Role playbook

AI for sales.

What wins deals is not what fills the day. Research, lists, notes, CRM hygiene, follow-ups and proposals are the tax around selling. AI automates most of it; whether that is good news depends on the selling left.

This assumes you have the base. Start there if not

The job

What this role actually does.

Six things every week. Model releases do not change them; this page should still hold in three years.

01Prospecting 02Outreach 03Discovery 04Proposals 05Objections 06Pipeline

Most examples are vendor customer stories. Their self-reported, unaudited numbers show something was built and used—not a benchmark to expect.

Prospecting and account research

Research

The use case: Building a real account picture at list scale: funding, hiring, org changes, earnings commentary, and which existing customer they most resemble.
Job to be done: When I plan my week, I want every target account researched to the standard I used to reserve for the top ten, so I choose where to spend my hours on evidence.
Value to the business: Better-qualified pipeline at the top, and rep hours moved out of browser tabs and into conversations.
How to evaluate it: Selling hours per rep per week, and meeting-accepted rate on researched accounts against the old baseline. Verify facts before they reach a customer: one wrong detail costs more than no detail.

What AI does here

AI assembles account pictures at volume: announcements, new hires, earnings calls and similar customers. Enrichment platforms make research a property of the list, not a heroic act.

Workflow map

InputAccount list, ICP, CRM history and public company signals.
AI-assisted processEnriches records, summarizes changes and surfaces likely stakeholders.
Human checkpointRep verifies material facts and qualifies which accounts merit outreach.
OutputA ranked account brief with a defensible reason to contact.

What is still yours

Choose who is worth the effort. A tool researches two thousand companies that will never buy. Qualify, disqualify early and find accounts with budget and a real problem.

Outreach and sequencing

Drafting

The use case: Drafting a first touch and follow-ups that open with something specific and true about that account.
Job to be done: When I contact someone cold, I want a message grounded in their actual situation, so I earn a reply instead of joining the generic pile.
Value to the business: Reply rates that survive the collapse in inbox tolerance for automated outreach.
How to evaluate it: Reply and positive-reply rate per sequence, with unsubscribe and spam-complaint rate as the counterweight. If volume rises while reply rate falls, the tool is burning the channel.

What AI does here

AI personalises at volume: an opening that proves you know something specific and true, attached to a consistent message. Generic AI outreach fills inboxes, so the bar for personal has risen.

Workflow map

InputVerified account brief, approved value proposition and contact role.
AI-assisted processDrafts a relevant opening and follow-up sequence from those facts.
Human checkpointRep checks relevance, claims, tone and send timing.
OutputA short, personalized message ready for accountable sending.

What is still yours

Have something worth saying. Personalisation cannot rescue a message without insight. If it is not interesting by hand, automation gets you ignored at scale.

Discovery calls

Meetings

The use case: Capture and structure (transcript, summary, next steps, CRM fields) so the rep can listen instead of type.
Job to be done: When I am on a call, I want the record made for me, so I can pay full attention and still leave with a clean account of what was said.
Value to the business: Better-qualified deals, and CRM data that actually exists.
How to evaluate it: CRM completeness on open deals, time to follow-up after a call, and whether qualification criteria are recorded rather than assumed. Reps correct the summary: an unchecked transcript becomes an unchecked belief.

What AI does here

Record and transcribe calls; let the summary track commitments and objections so you can be present. Then ask what you missed and failed to follow up. The review is uncomfortable and useful.

Workflow map

InputConsent, call recording, account context and discovery goals.
AI-assisted processTranscribes, summarizes commitments and flags unanswered topics.
Human checkpointRep validates the record and chooses the next question and action.
OutputShared call notes, follow-up tasks and a better prepared team.

What is still yours

The conversation. Hear hesitation, abandon the agenda when the real problem surfaces and ask the uncomfortable question. Recording a call is not running one.

Proposals and pricing

Drafting

The use case: Assembling proposals, RFP responses and security questionnaires out of a reviewed content library.
Job to be done: When a proposal is due, I want the already-answered questions answered from approved content, so I only write the parts that are genuinely new.
Value to the business: More proposals per person, and consistent answers on the questions where a wrong one is expensive.
How to evaluate it: Turnaround time per response, share of content reused from the library, and the rate of answers flagged in review. Nothing leaves without a named human owner.

