Writing

Start with the bottleneck, not the technology

aieconomyriskworking-note

As a tech entrepreneur, I use these questions to identify valuable product opportunities and manage the risks of bringing them to market.

Economic value

How much economic value can this technology create?

The following points summarize Chad Jones’s lecture, “A.I. and Our Economic Future”. [1]

  • Automating one task even extremely well creates limited value when the rest of the workflow remains bottlenecked.
  • Abundant capabilities become cheap; scarce complementary tasks capture more value. Computing power grew enormously, yet deciding which questions to ask remains a constraint.
  • In Chad Jones’s simplified accounting example, making software unlimited would initially add only roughly software’s current share of GDP (about 2%). The figure is illustrative, not a forecast: large gains require automating complementary tasks across the economy.
  • AI may ultimately be worth “multiple internets,” but adoption and organizational change could take decades. Economic gains may arrive slowly, while security and systemic risks can arrive much sooner.

Problem–tool fit

What problem are we trying to solve, and what can this technology reliably do today?

Start with the bottleneck, then choose the most reliable, repeatable, and economical tool for removing it. Otherwise, the hammer effect takes over: when OpenClaw is the tool at hand, every problem starts to look like a nail. [2]

Risk management

Given the realistic estimate of the potential economic value outlined above, we need to manage the associated risks carefully.

1. Market

  • Demand intensity
    Is it merely “nice to have,” or urgent, recurring, and worth paying for?
  • Market readiness
    Is the solution still in R&D and years away from mass production, or is it mature, scalable, and supported by steady demand? If it is still in R&D, how much further development is required before the product is commercially viable, and what happens to customers, revenue, and defensibility if R&D slows or stops?
  • Market size
    Is the serviceable market large enough to support the target business?
  • Supply and demand
    Is demand growing faster than effective supply?What needs remain unmet
  • Competitive structure
    Competition can validate demand. Avoid markets that are saturated, highly commoditized, and offer no distinctive niche, product advantage, or acquisition edge.

2. Product

  • Which tech stack should we use, and when?
  • How should we allocate time and resources?
  • Which foundations, such as data storage, version control, and code security must be handled early, to what depth, and in what priority order?
  • What laws, licenses, privacy obligations, or liabilities apply?

4. Financial

  • What is the goal?
  • How much money is wise to spend for this goal?
  • Am I profitable?
  • What are the dependency risks?

References

[1] Chad Jones. “A.I. and Our Economic Future.” Stanford Graduate School of Business, July 16, 2026.
[2] Wikipedia “Law of the instrument”.