Startup revenue is more fragile than it looks: customers leave as fast as they arrive

Research cited by TechCrunch recorded an uncomfortable shift for startups: annual recurring revenue (ARR) has become an unreliable metric. Customers — especially of AI services — sign up in a day and leave in a day, for a competitor or for an in-house build. "Recurring" revenue increasingly behaves like one-off revenue.

The cause lies in the nature of the new products: AI tools are easy to try and easy to swap, switching costs are near zero, and the differences between services melt with every base-model release. The winner is not whoever collects subscriptions first, but whoever is inconvenient to leave — thanks to data, integrations and embedding in processes.

For business in Kazakhstan the news cuts both ways. If you buy AI services — check the vendor's resilience and your ability to take your data out: a shiny startup may not survive the next year. If you sell — build value on depth, not novelty: integration into a client's accounting and processes retains better than any discount.