Every time I hear ”AI agent,” I think “secret agent.”

The allusion is obvious, but I'll blame it on the varnished set of Ian Fleming novels sitting on my office shelf, featuring a debonair Lotharian spy, James Bond. You know him: 007. Casino Royale. The infamous “license to kill.” Martinis shaken, not stirred.

Bond was skilled, suave, and seductive. Useful attributes in his high-stakes line of business. But when I think about this new AI nomenclature, another secret agent comes to mind:

Maxwell Smart.

If you’re not familiar, he was the titular character from the Mel Brooks TV series Get Smart. As a kid, I discovered this ‘60s gem in syndication. It concerned an aloof spy in an equally aloof world, satirizing the inept machinations of Cold War espionage.

Max worked for an agency called CONTROL – which, ironically, was never in control of anything. Not as “smart” as his name implied, he was always over his head and prone to mishaps, somehow fumbling his way to thwarting KAOS (an evil crime syndicate).

One of my favorite characters from the program was Hymie, Max’s humanoid robot sidekick. Hymie was an advanced artificial intelligence that could analyze physics equations while battling gorillas and dodging bullets. Despite that superior brainpower, he wasn’t exactly “reliable.”

In a lot of ways, Hymie mirrors the challenges we face with AI. On the surface, he’s the perfect worker, designed to help his human counterparts execute the mission. At the same time, he’s operating in a human world – one that’s not built for him to function optimally.

Like Hymie, AI agents are working on behalf of their human counterparts, adapting autonomously to environmental changes using sophisticated reasoning. What separates them from other forms of Gen AI is that they’re endowed with agency – allowing them to make decisions without any human intervention.

For years, websites and apps have been our windows into digital experiences. But they’re highly manual, requiring a dedicated level of human engagement. Now, AI agents can execute for us in real time, eliminating repetitive tasks and giving us back the most valuable currency of all: time.

Given that potential, it’s not a stretch to predict that AI agents could supplant entire swaths of SaaS apps that would otherwise require our attention. And with their uninhibited work ethic – no weekends or coffee breaks – they could deliver breathtaking levels of automation.

AI agents are already being employed to automate all kinds of functions in software. But the horizon of use cases stokes the imagination, from adjusting your smart home’s thermostat based on your preferences, to diagnosing health-related issues and even recommending treatments. One day soon, they might be managing your investments. That said, I’m not ready to trust Hymie with all my finances… not yet, anyway.

As we explore new ways for intelligent agents to book our next vacation, how will they effectively navigate our human landscape? Websites are, after all, built for people – not robots. Are we designing the experience infrastructure with agents in mind?

That’s where I started my conversation with Mathias Biilmann, who has pioneered the new concept of Agent Experience, or AX. As the CEO and co-founder of Netlify, Matt has been shaping the rise of AI agents as they autonomously execute more and more of our daily functions, specifically in the development channel – where Netlify plays.

As he sees it, Agent Experience is more than a technical evolutionary step. It’s an entirely new way to deliver business value and, in his words, “build a better web.” Right out of the gate, I asked him where the idea surfaced.

“The term ‘AX’ kind of dawned on me at the start of the year,” he explained. “We’ve been doing a lot of work on the agent experience of Netlify, but without having a sort of coherent internal term around it.”

In a recent article that Matt published on his blog, he expands on the genesis of AX and how it emerged from the progression of UX (“User Experience”) and later DX (“Developer Experience”). These concepts are well-established and have shaped the foundation for how we build products.

But to Matt’s point, as we enter this new era where agents interact with those products autonomously, we need to craft experiences specifically for AI agents, so they’re not hamstrung by navigating human interfaces to make requests or execute tasks.

“I think that humans will collaborate with a ton of agents that will do all sorts of things for them – both when they are directly asked to do it, and in the background,” Matt said. “And as builders, will have to really take a holistic view of how agents are able to do this.”

There’s a lot of chaos in the realm of AI, and the rise of agents isn’t quelling any of it. They're showing up everywhere, and it’s only getting louder. Is Agent Experience the control we need to help shape a more promising future?

Building a platform for DX – and now, AX

Matt has always leaned into “vision questing.” He and Netlify’s co-found Chris Bach are celebrated luminaries in the digital experience and development fields and have a long history of collaborative innovation. The pair met during high school in Denmark, and according to lore, Matt began as an aspiring musician and music journalist before pivoting to engineering in 2006. These days, he might be spending more time on his computer's keyboard, but he still has one of the “musical variety” in his office.

A consummate entrepreneur, Matt continues to weave his other startups into a networked web of capabilities, like Webpop – a hosted CMS based on a strict separation between content and design. In fact, his previous startup BitBalloon evolved into the Netlify Drop service in 2018.

