# What does it take to put intelligence inside a product people use?

It takes more than a model call. The product needs a fast path for the common case, a fallback when a provider fails, keys kept off the device, and a measure of the speed users feel. Most of the work is the part around the model, and it keeps running after launch.

0 posts

## Posts

Nothing in this topic yet. The longer arguments below cover the ground.

## Where this leads

- [The checklist](https://octyn.co/compare/production-ready-ai-system)
- [Six months later](https://octyn.co/compare/cost-to-change-ai-workflow-six-months)
- [Product system](https://octyn.co/glossary/product-system)
- [AI product integration](https://octyn.co/glossary/ai-product-integration)
- [p95 latency](https://octyn.co/glossary/p95-latency)

## Other topics

- [Build or buy](https://octyn.co/blog/topic/custom-ai-systems)
- [Client context](https://octyn.co/blog/topic/client-context)
- [Automation that holds up](https://octyn.co/blog/topic/reliable-automation)
- [Founder operations](https://octyn.co/blog/topic/founder-operations)
- [Growth systems](https://octyn.co/blog/topic/growth-systems)
- [Operations systems](https://octyn.co/blog/topic/operations-systems)

Canonical: https://octyn.co/blog/topic/product-systems