When we started building Wizen, the problem we were trying to solve was relatively straightforward. How do we give people better access to an organisation’s trusted information through AI? Rather than asking a general-purpose AI and hoping it finds the right answer somewhere on the web, Wizen works from an organisation’s own governed content, giving people a conversational way to find information while giving the organisation greater control over the sources and guardrails behind the answer.
As we’ve started building more sophisticated experiences, we’ve found ourselves thinking about a much bigger shift. People don’t just want AI to answer questions. Increasingly, they want it to help them get things done. That changes the role AI plays in the customer experience and, we think, changes what it means for an organisation to be ready for AI.
AEO is only part of the answer
There’s a lot of attention on Answer Engine Optimisation and Generative Engine Optimisation right now, and for good reason. As more discovery moves into ChatGPT, Gemini, Claude and other AI experiences, organisations need to consider whether their content can be found, understood and accurately represented by these systems. Much of the current thinking focuses on making existing web content more legible to AI through better structure, clearer language, stronger authority and content that is easier for machines to interpret.
We think that work matters, but it largely assumes AI is another way of finding information. An AI searches for something, reads what it can find and returns an answer. The organisation’s job is to make sure the right information can be found and understood.
Agents introduce a different proposition. An agent might not be looking for a page to send someone to at all. It might be trying to understand their needs, explore different options, check current information and complete a task on their behalf. Once that starts happening, optimising the content itself is only part of the problem.
We’re making agents use websites designed for humans
Today, when an AI agent wants to interact with an organisation, it often has to use the same experience we designed for people. It opens a website, interprets what’s on the page, works out the navigation, follows links, clicks buttons, fills in forms and tries to understand what happens next.
The technology that makes this possible is impressive, but it’s also a fairly crude way for two machines to interact. We’re effectively asking an agent to behave like a human so it can navigate an experience that was never designed for it.
For simple tasks that might be fine. Finding an opening time or retrieving a straightforward piece of information doesn’t require much interpretation. It becomes more problematic when an agent needs to compare options, understand eligibility, check live conditions, navigate rules or make a decision using information spread across multiple systems.
This is something we’re confronting directly in one of our current Wizen projects. We’re working with a large repository of trusted information designed to help people plan, prepare and stay safe. Making that information conversationally accessible is useful, but it quickly becomes apparent that simply answering questions leaves a lot of potential on the table.
Moving from answers to tools
Imagine someone asking an AI agent to help them plan an activity based on where they are, who they’re travelling with, their level of experience, accessibility needs, current conditions and what they need to prepare.
The agent could attempt to piece that together by searching a website, opening pages and interpreting what it finds. Or the organisation could give it specific tools designed to answer those questions reliably. Those tools might allow it to find suitable options, check current conditions and closures, identify accessibility requirements, retrieve relevant safety information and return the results in a structured format that helps the person make a decision.
That distinction is becoming increasingly important to how we think about Wizen. Instead of only creating a trusted knowledge layer that AI can draw from, we’re exploring how organisations can also provide clearly defined capabilities that agents can use. The organisation determines what information is available, which tools can be called, what rules apply and where the boundaries sit. The agent doesn’t have to reverse engineer the experience because the organisation has deliberately designed a better way for it to interact.
Knowledge, tools and experience
We’re starting to think about the evolution of Wizen across three connected layers.
The first is knowledge. This is the foundation Wizen already provides, giving AI access to governed organisational content and trusted sources with clear guardrails around how that information can be used.
The second is tools. These give an agent specific capabilities rather than expecting it to work everything out from pages designed for humans. Depending on the organisation, that could mean searching a particular dataset, checking availability, comparing options, understanding eligibility, querying live information or eventually performing an action.
The third is experience. Once an agent has the right information and tools, there’s no reason every interaction needs to come back as paragraphs of generated text. A comparison might be better presented as a set of cards. A destination query might need a map. A complex decision might be easier to understand through a guided interface. Generative UI opens up the possibility that the interface itself can respond to the task, while still working within the rules and design decisions established by the organisation.
Taken together, those three layers start to move the conversation beyond making content AI-friendly. They ask what an organisation would look like if some of its customers increasingly arrived through agents rather than directly through its website.
Getting ready for an agentic web
This is why emerging approaches such as WebMCP are interesting to us. They point towards a web where organisations don’t only publish pages for people to read. They can also expose structured capabilities that tell agents what they can do and how they can interact with them.
It’s still early, and there are important questions to resolve around standards, security, permissions, governance and how much autonomy people will actually want to give their agents. We don’t think that means every website suddenly disappears or every customer journey becomes agentic. What it does suggest is that organisations should start considering AI as more than another audience for their content.
We’re starting to add these capabilities to Wizen because we think the progression from knowledge to tools is a natural one. AEO and GEO help make an organisation understandable to AI. The next opportunity is to decide how AI should be able to interact with that organisation, then provide a governed way for it to do so.
For us, that’s the shift beyond AEO. The question is moving from whether an AI can find and understand your organisation to what you want an agent to be able to do when it arrives at your front door.