2.4 · How it works· 2 min read

People or agents: who leads at each stage

Who leads depends on how often something happens and how much is at stake. The goal is that the founder leads only where trust is made or lost: Choose, Sign and Stay. An agent prepares, a person sends.

The question isn't whether to use agents. It's where they should lead, and where a person should.

The answer depends on two things: how often the same thing happens, and how much is at stake when it goes wrong. In a small industry, a lot is at stake more often than you'd think.

Four ways to split the work

ModeWho does whatFits when
Rules run itFixed if-this-then-that. No judgement involvedIt happens often, and little is at stake
Agent leads, person watchesThe agent acts within limits. A person steps in on exceptionsIt happens often, in clear patterns
Person leads, agent helpsThe agent researches, sums up and flags. The person decidesJudgement matters, and speed still helps
Person owns itThe agent prepares. It never sends or decidesMuch is at stake, and trust is on the line

Stage by stage

StageWhat to watch forWhat agents doWhat people decideWho leads
NoticeA peer mentions you, an introduction, a first visitNote where they came from, and who mentioned youWho to thank, and howRules and agents
LearnRepeat visits, questions, a call bookedPull together what you already know about themWhat to say firstPerson leads, agent helps
ChoosePricing visits, more people on calls, a reference asked forPrepare the history, find a customer they'd trustReal budget, real urgency, real fitPerson leads, agent helps
SignProposal opened, contract questionsDraft the proposal, write down every promise madeEverything said to the buyerPerson owns it
StartSetup steps done or missed, first value reachedCarry the promises to onboarding, track milestonesWhat "live" means; step in when someone stallsAgent leads, person watches
StayUsage drops, sharp tickets, missed meetingsSpot risk early, pull together the full story, suggest a responseThe conversation with the customerPerson owns it
GrowA happy call, first value, a past user in a new jobSpot the moment, draft the introduction askWhether to ask, and the ask itselfAgent leads, person sends
IllustrationWho leads at each stage

The table shows who leads, not who takes part. Agents help in every stage. People answer for every stage.

What this means for a founder

Today you lead every row. The goal is that you lead only where trust is made or lost: Choose, Sign and Stay. Everywhere else, your team runs the loop, and you hear about the exceptions.

That's what "a sales loop your team runs" means in practice.

Exceptions, not approvals

There are two ways to get this wrong.

  • A person approves everything. Safe, and slow. Your onboarding lead can't check every message for every new customer. Things queue up, and the speed you wanted disappears.
  • Nobody watches. Fast, and risky. Mistakes reach customers before anyone notices.

The middle way: agents work inside agreed limits, and people get the exceptions, with context. Instead of checking every onboarding message, your onboarding lead gets the customers who stalled, with what happened and a suggested next step.

One agent, several stages

An agent doesn't belong to one stage. An agent that watches how customers are doing can track setup during Start, risk during Stay and the moment to ask for an introduction during Grow. Same agent, different jobs depending on where the customer is.

One rule we don't bend

From Choose onwards, nothing a buyer or customer receives should read as automated. An agent prepares. A person sends.

Questions this chapter answers

Which parts of selling should agents handle, and which should people keep?
It depends on how often something happens and how much is at stake. Rules and agents can note where a buyer came from and track setup. People lead where trust is made or lost: choosing, signing, and the conversation with an unhappy customer. From Choose onwards, an agent prepares and a person sends.
Do I have to approve everything an agent does?
No. Approving everything is safe and slow; nobody watching is fast and risky. The middle way is that agents work inside agreed limits and people get the exceptions, with context: what happened and a suggested next step.