Among his other claims to fame, Matt famously coined the term “Jamstack” in 2015, which became the de facto locution for progressive, ultra-fast web architectures that leverage static site generators. The acronym (which stands for JavaScript, API, and Markup) continues to be a zeitgeisty staple of composable solutions.

As far as Netlify is concerned, it has secured its place in the pantheon of web development. The platform powers developers to build dynamic, highly performant sites, e-commerce stores, and web applications with an extensive ecosystem of technologies, services, and APIs.

With its pioneering Jamstack heritage, Netlify brings together modern web frameworks powered by advanced serverless functions and edge computing. According to the company’s website, over 35 million websites are deployed on Netlify, and over five million developers use the platform. And that’s been at the heart of everything: understanding the developer experience.

“We want developers to know Netlify and have it in their toolbelt,” Matt said during a recent interview at Infobip Shift Miami. It's working – and the market agrees. In addition to scaling its revenue, the company raised $53 million in its Series C funding in 2020, which is helping advance the adoption of its web development platform.

Pioneering the path to AI

Netlify’s journey to generative AI started in the same place as its peers, with use cases that leveraged OpenAI. But Matt quickly understood how important it was to make the deployment process easier for developers.

“Some of our early experiences were with building our own ChatGPT integrations on top of some of the APIs we had,” Matt recalled. “We had built originally for resellers, and then white labeled for users and so on. We wanted to provide this frictionless ability to tell ChatGPT to just deploy this stuff you made on Netlify, so you’re not sent through a bunch of hoops, logging in back and forth. That ruins the experience of it, right?”

And this is where this agent “awakening” is making an impact. Now, more than 1,000 sites are being created on Netlify directly from ChatGPT every single day. As Matt noted, this happened because they focused on what agents would need from Netlify, identifying new interaction flows and how their API would need to be adapted and optimized for LLMs.

“We obviously have customers building on Netlify in different ways – chatbots and stuff like that. We also have AI-powered features within Netlify, so we had this notion: How do we make Netlify the best platform for AIs to deploy on?”

“A lot came from where we sit in the space. We obviously have customers building on Netlify in different ways – chatbots and stuff like that,” Matt said. “We also have AI-powered features within Netlify, so we had this notion: How do we make Netlify the best platform for AIs to deploy on?”

Now, with the mainstream adoption of agentic AI, Netlify has positioned itself as a go-to platform for building what’s next and enabling developers to lead the charge. As I detailed in a recent article, the number of new Git repos leveraging Python, PyTorch, and ollama exploded in 2024, and these AI-centric frameworks reflect a huge shift toward AI-specific projects.

Some of these projects might dabble in the “shallow” AI features we’ve seen over the last two years (think basic ChatGPT connectors for generating content). But many are pushing the boundaries in exciting ways, and Netlify is providing the foundation for building and shipping those experiences – particularly with agents.

Will AI agents “get smarter” with AX?

Matt’s vision for Agent Experience is fresh. But if the term seems familiar, there might be a reason for that: CSM technologies like Zendesk and ServiceNow have long championed the foundational idea of the CX Agent Experience, defining it as the satisfaction employees feel when interacting with technology, processes, support systems, and other facets of their work.

That said, it’s a perfect mirror for AI agents – which, in many ways, behave like employees (or perhaps even interns). The difference? AX concentrates on improving how AI agents engage with their environment.

Like many things in tech, realizing the value of AX came with trials and tribulations. As Matt described, he saw a micro example of where developer experience was still creating friction in the flow and forced some rethinking. It emerged while using Bolt.new – an AI-powered tool for prompting, running, and deploying full-stack apps via agents.

Bolt enables users to prompt, run, edit, and deploy full-stack apps via agents. Source: Bolt.new website

“Right after I launched my blog, someone asked me if I could add an email subscription form so they could hear about new articles. So I went into Bold and said, ‘Can you add an email subscription form?’ and it added a Netlify form perfectly. But when I deployed, it didn't actually work. It turns out that the normal flow for Netlify forms is based on our notion of developer experience, so it was centered around how a developer thinks and approaches our platform.”

To Matt’s point, if a developer wants to build a form with Netlify, they typically refer to documentation. For a human reader, the first prescribed step is to “enable detection of forms” during the processing of the deployment, which the agent didn’t account for.

“There's a clear difference between approaching it through the lens of DX or the lens of AX, and where we end up having to build something slightly different,” he continued. “And the more you go down this path, the more you start seeing where agents have different strengths and weaknesses and ways of knowing the information and not knowing how to act. I started sharing this concept of AX internally and telling my whole team in the same way we did with our innovation in the frontend space by focusing on DX. The next era is really focusing on AX.”