What AI does here

Most proposals are assembly. Keep cases, security answers, terms and pricing logic together; generate the tailored draft from call transcript and account research. RFP extraction against prior answers cuts days to an afternoon.

Workflow map

InputCall transcript, approved pricing rules, case studies and security answers.
AI-assisted processExtracts requirements and assembles a proposal or RFP response draft.
Human checkpointOwner checks scope, commercial terms, security claims and price.
OutputA tailored proposal with approved evidence and clear exceptions.

What is still yours

The number. Pricing weighs leverage, urgency and what the buyer can defend to finance. No model knows how badly you need the deal.

Objections and negotiation

Synthesis

The use case: Mining the recorded call library for how objections were actually handled, and rehearsing before the real conversation.
Job to be done: When I face a hard objection, I want to see how the best people here answered it, so I can prepare rather than improvise.
Value to the business: New reps reach quota sooner, and the answers of top performers stop leaving when they do.
How to evaluate it: Ramp time to first closed deal, win rate on deals where the objection appeared, and coaching hours per manager. Keep it coaching: call analysis repurposed as surveillance kills the recordings' value.

What AI does here

Use AI to rehearse against a hostile buyer and find which objections kill deals, when and how winners answered. Conversation intelligence on recorded calls is more honest than team opinion.

Workflow map

InputRecorded calls, deal stage, competitor context and prior outcomes.
AI-assisted processFinds objection patterns and creates role-play prompts from the evidence.
Human checkpointManager tests whether the pattern is causal and coaches the response.
OutputA practical rehearsal plan and objection library for the team.

What is still yours

Everything in the room. Negotiation is a relationship under pressure: silence, timing, walking away and costly concessions. Rehearsal helps; it does not participate.

Pipeline and forecasting

Extraction

The use case: Reading deal activity across the pipeline to flag risk, stalled deals and forecast variance without the spreadsheet ritual.
Job to be done: When I commit a number, I want the pipeline's real state surfaced from activity rather than from optimism, so the forecast survives the quarter.
Value to the business: A forecast leadership can plan hiring and cash against.
How to evaluate it: Forecast accuracy against actuals over several quarters, and hours spent assembling it. The rep still owns the call; a model that has never met the buyer is a second opinion.

What AI does here

Forecast accuracy is mostly data quality. AI writes CRM records from calls, flags quiet deals and exposes where nobody has met a decision maker—the flag that saves quarters.

Workflow map

InputCRM stages, activities, call signals, close dates and historical outcomes.
AI-assisted processUpdates records, detects risk and estimates forecast scenarios.
Human checkpointLeader challenges assumptions and owns the commit number.
OutputA cleaner pipeline and an explained forecast with actions.

What is still yours

Call the number and stand behind it. A model says a deal looks weak; it will not tell your VP. Owning an occasionally wrong forecast matters.

Start this week

Record your next five calls and ask, for each one, what you failed to follow up on.
Build one library holding your case studies, security answers and standard terms, then generate your next proposal from it rather than from the last proposal.
Take your ten best-fit accounts, research them properly with a model, and compare the result against the template you were about to send them.

Where it fails in this role

Volume as a strategy. Sending ten times more mediocre email is not leverage, it is a deliverability problem that becomes a domain reputation problem that becomes somebody else's job to fix.
Sounding human. Buyers now recognise the register instantly. Over-polished, over-structured, three-bullets-and-a-question email reads as automated whether or not it was.
Relationships. Trust is built by a person doing something slightly inconvenient on your behalf. That has not changed and is not going to.

Other playbooks

All nine playbooks Product management Marketing Venture capital Operations Consulting Forward deployed engineer Founder / entrepreneur AI product manager

Now do it somewhere real.

A playbook is a map. The Industry Fellowship is the terrain: a term spent talking to professionals who are implementing AI in one industry, working out where it actually creates value, and building a working prototype against what you find. Most fellows start as beginners.

Explore the Fellowship Apply for the Class of 2028