When runway is measured in days, spend the next month on customer conversations unless paid demand already tells you exactly what model access will unlock. More capability cannot rescue a product whose buyer, use case, and price are still guesses.
In April 1970, Apollo 13 was on its way to the Moon when an oxygen tank exploded. Jim Lovell, Jack Swigert, and Fred Haise had a spacecraft with rising carbon dioxide and a life-support problem whose outcome was uncertain. The lunar module carried square lithium hydroxide canisters; the command module had round fittings. Engineers in Mission Control had to make the available equipment work with what the crew had onboard.
NASA’s account of the mission records the improvised adapter that let the crew use the square canisters. The point was not to build a better spacecraft in real time. It was to keep the people in the spacecraft alive with the materials already available.
Twelve days of runway creates a smaller version of that constraint. Friday arrives, the model credits are nearly gone, and the question looks technical: renew access, test the newer model, improve the demo. The actual question is commercial: what evidence could change the company’s chances before the cash does?
Model access is useful only when the next test is clear
A model subscription can feel like progress because it keeps the product moving. There is always another prompt to tune, another agent flow to connect, another benchmark to improve.
That work earns its place when a specific customer has already shown the path. Perhaps a logistics lead in Accra has agreed to test a document workflow if it can handle a defined set of invoices. Perhaps a small team in Berlin has said they will pay if the system cuts a recurring manual step. In those cases, model access supports a known test with a real person on the other side.
The dangerous version is different. The team renews credits because the demo still feels unfinished, while nobody has committed to a pilot, described the cost of the current workaround, or named the budget owner. A smoother answer from the model may make the product more impressive. It does not make demand more likely.
The decision should turn on one question: what will the credits help you learn that five customer conversations cannot?
If the answer is “whether the model can perform the task at all,” run the smallest possible test. If the answer is vague, protect time for conversations.
Conversations can expose the problem behind the request
Founders often hear a request for AI and assume they have found a use case. A customer asks for an assistant, a chatbot, or an automation layer. Those are descriptions of a proposed solution. They are rarely a complete account of the problem.
The useful conversation goes one level deeper. Ask what happened the last time the work failed. Ask who notices, who fixes it, what gets delayed, and what the failure costs. Ask what they use today when your product is unavailable. Then ask whether they would put money behind a narrow pilot.
That sequence can be uncomfortable. It can also save weeks.
A prospective customer may say they want an AI support tool, then reveal that the real constraint is approval. An unreviewed reply could damage a retailer relationship. In that case, a fully automated demo is pointed in the wrong direction. A review queue, clear escalation, and a smaller pilot may be the product worth testing. That is the pressure behind Nia’s approval gap.
Customer conversations also reveal when a problem has urgency but no buyer. A founder can then stop treating interest as validation. That distinction matters more when the company has days, not quarters, to decide.
Use the remaining credits as an instrument, not a habit
There is a middle path between cancelling every tool and continuing as before. Put a hard boundary around the remaining model access.
Choose one customer-backed workflow. Define what successful output looks like before opening the model console. Decide how many examples are enough to tell whether the workflow is viable. Set a date when the test ends, even if the result is incomplete.
Then take the evidence back to the customer. Show the narrow version. Ask for the pilot, the introduction, the data sample, or the reason they cannot buy. Each response creates a decision the company can act on.
This is especially important for founders selling across African, European, and US markets. A buyer may like the same capability for different reasons, under different procurement pressure, with different tolerance for risk. A model demo cannot tell you which condition matters. A serious conversation can.
Apollo 13 did not have the luxury of designing an ideal system. The crew and Mission Control used the materials they could reach, for the immediate problem they could define. A founder with twelve days should apply the same discipline: keep only the technical spend that helps answer a live customer question, then put the rest of the week where the answer can change the company’s direction.
On Monday, write down the five people whose answer could alter your roadmap or cash position. Contact them before you renew anything.
Comments
No comments yet.