Opticon 2026: When AI stopped being a feature and became the operating model
Justin Masters
Before Opticon, I wrote that there was no shortage of conversation around AI, but that what mattered was how quickly the ideas were moving beyond the hype and into practical application, and where they could create meaningful value.
Coming out of Opticon New York, that still feels like the right question. What has changed is how close the next phase now feels.
There was a lot of new tech on show: autonomous experimentation, generative personalisation, Virtual Teammates, marketing-specific AI models and digital experiences designed to be understood and acted upon by AI agents.
Opticon 2026 though wasn’t really an AI product story. It was a preview of what happens to the digital operating model when AI stops being a feature and starts becoming part of the workforce.
The pace has changed
One of the comments from a panellist on the main stage was that “we are just at the start of a seismic change in customer experience.” That captured the mood well.
What Optimizely was showing a year ago was impressive in the context of where AI was then. The distance travelled since mirrors what we are seeing across the wider market.
McKinsey’s latest State of AI research found that 40% of large enterprises are now scaling AI agents somewhere in their organisation, up from 27% the previous year.
Optimizely’s Virtual Teammates are interesting in that context, less because of the functionality and more because of what they imply. When an AI has a defined role, access to organisational knowledge, objectives and the ability to undertake work, questions about ownership, accountability, governance and where people fit become operating-model questions rather than technology questions.
For digital leaders, the issue is increasingly not whether to adopt AI, but where to apply it and how quickly the organisation can learn to work differently around it.
AI will become more specialised
Optimizely developing its own models optimised for marketing was another key area of interest for me.
It is easy to think that AI strategy means adopting whichever frontier model currently sits at the top of the leaderboard. I don’t think that is where this goes.
Stanford’s 2026 AI Index shows leading models clustering much more closely on overall performance. As capability converges, cost, reliability and domain-specific performance become more important.
Sometimes the right choice will be a frontier model. Sometimes it will be a smaller or specialised model that is faster, cheaper and better suited to the task.
Optimizely training models around marketing is a good example of how AI will need to shape-shift around the work rather than the other way around.
Personalisation may finally become personal
Personalisation has been one of digital’s longest-running promises.
In reality, most personalisation has meant creating a manageable number of segments, producing variations for them and deciding which somebody should see. The constraint has always been as operational as technological: somebody still has to conceive, write, design, approve and maintain every experience.
What interested me at Opticon was the move from selecting between predefined experiences towards dynamically assembling an experience around an individual’s context and intent.
“The best version of your website for a particular customer may not exist until they arrive” was broadly a quote from Opticon, and this really resonated with me.
That gets much closer to the original promise of one-to-one personalisation and could make it achievable for organisations without Amazon-sized teams and budgets.
It also makes the fundamentals more important. Trusted data, strong content, clear brand rules and sensible governance matter even more when the number of possible experiences becomes effectively limitless.
The website is going to have a new audience
Another significant takeaway from Opticon was the addition of a third experience alongside the marketer and customer experience: the agent experience.
We have spent decades designing websites for people. As digital becomes increasingly agent-driven, AI will become an intermediary between organisations and some of their customers.
WebMCP, which Optimizely previewed at Opticon, gives a practical indication of what that could mean. Rather than an agent having to navigate a website in the same way a person does, a site can expose structured actions that an agent can understand and use.
That shifts the question beyond whether your content can be found by AI. Can your products and services be understood and acted upon by AI?
Getting ready for that will become an important digital priority over the next year or two. It affects content structure, APIs, product architecture, data and ultimately the role of the website itself.
The differentiator will be how quickly organisations learn
The mistake would be to respond to all of this change with a collection of disconnected AI pilots.
The opportunity should be to look at how work actually gets done: where decisions stall, where hand-offs slow things down, where knowledge sits with individuals, and where AI could change the process rather than simply make one task faster.
Not everything needs a three-year business case before it can be tried. The pace of change is too fast for that. Organisations still need commercial discipline, but they also need to recognise that learning has value.
The competitive advantage may ultimately be “cycle time”, which was discussed in the opening keynote: how quickly an organisation can turn an idea into something real, learn from it and make the next decision.
Technology advantages get copied. Platforms are available to competitors. Models become commoditised. An organisation that has learnt how to learn faster is much harder to replicate.
There is an important caveat to that speed. Optimizely CEO, Alex Atzberger used a phrase at Opticon that stuck with me: “garbage in, gospel out.”
AI can take weak assumptions, incomplete information or poor data and turn them into something polished and convincing. Moving faster only helps if the inputs, objectives and judgement behind the work are sound.
What I took away from Opticon
The most interesting thing about Opticon wasn’t any individual announcement. It was how clearly the direction of travel is now coming into view.
Some of this is already here. Much more is coming quickly.
For digital leaders, the job isn’t to predict exactly where this lands. It is to make sure their organisation is learning and adapting quickly enough to be ready for it.
Couldn’t make it to New York? There’s still time to join the conversation.
We’ll be at Opticon London on 13 October, where we’ll be exploring what this shift towards agentic digital experiences means in practice. Join us to hear from digital leaders, see the latest developments from Optimizely and continue the conversation around what comes next.