A working demo answers whether the product can perform the task. With six weeks of runway left, the harder question is whether anyone will pay soon enough to keep the company alive.
In 1999, Louis Borders had already proved that Webvan could build an impressive online grocery operation. The company had warehouses, software, delivery vehicles and a plan to serve multiple US cities. It had also entered the dangerous period between proving that a system could work and proving that demand could support the system being built.
The test that does not appear in the demo
Webvan’s technology did real work. Customers could order groceries online and receive deliveries at home. The warehouses could process orders. The vehicles could move them.
That technical success created another decision: keep expanding the system or slow down until customers demonstrated repeatable demand.
Webvan chose expansion. The company committed heavily to warehouses and new markets before it had enough evidence that order volume, purchase frequency and delivery economics would support those commitments. In 2001, Webvan filed for bankruptcy.
Wired documented the collapse in “Why Webvan Drove Off a Cliff.” The lesson was larger than online groceries. A product can function as designed while the business around it remains unproven.
That is the same problem hiding inside a successful AI demo.
At 4:47 on a Friday afternoon, the final test passes. The extraction is accurate. The agent completes the workflow. The response arrives fast enough to show a buyer on Monday.
For ten minutes, the team has what it has been chasing for months.
Then the founder opens the runway sheet.
Six weeks remain.
The demo has removed the technical excuse. Now someone must decide whether to spend the remaining cash turning it into a product, use it to win a paid pilot, or stop and test whether the problem matters enough.
A passing test creates three different products
I have seen founders treat a working demo as permission to continue building. Usually, it creates three possible products.
The first is the product the team imagined. It needs onboarding, permissions, billing, monitoring, support and enough reliability for customers outside the founders’ laptops.
The second is the product a specific buyer will pay to test. It may handle one workflow, accept manual support and avoid the broader architecture for now.
The third is the product the market is already asking for, which may use only a small part of the demo.
These paths consume runway differently. The first asks the company to finance its own assumptions. The second can convert learning into revenue. The third may force the founder to abandon work that was technically difficult and emotionally expensive.
That is why the commercial decision gets harder after the demo works. Failure would have closed a path. Success leaves several open, and the company cannot afford to follow all of them.
Six weeks changes the burden of proof
With eighteen months of cash, a founder can fund exploration. With six weeks, each additional build needs to answer a commercial question.
Which named buyer has agreed to test it?
What must be true before that buyer pays?
Which part of the product blocks that payment today?
How much manual work can the team tolerate during the first delivery?
The answers should determine Monday’s work. A generic request for “production readiness” should not.
A buyer may need audit records before giving the product real data. Another may need one existing tool connected. A third may like the demo but have no budget owner, purchasing route or urgent deadline. Those are three different signals. Only the first two identify work that could move a purchase forward.
This is where a paid pilot can help, provided the buyer pays for the variation it requires. In Kwame’s pilot decision, the useful constraint was commercial: protect the core product while using a contract to extend runway.
A pilot should buy evidence. Payment confirms that the problem has some value. Usage exposes where the workflow fails. Renewal reveals whether the result matters after the novelty wears off.
Free interest proves less.
Decide what Monday is for
Before adding another feature, I would write three lines on the Friday runway sheet:
- The buyer whose decision matters next.
- The evidence missing from that decision.
- The smallest piece of work that can produce that evidence.
If no named buyer exists, Monday is for customer conversations and direct selling. If a buyer exists but cannot explain the cost of the current problem, Monday is for validating urgency. If payment depends on one narrow product gap, Monday is for building that gap and leaving the broader roadmap alone.
Webvan’s operations worked. That achievement made continued expansion feel supportable, even while the commercial evidence remained weak. The infrastructure then reduced the company’s room to change course.
A founder with six weeks does not have to repeat that sequence. Keep the demo. Cancel the celebratory roadmap meeting. On Monday morning, ask the closest buyer what has to happen for money to move, and build only against an answer you can verify.
Comments
No comments yet.