Sales — Charter
What sales owns in an AI-operated company — pipeline, qualification, forecast discipline, and where automation genuinely helps versus where it damages the relationship.
Sales owns revenue and the truth about what is coming. The second part matters more than most sales functions admit: a forecast everyone privately discounts is worse than no forecast, because Finance plans against it and Operations staffs against it.
In an AI-operated company research, note-taking and follow-up automate well. Qualification and trust do not.
What this role owns#
Pipeline. What is in it, at what stage, and honestly what it is worth.
Qualification. Whether an opportunity is real. The discipline to disqualify early is the highest-value habit in the function.
Forecast accuracy. Not optimism — accuracy.
The customer relationship through to handover.
Commercial terms within agreed boundaries.
KPIs#
| Measure | Why this one |
|---|---|
| Forecast accuracy | Actual against forecast at each stage. The credibility metric |
| Win rate by segment | Where you actually win, which is often not where you spend time |
| Sales cycle length | And where deals stall within it |
| Pipeline coverage | Pipeline against target. Below ~3x, the target is at risk regardless of effort |
| Disqualification rate and timing | Disqualifying late is expensive; disqualifying early is skill |
| Cost of acquisition against lifetime value | The unit economics test |
| Time spent selling | Share of the week actually with customers rather than on admin |
The last one is the metric AI most directly moves, and it is worth measuring before and after.
Where AI genuinely helps#
Research before a conversation. Company, market, likely pressures, recent news. Turns a generic call into an informed one.
Note-taking and follow-up. Capturing what was said and what was promised, then producing the follow-up. High-volume, mechanical, and the thing that slips when a week gets busy.
CRM hygiene. Keeping records current without a person retyping them.
Proposal drafting from a template plus the specifics of the conversation.
Qualification prompts. Surfacing the questions that were not asked.
Where it damages the relationship#
Worth being blunt, because the tooling encourages it:
Personalisation at volume that is obviously automated. A message that references your recent funding round and nothing else signals a script, not attention. The floor for "personalised" has risen; generated outreach that reads as generated is worse than an honest generic message.
Volume outreach. Sending more because sending is cheap. It burns the domain, the list, and the brand.
Automated responses to a real objection. If a customer raises a concern, a fast generic reply tells them nobody read it.
Anything that misrepresents who they are talking to. Beyond the trust question, EU AI Act transparency obligations applied from 2 August 2026 — people must be told when they are interacting with an AI system.
AI agents in this function#
Account research agent — assembles a briefing before a call, with sources.
Call notes agent — captures the conversation, extracts commitments and next steps. A human confirms before it enters the CRM.
CRM update agent — keeps records current from meetings and email.
Proposal assistant — drafts from an approved template plus the specifics.
Pipeline hygiene agent — flags stale opportunities, missing next steps, and forecast entries with no recent activity.
What stays human: every customer conversation, qualification judgement, pricing decision, negotiation, and the forecast commitment. An agent can tell you a deal looks stalled; it cannot tell you whether the champion has gone quiet because they are busy or because they have lost.
SOPs#
- Qualification — a defined framework applied consistently, with the discipline to disqualify.
- Pipeline review — weekly, with next steps and dates. An opportunity without a next step is not an opportunity.
- Forecast commitment — what "committed" means, and what it costs to miss one.
- Handover to delivery — what is passed, what was promised, and who accepts it. Most delivery disappointment starts as a sales promise nobody wrote down.
- AI disclosure — any automated interaction discloses that it is automated.
Templates#
Qualification framework · proposal · BRD for scoping an engagement · handover record · pipeline review format.
Workflows#
In: leads from Marketing · inbound enquiries · renewals · referrals.
Out: closed business · forecast · handover to delivery · market feedback to Product and Marketing.
Handoffs: Marketing for lead quality feedback · Legal for non-standard terms · Finance for forecast and pricing approval · Delivery for handover.
FAQ#
Can AI do outbound sales?#
It can research, draft and schedule. It should not be the one having the conversation, and it must disclose itself if it is. The reason is commercial before it is legal: volume outreach that reads as automated devalues every future message you send.
How does AI improve forecast accuracy?#
By removing the excuses — records stay current, stalled deals surface, and activity is visible. The forecast still depends on honest judgement about whether a deal will close; AI just removes the places to hide.
Should sales use AI-generated proposals?#
As a first draft from an approved template, yes. Sent unreviewed, no — the errors are exactly the kind a customer notices, and a wrong figure in a proposal is expensive in two directions.
What is the biggest mistake?#
Automating the activity without improving the qualification. More outreach into a poorly qualified list produces more noise, a worse reputation, and a forecast nobody believes.
What else is coming for Sales
Charter Ready
What this department owns and is accountable for.
KPIs Not yet
The numbers it is judged on.
AI Agents Not yet
What is automated, and what stays human.
SOPs Not yet
How the recurring work is done.
Templates Not yet
The documents it produces.
Workflows Not yet
How work enters, moves and leaves.