Agents Do the Work. Humans Own the Deal.

Where AI should lead in sales, and where accountability stays human

Imagine your new AI agent has been working on your top 200 accounts for a month. Meetings are up. Then it sends a CFO last year’s pricing, with a discount nobody approved. The CFO forwards it to your CEO. Now answer one question: whose name is on that email? If the room goes quiet, you have found the real AI problem in go-to-market. We talk about what agents can do. We hardly talk about who answers for what they do.

Agents are brilliant interns, not partners

Let me be clear: I am not an AI sceptic. Agents already do work that used to eat a big part of a seller’s week. Account research, call summaries, CRM updates, meeting prep, call scripts. If not done yet, hand it over, today. But an agent is not a colleague. You can’t coach it over coffee, you can’t sue it, and you can’t fire it. It will never look a customer in the eye after a mistake. The work can be delegated. The accountability can’t.

Where the line runs: two questions about every sales task

  1. How many right answers are there? A prospecting email, a discovery-prep brief or a call summary has many good versions. A price, a contract term or a forecast commit has one right answer and a lot of expensive wrong ones.
  2. What does a mistake cost? A clumsy first draft costs a minute. A wrong discount costs margin for three years. A wrong promise to a buying group costs the customer.

Put those two questions together and the line draws itself:

Many right answers One right answer
Mistake is cheap Agents lead. Research, briefs, call summaries, first-draft outreach, CRM hygiene. Rules automate. Routing, territory assignment, approval workflows.
Mistake is expensive Agent drafts, human decides. Business case, proposal, mutual action plan, value story. Human owns. Pricing and discounts, contract commitments, the forecast commit, the hard conversation with your champion.

Most AI disappointment in sales comes from skipping this map. Teams give an agent a job from the bottom-right corner because the demo looked impressive. A demo is not a track record.

Autonomy is earned, not installed

Treat an agent the way you treat a new hire. It starts in the safe corner. It earns more freedom when it proves, over weeks, that it is accurate, consistent and knows when to escalate. Not because the vendor said it is ready. Give every agent a boss. My rule: every agent-driven workflow in your GTM organisation has one named human owner. Not “RevOps”. Not “the AI team”. A name. That person defines what good looks like, reviews samples every week, decides when the agent may do more, and stands up when it goes wrong.

Plan for the failure, not the perfect run

Agents will fail. Not often, but unpredictably and at scale: the same mistake, sent to 4,000 contacts before anyone notices. So design for the moment after the failure. Who gets alerted? How do you stop the sequence? Who calls the customer? Klarna learned this in public. After replacing a large part of its customer service with AI, its CEO told Bloomberg in May 2025 that quality had suffered and that the company was hiring humans again. Savings that cost you trust are not savings.

Why this matters even more in sales

Buyers don’t buy from agents. They buy from people they trust to stand behind a promise. It is possible to win the sale and lose the customer, and an agent that overpromises will do exactly that, at machine speed. Look at what really moves a complex deal. An agent can collect the metrics. It cannot earn a champion. It can map the buying group. It cannot build consensus in a room where the CFO and the Head of Operations want different things. That is human work, and it is becoming the most valuable work in the deal.

Call to action for sales leaders

  1. Map your top 20 GTM tasks on the two questions. Hand the safe corner to agents this quarter. Stop debating it.
  2. Put a name next to every agent. No owner, no agent.
  3. Write the “when it goes wrong” playbook before go-live, not after the first incident.

Agents do the work. Humans own the deal. That is not a limitation of AI. It is the definition of selling.

Go deeper. Qualifying a deal and building the business case are exactly the human work an agent can’t own. Both are part of my 2ENGAGE Training Catalogue: Deal qualification & progression – inspired by MEDDPICC and Acquire customers by creating business value.

Scroll to top