The Age of Agents – Product Dev for AI

Way back in the “nineteen hundreds” (as my kids call it), product development was relatively simple. We lived in an offline, disconnected world, where great products were largely defined by their feature set rather than experience. If your product had the right features, and it worked, that was often enough to win. Customer Experience (CX) mattered, but the bar was low.

Then, the world of product development changed.

“Software Is Eating the World”

In the 2000s, connectivity exploded. The rise of the internet, then modern web applications rewrote the rules of product development. What was once considered “good enough” CX became outdated overnight.

By 2011, Marc Andreessen wrote Software is Eating the World, predicting that every company would soon become a software company. And he was right — entire industries were transformed by frictionless, always-on, real-time experiences. That then got accelerated (and shrunk to pocket-size) with the rise and domination of what we used to call “smartphones”. Now, they’re just phones, and everyone has one.

The DX and EX Revolution

Winning in this new world wasn’t just about CX anymore. Companies had to build and ship high quality software faster. This led to a focus on:

Developer Experience (DX): Making it easy, fast, and safe for developers to build and deploy software. You also needed a great culture to attract talent and hold on to it.
Employee Experience (EX): Equipping internal teams with seamless tools to operate at scale.

The big cloud providers became the backbone of this transformation. Instead of wrestling with supporting on-prem infrastructure, developers could spin up scalable services instantly. Companies like Auth0 made authentication plug-and-play, while Stripe turned payments into a simple API call. The easier you made things for developers, the faster they could build, and the more dominant your platform became.

For a decade, the secret to better CX was better DX and EX. Companies that understood this, like Amazon, Stripe, Twilio, and others – they became category-defining winners.

Enter the Next Wave: Agent Experience (AX)

Now, we’re on the verge of another shift — one that will leave many companies gasping for air again. The world is moving beyond just EX and DX.

We’re entering an era where Agent Experience (AX) is the next battleground. This is the rise of AI as a New User/Persona that we need to design for.

Until now, software was built for two groups: developers and end users. But AI agents are fast becoming a third category of user. Products that optimise for AI-driven interactions will thrive. Those that don’t? They’ll struggle to remain relevant.

A few early signs of this shift:

Supabase provides backend “database as a service”, with related authentication and other services on top of a solid open-source PostgreSQL DB. Their DX is fabulous, so they were already doing amazing. More recently though, Supabase has emerged as the dominant player with a number of AI prototyping and development tools, like Lovable.dev and Bolt.new. Seeing the AI tools interact with the Supabase service and spin up a solution is worth trying.
Netlify and Vercel provide hosting for web apps, which again started with super nice DX. But now, they are both emerging as dominant in adoption by agents, because they’re easy for the Gen-AI tools to interact with. Vercel has it’s own GenAI dev tooling, v0.dev, which of course uses Vercel as the default backend.

These are just a few early examples, but I’m sure we’re going to see many more. These companies aren’t just easy for developers to use — they are also easy for AI-powered tools to plug into and play with. And that’s a key lesson for the future.

Where this goes, who knows. But as an example, I’m willing to bet that the first bank to provide not just great EX and DX, but also great AX will see it leap ahead of the competition. Same with HR systems, Legal systems, and all of the other things that are considered the basic building blocks of starting a company. I should be able to not just have Generative AI tooling assist with building a website or apps that solve real problems for my customers, I should be able to spin up the whole company with a few prompts. Company formation, bank account setup, writing contracts, onboarding employees, all of it.

And that’s just the B2B space. I can only imagine what the B2C space might look like. It’s worth pondering though. It’s coming, whether we like it or not.

How to Build for the Era of AX

If DX was about making it easy for engineers to build, and EX was about making life easier for operational employees, then AX is about making it easy for AI agents to interact directly with your product.

So, what does that actually mean? How should you be thinking about Product Development in this new era?

1. Expose clean, structured data

AI agents don’t “see” buttons and screens with words on them — they consume data. If your product buries key information behind complex UIs, lacks API access, or provides inconsistent outputs, agents won’t be able to use it effectively.

Ensure data is well-structured, machine-readable, and accessible via APIs.
Embrace semantic formats (e.g., GraphQL, structured JSON) over raw dumps of poorly structured text.
Offer clear, well-documented API endpoints to make agent-interactions seamless.

2. Make actions executable, not just visible

Today’s AI tools don’t just retrieve data—they take action. A product that only presents information without allowing agents to act on it programmatically is going to feel outdated very quickly in this new era.

Turn UI-driven workflows into API-driven workflows. It turns out that headless-mode wasn’t just for delivering content and interactions to multiple end-points, it’s also going to be key for AX.
Make sure key actions (e.g., transactions, configurations, automations) can be triggered via API, not just via a human clicking buttons.

3. Design for interoperability

AX isn’t just about your product—it’s about how well your product plays with other tools in an AI-driven workflow. If your platform or service is behind a big wall that the agents can’t access, it won’t be the first choice for AI-driven adoption.

Make integrations dead simple.
Offer native AI connectors (e.g., OpenAI Plugins, LangChain integrations).
Provide webhooks and event-driven architectures to let agents respond to changes and errors in real-time.

4. Optimise for AI-driven decision-making

Many traditional systems rely on human judgment and intervention, but AI agents make decisions differently. If your product is built for manual review instead of automated reasoning, it may struggle in an AI-first world.

Expose confidence scores and reasoning in API responses to help AI agents make informed choices.
Build in fallback mechanisms—let AI agents request clarifications rather than failing outright.
Reduce unnecessary complexity in decision-making processes and workflows to make automation easier.

The bottom line: AX is already a competitive advantage

We’ve seen this play out before—companies that optimized for DX and EX dominated the last decade. The same will be true for AX. Products that embrace AI as a first-class user—by making their data, actions, and workflows AI-friendly—will have a massive edge.

This is the key question to ask yourself: If an AI agent wanted to use our product today, could it?

If not, now is the time to start making that shift — because AX isn’t the future. It’s already here.