When AI Makes Recognition Look Like Understanding

Updated: Oct 19

A good demo can make weak discovery feel productive. AI can make the story afterwards look complete.
Why technical interest, detailed questions and an agreed proof of concept can still leave the buying case unresolved.
I have been reviewing a set of recent customer conversations that, at first glance, looked encouraging.
The details here are anonymised, and the scene combines elements from several calls. The wording is reconstructed, but the pattern is real.
A customer described how their team investigated a difficult technical issue. They had dashboards, logs and experienced engineers, but the work still involved moving between tools, reading code and asking people to connect pieces of context.
The seller heard a clear connection to the product and opened the demonstration.
The customer leaned in. They asked about integrations, deployment, data, security and what the product could detect. The conversation became detailed. A proof of concept was discussed. Everyone left with the sense that the meeting had gone well.
When we reviewed the notes later, one question was surprisingly difficult to answer.
What decision was the customer trying to make better?
We could describe the capability we had shown. We could list the technical questions. We could explain what the proof of concept might demonstrate.
We were much less clear about what the customer would do differently if it worked, what consequence mattered, or who would decide whether the result justified a change.
The meeting had created momentum. I was no longer sure it had created enough discovery.
Why the meeting felt productive
Nothing about the seller’s move was irrational.
The customer had described a situation that sounded relevant. The product could contribute. Time was limited. A demonstration gave both sides something concrete to discuss.
The customer’s questions mattered too. Technical evaluators do not ask detailed questions for no reason, and it would be a mistake to dismiss their interest because it was not yet expressed as a business case.
That is why this pattern is easy to miss.
The meeting contains many of the signals we normally associate with progress: attention, specificity, challenge, curiosity and an agreed next step.
But those signals can tell us that the customer is evaluating the technology without telling us whether the technology is material to a decision they need to make.
A product connection can be correct and still arrive too early.
The answer that closed down the question
The most important point in the conversation may have been the customer’s first technical question.
A question about deployment might have been a route into the environment they were trying to change. A question about data might have revealed which evidence they trusted today. A question about integration might have exposed the manual work, handoffs or delays that remained after their existing tools had done their job.
Instead, the question was answered.
The answer was useful. It also moved the meeting further into the product.
Once the seller could explain the capability, the reason behind the customer’s question received less attention. The conversation became clearer technically while the customer’s situation remained only partly understood.
I used to think strong discovery was mainly about asking better questions before the demonstration began.
I now think it also depends on what happens after the customer asks a question that the seller is eager to answer.
The question can be treated as a request for information, or as evidence that there is more to understand. Often it is both.
Technical interest is not yet a buying case
Technical interest is valuable evidence. It shows that the customer can see a possible use, that the capability is credible enough to examine, or that an evaluator has identified a risk they need to resolve.
It does not yet tell us whether the issue is material.
For that, the conversation needs to reach the work around the technology.
What event or decision creates the need for attention? How is it handled today? What uncertainty remains after the current tools and people contribute? What happens because that uncertainty remains? Who feels the consequence? What would have to change for the customer to act differently?
These are not questions to force into every call as a script. A customer may not know the answers. They may not be ready to share them. The person in the room may be responsible for technical validation rather than the commercial decision.
The purpose is not to make the customer perform a business case for the seller.
It is to avoid treating product relevance as evidence of commercial materiality before the connection has been examined.
A proof of concept can succeed and still prove very little
The same distinction matters when the meeting ends with a proof of concept.
A good proof should reduce uncertainty that matters to a customer decision. That requires more than a list of technical tests.
It needs a baseline: what happens now.
It needs a question: what remains uncertain.
It needs an expected contribution: what the product may reveal or change.
It needs a decision consequence: what the customer will do differently if the evidence is strong enough.
Without those links, the proof can succeed technically and remain commercially unproven.
The product may integrate. It may detect the expected condition. It may perform well in the test environment. The customer may be impressed.
The team can still reach the end without knowing whether the result changes an operational decision, removes enough friction, reduces a risk the customer considers material, or justifies adoption alongside competing priorities.
That is when value gets reconstructed after the technical work. Estimates, benchmarks and seller assumptions are asked to do the job that discovery did not complete.
Sometimes that reconstruction is legitimate. New evidence can emerge during a proof. The risk is pretending that technical success established a buying case when the customer never agreed what the proof was meant to change.
Discovery needs permission to disqualify
This has changed how I think about planning customer meetings.
Before the call, I want the team to be clear about its purpose. Are we discovering a situation, demonstrating a capability, validating technical fit, or designing evidence for a decision? A call can do more than one of those things, but confusion about the primary job makes it easy to mistake activity for progress.
I also want to know what we need to learn before a proof of concept would be meaningful.
That usually includes the customer decision or event, the current mechanics, the uncertainty that remains, the consequence if nothing changes and the person who can judge whether the evidence matters.
The meeting does not have to produce a dramatic problem.
One valid outcome is that the existing process is sufficient. Another is that the problem is real but the product would not make a material contribution. A third is that the product may matter and the next step can now be designed around a decision rather than a demonstration alone.
If every discovery conversation has to end with product progression, the questions are not really free to discover very much.
The pause before the pitch
I still believe in demonstrations. For technical products, seeing the capability can be the fastest way to make a conversation useful.
The change is smaller.
When a customer says something that connects neatly to the product, I try not to treat the connection as permission to stop discovering.
I want to understand what the sentence refers to in their work, why it matters, and what evidence would change a decision.
Then the demonstration has a job.
Without that, it may still be a very good demo. It just may not tell us why the customer would buy.




Comments