The strongest signal came from what customers repeated every Monday: they returned to the weekly workflow and treated the polished AI output as optional. The company therefore had to sell the recurring decision process, with AI supporting the work where it earned a place.
At 9:07 on a Monday morning in Accra, Kojo sat beside a cooling mug of tea and watched the session recordings from Friday’s pilot. Kojo is a composite founder, but the decision in front of him is common: his small team had spent six weeks improving an AI-generated planning brief, while customers kept opening the plain weekly board underneath it.
One customer read two lines of the brief, closed it, and moved three overdue items into the current week. Another skipped the brief entirely and updated the person responsible for a delayed task.
Kojo replayed the recordings. The same thing happened each time.
The team had enough runway to keep building, but not enough to misunderstand the product for another quarter. If Kojo continued selling the AI brief as the centre of the experience, the next pilot could end with polite praise and no renewal.
The behaviour contradicted the pitch
The demo looked good. A customer could enter project information and receive a structured AI summary covering delays, priorities and possible next actions. During sales calls, that moment held attention.
Monday exposed a different job.
Customers needed a reliable place to decide what would happen that week, who owned it and which delay required intervention. The generated brief occasionally helped them see a pattern. The board helped them run the meeting.
That distinction matters because founders often give the most advanced feature the most strategic weight. It consumed the hardest engineering work. It produced the best demo reaction. It also made the product easier to describe as an AI product.
None of those facts proved that customers would return for it.
Kojo had been reading demo enthusiasm as product demand. The session recordings gave him a harder form of evidence: repeated behaviour when nobody from his team was present to guide the customer.
The decision was larger than a feature change
Removing the AI brief would have been premature. Keeping it at the centre would have ignored what customers were showing him.
Kojo wrote two possible product statements on paper.
The first described a tool that generated intelligent project updates. The second described a weekly operating workflow that helped small teams turn scattered work into named decisions.
The second statement matched the customer’s Monday. It also changed the roadmap.
The next engineering cycle would focus on making ownership, carry-over work and unresolved decisions easier to see. The AI brief would remain available, but it would draw from the workflow and point back into it. Its value would depend on helping someone make a specific call, rather than producing an impressive block of text.
This is the uncomfortable part of product strategy. A founder may need to reduce the billing of the feature that took the most effort to build. That effort is already spent. The next decision should follow observed value.
I have seen the same tension appear when a credible request starts pulling a product toward a second identity. The issue is rarely whether the request sounds reasonable. The issue is whether the company can support the new direction without splitting its limited runway across two products. [Kojo’s two credible requests](\/blog\/kojo-s-two-credible-requests-one-runway-could-become-two-companies-39225eb8\/) explores that decision from another angle.
Monday use deserves more weight than demo applause
A demo measures whether someone understands the possibility. Recurring use measures whether the product fits into work that already has consequences.
That does not make demos useless. A strong demo can open a serious conversation. It can also hide the manual steps, existing habits and internal pressure that determine whether the product survives after the call.
The useful questions begin after the customer logs in alone:
What did they open first?
What did they ignore?
What did they copy into email, WhatsApp or another tool?
What action did they complete before leaving?
That last question often reveals the product more clearly than a list of requested features. In one AI procurement workflow, the revealing behaviour was people copying answers into WhatsApp because the decision continued there. The output mattered, but the destination showed where the work actually moved. [Ama’s procurement assistant story](\/blog\/ai-procurement-assistant-what-copying-answers-to-whatsapp-taught-ama-about-decisions-4d08affc\/) follows that signal further.
The next demo started with the weekly meeting
By Thursday, Kojo had changed the order of the pitch.
He began with the Monday board: overdue work, named owners and the decisions still blocking progress. Only then did he show the AI brief, attached to a real question inside that workflow.
The screen looked less dramatic. The product made more sense.
Kojo still had an unresolved question about pricing. Customers might pay for the workflow, the AI assistance or the combination. He did not pretend one week of observation had answered it.
But the next test was now precise. He would ask pilot teams to run their weekly meeting without guidance, watch which actions they completed and see whether the AI brief changed a decision. On the following Monday, the board would carry the meeting. The side feature would have to earn its way back into the centre.
Comments
No comments yet.