Ahead of Platform Summit 2025, keynote speaker Kristen Womack shares the current state of the agentic coding revolution and its impact on APIs.

To say that AI has affected software development in the last year or so would be an understatement. The rise of agentic AI has been incredibly transformative and downright impossible to ignore for all involved in the software development sphere.

Kristen Womack will give a talk on the rise of proactive agents at Platform Summit 2025

One person who has a keen perspective on both the evolution of AI and its impact on the API space is Kristen Womack, a principal product manager at Microsoft building the Azure Developer CLI.

Kristen has previously presented at the Platform Summit, having explored how to create an awesome developer experience (DX) in 2023 and, in 2024, how to connect AI APIs to enhance user-centric applications.

According to Womack, the evolution of DX and AI is “a world apart” from when we met just last year. New agentic capabilities have fundamentally changed daily developer workflows, and the Model Context Protocol (MCP) has opened a plethora of opportunities for syncing agents with external resources. That said, the fundamental API best practices we’ve been preaching for years still ring true, she added.

Read the full Q&A below with Womack, and consider making it out to this year’s Platform Summit and the API Security UnConference in Stockholm for a deeper understanding and the ability to connect with some great minds currently exploring these areas.

Interview with Kristen Womack

Both developer experience and AI are fast-moving areas. How would you say AI has evolved since our last Summit?

Last year’s Summit seems a world apart from where we are today. We were describing and discussing the unanswered question of how AI models/chat use and access data and services via APIs. Particularly in Zdenek Nemec‘s talks over the past two years, he did a brilliant job describing the problem.

Today, protocols like MCP are bridging these gaps between AI and APIs by enabling language models to operate with agentic capabilities. Models enabled with agent protocols can process structured environmental data and context, make informed decisions based on this information, take action through function calls and API integrations, and adapt through feedback loops. These protocols provide a framework for agent-environment interactions. This is, by far, the biggest change I’ve seen in the last year related to AI, prompt engineering, and APIs.

Agentic AI is a hot topic in tech in 2025. Are you seeing actionable results using AI agents that go beyond the hype?

Yes, agentic AI has fundamentally accelerated the pace of innovation in software development — specifically, how we build software applications. And I don’t just mean vibe coding for prototypes. The inner developer loop is getting tighter as we resolve problems much faster. For example, an agent given the right context and tools can find problems across the repository faster than a text-only search or using grep, and in more complete ways. And with MCP servers, we can pull context into our workspace so we don’t have to leave our code editor.

Even in my own role as a product manager, almost everything I do has changed. For example, the workflow of how I collect the contents for a release blog, how I write a product spec, how I prototype, how I analyze customer feedback and market data — all of this has changed. It’s a higher abstraction of what we can build because our time is freeing up from shortening the loop in debugging, gathering research, and even finding research in the depths of places I didn’t know existed. This is just the short list. I’m amazed listening to stories of what so many people are doing.

If I could offer an analogy, it’s akin to spending less time churning butter and more time creating new ways to cook with the butter.

There are case studies on developer productivity and medical breakthroughs, but there are also just the very real, everyday things I’m seeing that go beyond the hype. Recently, Liz Fong-Jones shared a cool story on Bluesky about customer turnaround time with her creativity. I’m seeing things like this every day, and it’s amazing.

What does the use of APIs by AI agents look like lately? What sort of capabilities is this helping them achieve?

Integrating LLMs into software applications relies on data, such as from the internet or using a retrieval augmented generation (RAG) model against some stored data. This is the grounding data it reasons about in its actions or chat. And since there have been more than a dozen protocols that have emerged since August 2024, with MCP quickly becoming the front-runner after its launch in November 2024, now models and APIs can be friends with less heavy work for the developer. MCP and other protocols allow more dynamic context management and bidirectional agent communication. Through standardized function calling interfaces, models can now run operations via the MCP server, with proper error handling, state management, authentication, and more.

What are the end benefits to the business if we can effectively arm AI agents with APIs?

There are a few ways to look at this. From the perspective of API producers, those with APIs ready to extend their business functionality will thrive in the agent era. It’s similar to the mobile era paradigm — if a company wasn’t interoperable for mobile, their websites didn’t work or render properly on phones. The same principle applies today with AI agents.

From the aspect of business operations, AI agents can aggregate and analyze data from multiple APIs simultaneously, providing real-time insights for business decisions. This enables faster decision-making with richer information context. For this to work effectively, businesses need to work with tools that have well-designed, accessible APIs.

Businesses can also extend these capabilities across departments. For instance, with observability, AI agents can detect patterns and anomalies across different systems by correlating API data streams. While this already proves valuable in fraud detection and security, the business benefits extend much further.

The key transformative aspect is that AI agents can understand business context and orchestrate APIs accordingly, rather than just executing predefined sequences. This leads to more intelligent, adaptive, and valuable business processes.

What are some gotchas in the way of seamless agent-to-API connections, and how can they be overcome?

One of the biggest things to overlook is that agents thrive on context, and it’s one of the easiest things we can do to help the agent-to-API connection. For example, one thing API providers might do is an audit of their APIs specifically for agents. The audit might include reviewing the metadata for each endpoint about the API semantics, checking to see if business rules exist in the API specification (not just the technical aspects), reviewing the API descriptions in the OpenAPI specification, and checking the ability to maintain context across API call chains. Basically, there is the programmable interface, the human interface, and the agent interface. They all require slightly different tuning to describe how to use your API.

There are numerous other “gotchas” like data quality issues, not checking the agent’s work, giving too much trust or access, specifying workflows, and layered orchestration. We have to remember that all the things that mattered before (such as auth, security, monitoring and observability, and performance) all still matter just as much as before we were working with agents.

What do you like most about Stockholm?

There is so much to love about Stockholm. Bikes, nature, food, museums. The people are welcoming and kind. When I’m in Stockholm, I find there are endless running routes and there’s always a water’s edge to follow, given all the connected islands. The transportation by train and bike is amazing. Stockholm is such a lovely city, hard to pick one thing I like most.

We’re thrilled to feature you as a speaker back in Stockholm this year! What keeps you coming back to Nordic APIs in particular?

The speakers, the themes, the location, and the incredible conference organizers and staff that curate it all. The Summit attracts a mix of API practitioners: Ph.D researchers, start-up founders, technologists from large companies, developers, architects, product managers — the range is vast.

Are there any particular themes, hallway discussion topics, or sessions you have your eye on for this year?

I’m thrilled to see my good friend Adam DuVander close out the Summit. He’s an amazing storyteller and is bound to leave us with much to think about as we travel home and head into 2026. I’m also particularly interested in the conversations and talks on security. With rapid advancement in technical tools it gives rise to new security considerations and risks. How we are mitigating those and learning from the industry is important to me. And of course, I’m looking forward to returning to Stockholm.