intelligence
- Existing workflow
- AI appears where it's needed
- Work continues