Ahead of Nordic APIs Summit 2026, we check in with Kin Lane (The API Evangelist) to talk AI, APIs, and the growing risk of data becoming trapped in walled gardens.

There’s a ton of hype around AI in the API space, and beyond, right now.

Many are talking about agentic consumption, Model Context Protocol (MCP), and deterministic workflow standards like Arazzo. But beneath that hype there sits an uncomfortable question: who controls the data, infrastructure, and APIs that AI increasingly relies on?

Whether this level of excitement is sustainable, and indeed whether AI’s continued widespread use is actually economically viable, remains to be seen. To kick off this conversation, we started by asking Lane for his take on its rapid rise within the industry.

“For me, this is just another cycle that I’ve already seen several of. Definitely the largest and most hyped cycle that has come along, but I remember AI circa 2012/2013 — Wolfram Alpha and IBM Watson — and everyone thought it was the future. But, no doubt, it wasn’t as big back then as it is now, and there wasn’t as much money in it.”

He adds that for API practitioners, AI doesn’t change things as much as some people would have us believe. “You still have to have your house in order. Agents aren’t going to magically understand your processes, just like a human doesn’t. Microsoft, Google, AWS, they all have different approaches, different ways of bundling tools, and giving access to infrastructure, but a lot of it boils down to API design problems.”

And, as ever (just ask Notorious B.I.G), more money rarely means fewer problems.

From Database to API to Agent

One of the biggest shifts we’re seeing in the space right now relates to patterns of consumption. We’ve written extensively about the importance of making documentation — and APIs themselves — machine-readable, since it’s increasingly agents who are handling discovery, parsing, and even workflow creation.

“The power within the enterprise was always around the database,” Lane points out, “which was very centrally controlled. Internally, that has somewhat shifted from the web to API gateways — trying to control the flow of data as it goes out of the enterprise.”

As we’ve seen from the wave of businesses shuttering public APIs, the rise of agentic AI consumption has led some providers to become increasingly cautious about the data that they expose. This is, Lane suggests, straight out of the venture capital playbook:

“Everything starts off very kumbaya, with open APIs at launch. Then, by the time you get to a Series D [funding round], everything is locked down. And SaaS data lock-in only compounds that problem.” He provides the example of LinkedIn, which has a heavily gated invite-only API, “so the only way to extract your own data is using a local proxy.”

Although small language models, open source, and “bring your own” LLMs may look like a suitable pressure valve, they aren’t necessarily the answer either — if you can’t get your hands on the data you need, you can’t deploy the right tools to achieve your goals.

Bigger Gardens with Higher Walls

Lane jokes that “the best way to get your much-needed API project approved right now is to disguise it as an AI project so you can get the budget.” The flip side of that, of course, is that if the AI bubble bursts (or becomes prohibitively expensive), API developers risk having their projects shuttered or needing to significantly rework them.

Budget is undoubtedly an important factor in getting projects off the ground right now, but growing anxiety around security and access is another. The potential damage of risks like Broken Object-Level Authentication — see the exfiltration of millions of Optus customer records back in 2022 — rises with automation and agentic consumption.

We’re now living in a world where a faulty agent, or a bad actor using agents to take advantage of exploits, can make thousands of requests per second or disregard the API contract in other ways. Perhaps it shouldn’t be surprising that API providers and consumers are making development decisions driven by fear, or at least caution.

Enterprises, Lane states, increasingly live in curated ecosystems. “These walled gardens, long owned by Oracle and the like, are now increasingly owned by Microsoft, Amazon, and Google,” he says. “If they’re a Google app shop, their next solution will come from Google Cloud. If they’re a Microsoft shop, it’s going to come from Office 365, etc.”

The result of all this may be that API developers steer clear of (or, indeed, may be prohibited from using) a quirky third-party API that offers an ideal solution in favor of a middling solution from a trusted infrastructure provider. It’s an intensified version of vendor lock-in that can have wider implications on costs and functionality, not to mention the lack of innovation that so often results from widespread homogenization.

Boom, Bubble, or Bust?

To keep tighter control over their own data, many companies that are building agentic tools are really just wrappers for Claude or ChatGPT, or closely mimic those services. This could have some serious implications on pricing models moving forward.

“When AWS EC2 came out, everyone thought everything would be cheap,” says Lane. “People would say ‘our server costs are going to be near zero!’ And now your cloud bill is your biggest challenge. I think AI is just going to be that on steroids.” Indeed, some of the initial optimism around AI exhibited by large companies is already fading in the face of impending cost increases. And that’s not the only hurdle to clear here.

“AI and agentic have clearly found a use within coding circles. But as those coding projects come up against the realities of mainstream industries and businesses — healthcare, retail, the physical world — I think that’s going to slow the roll a little bit.”

Here Lane is alluding to everything from the regulatory and compliance issues that come up when using AI alongside sensitive data to the myriad obstacles faced when deploying it in the real world. (Because, while a robot falling over as it dances to Thriller is sort of funny, an autonomous vehicle making unsafe driving decisions is not.)

There may come a point at which some business leaders decide “this just isn’t worth it.”

For More on AI and APIs: Attend Nordic APIs Summit

When Kin joins us in October at Nordic APIs Summit 2026, he’ll talk more about the intersection of AI and APIs. Specifically, he’ll dig into the business side of that relationship, how API developers can take advantage of it, and how to avoid falling prey to hype cycles.

“The engineering side of operations and developers have wielded a lot of power over the last decade. I think some of what we’re seeing now is takeback; business people getting more control, because they have the budget right now and the money dictates things.” That will, or at least should, impact how developers position their upcoming projects.

In the meantime, we ask him where he sees the AI/API relationship heading next:

“What’s next is not just technical: it’s venture capital-fueled, it’s data center-fueled. So I think the hype curve is going to slow: LLMs will shrink back down to a certain size… but it’ll still be pretty big. Meanwhile, though they may not be seen as cool to talk about, APIs are 100 times more important today than they were five years ago.” Amen to that.