AI in sales: Land of Confusion

Land of Confusion is a Genesis classic from 1986. Phil Collins sang about a world in which too many people talk and too little actually moves. Forty years on, the song describes the AI market better than any study.

I am having a lot of conversations with larger companies at the moment. AI is on everyone’s agenda. But as soon as it gets concrete, things stall. Pilots start and go nowhere.

The numbers are uncomfortably clear

This is not an isolated case. According to IDC, around eighty-eight per cent of all AI pilots never make it into broad rollout. For every thirty-three pilots started, four reach production.

Burned energy for everyone involved. For IT, which provisions the environments. For the business units, which invest time. And for the vendors, who run a proof of concept that never turns into a contract.

What actually causes the failure

What I hear in conversations matches what the numbers show. Most projects do not fail on the technology.

They fail because the data is not ready. A language model reading a CRM in which half the contacts are out of date and opportunities have not been touched for two quarters does not produce insight, it produces well-phrased nonsense.

They fail because the system landscape is too complex. Between CRM, marketing automation, ERP and three home-grown islands there is no interface, there is a negotiation.

And they fail because the scope was chosen too large from the start. Anyone beginning with a company-wide AI approach gets lost in alignment before anything runs.

On top of that come data protection concerns, regulation, cost pressure and a range of vendors that creates more confusion than clarity.

What really helps in sales

My observation is simple: whoever takes one concrete, specific problem and works on it actually makes progress.

In sales that problem is usually the same one. The average salesperson spends a considerable part of the day on research and administration. Time that generates no revenue and that nobody misses once it is gone.

That is exactly where the lever sits. Summarise customer information automatically. The relevant news, the structure of the buying centre, the history so far, the open points from the last conversation, all in one place. The salesperson walks into the meeting prepared instead of losing hours.

This is not a visionary scenario. It is the kind of tool you can build in a manageable amount of time, and I have built exactly such tools myself.

The second lever is momentum

The most expensive sales call is not the one that goes badly. It is the call to somebody who does not have the problem at all.

Even so, many companies respond to a thin pipeline with more activity. More lists, more calls, more contacts. What is often missing is not activity but direction.

What really helps is recognising momentum and using it. A vendor announces end of life for a product. A regulation comes into force. A migration becomes unavoidable. In those moments there are customers actively looking for solutions, and customers for whom the topic has not landed yet. Telling the two apart is the actual job.

And that is a job AI is good for. An agent can identify the accounts in the CRM where such momentum applies, including the matching use case and the right contact in the buying centre. That turns a cold call into a conversation at the right time about a relevant problem.

What AI does not take over

In its Portrait of a High-Performing Seller, Gartner describes three capabilities that make the difference. Working with AI has the largest effect on quota attainment. The other two are tactical flexibility and mentalizing, meaning the ability to switch perspective and understand what actually moves the customer.

And that is the part AI does not take off your hands.

In Germany, by the way, the biggest fear in the buying centre is rarely the wrong technology. It is the prospect of having to explain to your own boss that you bought the wrong solution. Anyone who understands that sells differently. No model in the world supplies that understanding.

The best combination is therefore both. AI removes the time sinks. What happens with the time gained is still decided by a person.

The biggest pilot does not win. The one that solves a real problem does.