What happens when the customer of your API is no longer a person, but an AI agent?
For years, we've built software around people.
People open the product.
They read the UI.
They compare options.
They decide what to use.
That assumption has influenced how we design software, APIs, dashboards, documentation, and developer tools.
But AI agents are beginning to change how software gets used.
Agents are starting to choose the tools they need, call APIs, read documentation, and get work done with much less human involvement.
And that creates an interesting question for developers:
What happens when the customer of your API is an agent?
An agent doesn't care about a beautiful dashboard.
It cares about whether your API works.
It cares about whether the data is reliable.
It cares about whether it can understand your product quickly.
And it needs to be able to actually use the capabilities you've exposed.
This makes things like APIs, documentation, reliability, and data quality increasingly important.
Observability is an interesting example
Today, observability data is mostly something engineers look at.
An engineer sees an issue.
They investigate.
They look at the available data.
They identify what happened.
And they decide what to do next.
But consider a future where agents consume much more of that data.
Agents could continuously monitor systems.
They could find issues.
They could interpret what they're seeing.
And they could decide what to do next.
That means observability data may increasingly become something consumed by software, not just humans.
And that changes how we should think about the product itself.
The API becomes part of the user experience
If an agent is choosing and using your product, the API isn't just an implementation detail.
It becomes part of the experience.
The agent needs to know:
What does this product do?
What can I call?
How do I call it?
Can I trust the data?
Is the response predictable?
Can I understand the documentation quickly?
Can I actually complete the task using this product?
These questions lead to a broader shift in product thinking.
The old question was: “What do our users want?”
The new question might be: “What does an agent need in order to choose and use our product?”
That could become increasingly important across infrastructure, observability, APIs, and developer tools.
The next generation of products
Maybe the next generation of great products won't just be human-friendly.
They'll be agent-friendly.
Because the users of today may not be the users of tomorrow.
And if agents are increasingly choosing tools, consuming data, and taking action, products will need to work for them too.
Build something agents want.
Top comments (0)