Value accrual for AI apps

Over the past few years, we've seen apps evolve from gpt-wrappers (assistants and companions) to fully-fledged AI apps.

1. The first criticism of AI apps was that they were copycats and wouldn't grow.
   The utility they provide has enabled them to grow to millions of users extremely fast.

2. The second criticism of AI apps was about their capacity to generate revenue.

   AI apps like Perplexity, Cursor, Lovable, ElevenLabs and Midjourney now generate hundreds of millions in ARR.

3. Today, the main critique for AI apps is about their margin profile (they have fixed prices and variable costs).

Chris Paik articulated the fixed price, variable cost problem whilst A16z rebutted explaining that looking at margins this early in AI cycle shouldn't be the focus. App builders should instead focus on value to users, retention and expansion.

The retention problem

The perfect storm is brewing for AI apps: models are improving fast, research labs are becoming product companies (chatGPT, claude code), the barriers to building apps is lowering, the influx of new apps make attention increasingly scarce and the pace at which AI moves is preventing users from building new habits.

App builders face a balancing act. Grow userbase asap to capture as much of the market as possible, stay afloat financially and generate revenue, surf the reduction of inference costs etc.

generated by GPT

As consumers, we're all guilty of this: a new app comes out, we play with it to try it out it and never return.

How to solve the retention problem

The more context I've built in an app, the less likely I am to switch. Context is the moat and most AI startups are building teams internally to automate context engineering.

Instead of being used to lock-in users, context should be portable from app to app. This is what led us to build Mio — a context tool for AI apps. Mio derives preferences from existing datasources (calendar, email etc.) letting AI apps offer personalisation from the very first interaction. Signup to the waitlist here.