Why AI in Your App? The 2026 Landscape
After this chapter, you can look at any app feature and decide — deliberately — whether AI belongs in it, whether it should run in the cloud or on the device, and what it will cost you in money, latency, and risk.
- What a large language model actually is from your app's point of view — and what it is not
- The handful of things AI is genuinely great at, and the ones that will burn you
- The three failure modes you sign up for: hallucination, prompt injection, non-determinism
- How to choose between a cloud model and one on the phone — cost, latency, privacy, quality
- When the right amount of AI in a feature is none
In February 2024, a grieving customer named Jake Moffatt asked Air Canada's website chatbot how bereavement fares worked. The bot told him, confidently, that he could book now and apply for the discount within 90 days. That was wrong. When the airline refused the refund, Moffatt took it to a tribunal — where Air Canada argued the chatbot was “a separate legal entity responsible for its own actions.” The tribunal disagreed, held the airline liable for what its bot said, and ordered it to pay.
That is the whole subject of this book in one story. An AI feature is not a clever widget you bolt on; it is a thing that speaks for you, to your users, and you own every word it gets wrong. Used well, a language model can do things no amount of hand-written code ever could. Used carelessly, it invents policies, leaks data, and freezes your UI — and you find out in a courtroom or an app-store review.
Before you write a single await model.generate(...), you need a clear-eyed map of what this technology is good for, what it costs, and where its sharp edges are. That map is this chapter.
That’s the free sample.
The rest of Building AI-Powered Flutter Apps keeps going — 186 pages of it, every line of code verified for June 2026.