Numerous companies are already embracing AX as a discipline, demonstrating how sticky the concept is becoming. Matt cited examples in his blog of tools like Clerk, an authentication platform that’s putting AX thinking into action, making it simpler for agents from outside sources to build apps or sign into apps using Clerk as its authentication layer.

The broader potential is playing out in some fascinating and unexpected ways. More in the domain of sci-fi, a recent video showcased two conversational AI agents chatting over the phone about a hotel booking – and then switching to a more efficient data-over-sound ggwave protocol from ElevenLabs called GibberLink.

It's a bit like R2-D2 on steroids:

Ideas like GibberLink (you can check out the Git repo here) are going viral because they give us a glimpse of the future while making the hair on our necks stand up. Still nascent, this technology is already demonstrating how Agent Experience can transform the productivity and efficiency of apps. It also raises serious questions about privacy, transparency, and where human oversight fits in.

As agents roam into the wild, AX has become a company-wide “obsession” at Netlify. This includes divining the shortest path for an agent to go from user input to URL in production, ensuring Netlify’s APIs are designed optimally for LLMs to build with, and making the collaboration between developer and agent more efficient.

They’re already mapping this out in a tangible way. Sean Roberts, one of Netlify’s distinguished software engineers, recently penned an article that focuses on the challenges of adapting the modern web's “human” orientation to an AI agent world.

As he outlines, agents have different needs and requirements for achieving a successful AX. Without applying the right practices, agents will struggle in a less optimal environment – so enhancing the agent interaction is essential. He also delves deeper into the distinctions between APIs and the “agent web,” how content should be optimally exposed, and the role of caching in the mix.

A diagram of a brain

AI-generated content may be incorrect.

Source: agentexperience.ax website

One thing I love about this article is how Sean diagrams the typical human and browser processing and provides a clear pathway for realizing an optimized AX – culminating in a place where “agents and the web can work together to feel like magic to customers.” These simple workflow sketches help visualize the shift in thinking.

If you’re looking to stay on top of information like this, Netlify just launched a new website dedicated to AX, a “center of excellence” that will focus on learning, researching, and building the future of Agent Experience. It’s early days, so the content is light. But I’d still recommend bookmarking it.

This is where agents – and people – can get smarter with AX.

Unlocking agents with open, composable thinking

According to Matt, there's a familiar fork in the roadmap, one where platforms might choose to close off their innovation in proprietary systems or open it up to advance innovation. Some companies are already designing their agents to function within a narrow vertical spectrum, tightly integrating the use to their own software.

Matt sees that as a key limitation and advocates for an “open agent world,” which evokes the tenets of composability and the open web – an ecosystem where users have choice and flexibility. In this topology, a tool's AX posture will be a critical consideration.

“You’re almost at ‘Composability 2.0,’ where a lot of tools that are composed together will also depend on the agent experience of those tools,” he said. “It’s a software stack emerging for agent-built software.”

We also discussed how this is manifesting at the UI layer in the wrong ways. As Matt pointed out, Google has been feverishly adding buttons for Gemini AI across their products, with no clear path for users to bring outside agents for content creation, visual design, data manipulation, and presentation development.

Not very open. And, sadly, predictable.

That’s where Matt is continuing to evangelize for change – and it’s bigger than just these proprietary clouds. He believes the web needs to evolve to the next generation of experiences, and we need to invest in the force multiplication of AI as a community, not as isolated vendors.

“I do think there's a risk that a lot of the experiences we have just get pulled into proprietary platforms if we don't start innovating more in the experience level of the web,” he explained. “For the last couple of years, there's been a tendency to spend a lot on tools and technology that lets us build the same thing, but maybe a little faster. The first AI wave was about making the cost of building lower and easier. But I think, as a web community, we need to explore the experiences we can build on the web that we simply couldn't have before this availability of AI.”

It’s an AI agent world. What about the humans?

Good question. Are we being retired? Will agents just take over?

AI has always been a panic button for the naysayers. Some say it’s robbing us of our creativity – and the rise of agents has amplified the dystopian narrative of a world where people are sidelined by intelligent robots. Matt disagrees.

“I think we all understand those concerns, but at the same time my view on agent experience is very optimistic,” he said, smiling. “I don't really believe we’ll end up in a world where everything is just agents because, in the end, we build all of this to do stuff for humans, and humans crave shared experiences and places to go to online to view things.”

We may not be giving up the web entirely to agents, but we’re certainly going to share it with them. It’s happening already. To that end, Matt sees an ecosystem where we work together in efficient, automated ways to unlock more creativity than ever. To get there, developers will need to embrace new ways of thinking and architecting.

