How I think about the Agentic Web

Welcome to my first real post on this blog 😄🎉! Less technical, more general thoughts.

The way I like to think about the Agentic Web is:

The Agentic Web is a web where agents complete workflows on behalf of humans.

First of all - agentic systems are not new 🤷‍♂️.

In computer science an agent is an autonomous entity that observes its environment, processes the collected information, takes actions to achieve a specific goal, and then continues.

What changed in recent years is that we unlocked new capabilities: we gave agents access to LLMs.

When it comes to agents interacting with the web, I like to separate the experience into two distinct use cases. This difference is important for knowing exactly what to optimize, and how to do it:

1️⃣ An agent is searching for a website (for example, using Google Search). In this case I highly recommend guidance published by Google Search team.

2️⃣ An agent is working with a currently active website. In this case we need to understand how agents can perceive our websites - links in Around the web.

Whenever I chat with people in the SEO world who are exploring agentic interactions, they tell me the same thing: they don’t just care about traffic; they care about the full funnel. They’re looking at the entire journey, from how someone finds a website to that final click - be it a purchase, a subscription, or a contact form submission.

In the context of the Agentic Web, the bottom line is simple: if your website isn't optimized for an agent to actually use it, your users lose out on automation, and you lose out on conversions.

Evolution of the web 🧬

To imagine the future of the web, I first look at how much things have changed in the last four years.

Let's start with coding agents - I first used them in 2022. Safe to say I was not very convinced back then 😁 - even within a single function, the coding agent struggled to perform simple changes. Today, I offload most of my coding to agents 🤖, while still manually reviewing the code and fixing it as needed. It's a big change in my day-to-day work.

During one of my recent talks I asked the audience:

How many of you used any form of AI in your work in the last week?

... Roughly 90% of the room raised their hands. 💡🤔

My take is that increasing AI usage and quality will drive higher user expectations for the web going forward. If modern agents can write a complex script or analyze a document in seconds, why should a user have to spend twenty minutes filtering flights, dates, and hotels on a travel site just to book a simple trip?

On one of the local meetups I've heard an opinion that in the future we will not need websites - agents will just go and talk to MCP Servers and/or each other. I think I've heard something similar before 🤔 - I remember when online stores became popular in Poland in the early 2000s, and people were saying it was so over for physical stores, that nobody would want to shop in person anymore. Guess what? I'm in a physical store at least once a week 🤷‍♂️.

Some users will continue to browse the web as they always have, and they might use agents embedded on the websites, others will delegate that task to co-browsing agents, or OpenClaw-style agents.

To support this we don't need a second, separate web.

The Agentic Web is an evolution of the web we already know, not a parallel reality. The main thing changing is the interface (Hot take? 🌶️😁). And it will be up to the businesses to understand the expectations of their users and provide right channels of interaction - whether that means supporting co-browsing, full agent delegation, or a pure on-site experience.

More soon. ✨