Before Platform as a Service (PaaS) and serverless computing took off, developers spent more time managing infrastructure than building. Then AWS Lambda changed the game with serverless functions, and others followed suit with serverless containers. Developers were finally free to focus on building while offloading infrastructure concerns.
AI is now at the same crossroads. Developers don’t just need powerful models—they need AI that integrates without friction, scales effortlessly, is secure by default, and fades into the background. This is what unlocks rapid innovation.
What if AI worked like an API?
Plug AI into workflows with zero friction Validate and iterate AI prompts like API testing Scale intelligent automations without worrying about infrastructure
Postman is making this happen—turning AI workflows into intelligent agents with Postman Flows. The question isn’t if AI will follow the API-first path—it’s how fast we can get there.
What’s the biggest blocker to treating AI like APIs today? Drop your thoughts below.
One of the problems I’ve been facing is that websites that don’t have APIs. There are websites that have great information that I want to include in my workflows, but without an API, it’s near impossible to create an agentic workflow that doesn’t include a human in the loop.
Is this a hot take, @Rodric-Rabbah? Seems like this is what everybody actually wants
To answer your question though, I feel like the speed of LLMs is a big inhibitor for practical use. We’re in an era where APIs that respond slower than a second are trash (to put it lightly). But even the fastest LLMs aren’t reaching those speeds. If we can get to a point where we’re generating responses in a time that meets or exceeds user expectations, then we can start treating it like APIs. Otherwise there’s going to be friction.
I’m curious about how standardization and interoperability could make AI more like traditional APIs, especially in terms of integration and reliability. Could unified standards help address challenges like versioning, compliance, and scalability? @Rodric-Rabbah
Plug AI into workflows with zero friction