Displacement is another concern. Copilots are one thing, but agent-led automation – and the potential obsolescence of SaaS apps – could result in an even more precipitous decline. According to ADP, there’s been a huge drop in developers entering the field since 2019. Matt believes agentic AI will spur the opposite.

ADP research chart showing the decline in software developer employment

Source: ADP Research

“I think we’ll probably get 100 million more developers over the next five years,” he boldly predicted. “But as always, what you're developing – and what are developers – will be different. When I launched Netlify, the normal developer was building mainly backend with a little hint of some frontend thrown in. And it turned out we probably added 10 million developers at that pure frontend layer.”

Eliminating the cost barriers will be part of this transformation opportunity. As agentic AI blossoms, Matt sees a proliferating market for builders as they shift to more specialized development.

“It will massively drive down the cost of custom developing something,” he reflected. “When the cost of something goes down, demand always goes up. And I think in this case, the demand for custom building interactive experiences and apps will dramatically increase – and companies should be ready for that transition from generalized to specialized in the SaaS landscape.”

This could have a profound impact on companies buying generic SaaS applications. If the cost of building specific tools decreases, they could directly engineer hundreds of internal applications to solve their problems. In this world, developers could thrive.

Why it matters

One of the things I love about Get Smart is the show’s intro, in which Max (played by the comical everyman Don Adams) arrives at CONTROL’s headquarters and traverses its nonsensical security system:

Sometimes, our tech world feels exactly like this: Layers upon layers of conflated code, systems, and processes that ultimately prove ineffective. AI has been problematic in a similar way, from hallucinations to exorbitant cloud bills to shallow features that deliver limited value. We need to avoid these hallways of doom by thinking, talking, and collaborating to unlock the real value.

Matt’s focus on Agent Experience is timely – and another example of his forward-thinking mentality. The agent landscape is exploding, and as he pointed out, there are implicit challenges on many fronts. This includes the development of closed or proprietary agents that limit the extensibility of their potential.

In the “here and now,” AI features have been wedged into CMSes, DXPs, and other platforms at a dizzying rate. Some of these capabilities have proven their value, working tirelessly to generate content or help simplify complex workflows. Others are less novel and more… novelty.

We might be past the era of low-utility features, but AI agents aren’t some “silver bullet.” If anything, they have the potential to exacerbate our problems. To avoid more chaos, developers need to embrace the right mindset and practices, and businesses need to ask serious questions about privacy and transparency. This is where Netlify’s leadership around AX can be a powerful platform for building community consensus and trust.

By establishing a voice for Agent Experience and creating a “town square” for collaboration, Matt is also positioning to tackle big questions like industry standardization. He’s also at the forefront, clarifying what it is and why it's essential. Even if the new “agentexperience.ax” website is currently a skeleton, we need a backbone to support what Agent Experience means to the world – and this is a good first step.

Plenty can go wrong. With everyone making agents at a breathtaking pace, we’re bound to see mishaps worthy of Maxwell Smart. But AX might just be the secret weapon for building a smarter, safer, and more efficient ecosystem. One that’s designed for agents, but serves us.

“In the end, we build all of this to do stuff for humans,” Matt said.

Sounds pretty smart to me.

If you’re interested in learning more, Netlify is hosting an exclusive webinar on February 27th about AI development and the importance of embracing an open, AI agent-friendly future.

Looking for guidance on Netlify? Talk to an expert.

Upcoming Events

CMS Summit 25

May 13-14, 2025 – Frankfurt, Germany

Don't miss the first European edition of our prestigious international conference dedicated to the global content management community! CMS Summit 25 will bring together top-notch speakers, our renowned learning format, and engaging social events. Hear from leading practitioners like Deutsche Bahn, Dr. Oetker, and more – and join customers, agencies, and CMS vendors as we discuss current trends and what's ahead for the content and digital experience fields. Connect and network at the only vendor-neutral, in-person conference focused on CMS. Space is limited for this exclusive event, so book your seats today.

CMS Connect 25

August 5-6, 2025 – Montreal, Canada

We are delighted to present the second annual summer edition of our signature global conference dedicated to the content management community! CMS Connect will be held again in beautiful Montreal, Canada, and feature a unique blend of masterclasses, insightful talks, interactive discussions, impactful learning sessions, and authentic networking opportunities. Join vendors, agencies, and customers from across our industry as we engage and collaborate around the future of content management – and hear from the top thought leaders at the only vendor-neutral, in-person conference exclusively focused on CMS. Space is limited for this event, so book your seats today.

News

It’s an AI Agent world. Is ‘Agent Experience’ the secret weapon for controlling the chaos?