Building an agentic commercial workforce

AI agents can now run most of the work that sits around business development: monitoring accounts, preparing briefs, keeping records current, drafting follow-up. What they cannot do is create access to someone who does not know you. The opportunity is to automate the first and reinvest the time in the second.

What actually changed

The useful reading of the last few years is not that AI made selling easier. It made one specific part of selling almost free: producing plausible, personalised, well-researched contact at volume. Something that took a skilled person twenty minutes now takes seconds.

When the cost of an action falls to near zero, everyone takes that action. Buyers in most European B2B segments now receive more well-crafted, apparently personal outreach than they can process, so they have stopped processing it. The answer is not better emails. Everyone has better emails now.

The real gain sits elsewhere. Agents are genuinely good at the continuous, unglamorous work that used to consume a senior person's week, and that work is where most commercial time actually goes.

What an agentic commercial workforce looks like

The word "agent" is used loosely. What matters in practice is narrow: an agent is a defined job, run continuously, with an output someone actually uses. Five that earn their place in a commercial function:

Note what these have in common. Each one has a single job, a defined trigger and an output a person reads. None of them is asked to hold a relationship or make a commercial judgement. That boundary is the difference between an agentic workforce that compounds and one that quietly generates work nobody trusts.

Where the human stays

The parts of business development that resist automation are the parts that depend on someone being accountable rather than merely informed.

• {{ l }}

This is not a claim that these things are permanently beyond automation. It is an observation about where the constraint sits today. In practice the bottleneck in most commercial functions was never writing an email or building a list. It was having someone senior enough to open a real conversation and stay with it until there is a decision.

Why partner channels compound alongside agents

A partner channel is a way of borrowing trust that already exists. The reseller, consultant or technology partner has spent years earning access to a customer base, and an introduction from them carries weight a first-time approach cannot.

That matters more, not less, once everyone has agents:

Agents also make partner channels more practical to run. The reason most partner programmes fail is neglect, and neglect is a monitoring problem: nobody noticed a partner had gone quiet. That is precisely the kind of job an agent does well.

Building it, in the right order

Most companies approach this backwards. They deploy agents first, then discover the agents have nothing reliable to operate on. The dependency runs the other way: an agent amplifies whatever your targeting already is, so unclear targeting produces noise faster and more expensively.

Three layers, built in this order:

A practical sequencing note: build one agent, use it for a month, and only add the next once the first is genuinely relied upon. Commercial teams abandon agents that produce output nobody reads, and five mediocre agents are worse than one that is trusted.

What to be sceptical of

A short list, based on what tends to disappoint.

• {{ s }}

A reasonable position to take

Build agents for the information work: monitoring, research, preparation, admin, reporting. Assume they give you no advantage in outreach itself, because your competitors have the same tools. Then spend the time they free on what has become scarce, which is developing a small number of real relationships, directly and through partners.

The companies that do well over the next few years will not be the ones with the most agents. They will be the ones who used agents to buy back senior commercial attention, and then spent it on access.