Optimizely’s native AI agent capabilities: Five takeaways from the Opticon keynote

Optimizely recently held its annual Opticon event, both with a physical event in the US and in London, and an online version which includes some recordings from the in-person conference. Like other flagship events from CMS and DXP providers such as Umbraco’s Codegarden and Sitecore Symposium, Opticon includes the usual mix of practical sessions, celebrations of success and key product announcements.

This year’s Opticon was less about new revelations and more about showcasing current capabilities around agentic AI that are embedded into Optimizely Opal product. Opal is the main branded offering for the AI features integrated across the Optimizely One platform.

In this post we’re going to cover five of the key takeaways from the keynote session – featuring CEO Alex Atzberger and Chief Product Officer Rupali Jain – with a focus on the new agentic AI features within Opal.

1. Optimizely’s core principles are around content, customer superpowers and ROI

At the start of the keynote session, Alex Atzberger, Optimizely’s CEO, gave a rousing speech where he positioned Optimizely around three key principles:

1. Content is at the heart of the digital experience: Atzberger emphasised that Optimizely’s offering is crafted around the entire lifecycle for content and digital experiences, which echoes the positioning of the overall “Optimizely One” platform.

2. Give customers superpowers: Optimizely is designed to provide “scientific insights” for higher conversions, workflows to avoid menial tasks, and the opportunity to understand the outcomes of testing using advanced data; for this last point, Atzberger singled out the acquisition of data warehouse and analytics provider NetSpring, which has now rebranded as Optimizely Analytics and is integrated into the platform.

3. Provide the highest ROI: Atzberger declared Optimizely’s model for achieving ROI was to provide a full-featured platform that reduces costs by eliminating the need to deploy multiple other solutions, while also increasing revenue by generating more conversions. He also pointed out that over a third of customers (35%) are now using more than one Optimizely solution. Notably, he hinted that it was a different path to other DXPs that had gone down more of the composable route.

2. Optimizely Opal is being positioned as “the agent orchestration platform”

Atzberger stressed that AI is not changing Optimizely’s core principles and also should not change customer’s visions either – it was just changing how things are done:

• We all need to remain customer-first rather than AI-first, but AI is changing how customers are behaving.
• We need to change the way we work to break down silos, reduce bottlenecks, and automate manual work.
• We need to tap into the potential for AI.

At the end of the speech Atzberger then finally came on to the real focus of the keynote session – the promotion of Opal and its agentic AI capabilities. Atzberger positioned Opal as “the agent orchestration platform for marketing” with agents that:

• are fully embedded within workflows.
• understand your context such as brand governance.
• have been purpose-built for specific use using sophisticated tools.
• act autonomously on your behalf.
• have the potential to drive exponential value at scale.
• are able to orchestrate using the platform.
• are enterprise ready.

3. Opal’s AI agents support the rapid creation of sites from designs

Rupali Jain, Optimizely’s Chief Product Officer, led the next session which included a deeper dive into Opal’s agentic capabilities, illustrated by product demos.

Jain positioned the concept of “agent orchestration” as the intersection between the ability for humans to create agents in a simple and intuitive way, and the ability to execute the agents autonomously, without the need from help from IT. The point of marketers being fully enabled to design and execute agents without technical help is an attribute that Optimizely are keen to stress.

To illustrate the full power of Opal Agents we were presented with a slick but relatable demo. This focused on the example of an imaginary fitness brand called Stride which is sponsoring the NYC triathlon. Stride’s marketing team want to create an accompanying campaign website and the demo focused on how AI agents support the process.

The agents are accessed from a home screen which presents various prompts including:

• Optimize webpage
• Plan campaign
• Create landing page
• Find the right audience
• Analyse page performance
• Repurpose your assets

In the scenario presented in the demo, the internal team had already created a proposed campaign site design in Figma, but this required some design tweaks and also included components within the Optimizely CMS that did not exist.

In what was a genuinely impressive sequence showing how a design could be converted into a live website within minutes:

• The “create landing page” prompt within Opal was selected.
• It was explained that brand and design context in terms of various rules were already inherent in the platform.
• Just by providing a URL to the Figma design, the agent automatically broke the design down into potential frames or components, and then assessed which ones were covered by existing components within the CMS, and which ones would need to be new.
• The user could then select all the potential components to take over to the campaign site, adding qualifiers and adjustments to the design for each component.
• Opal created new components (content types) based on the design, all within the Optimizely CMS which could then be reused across the entire website in the future.
• The website for Stride was actually headless, and the demo showed how Opal could automatically generate the code for all the new content types, again with the ability for any code to be changed if necessary.
• The code could then be automatically pushed to a staging branch of GitHub before deploying them to the live site.
• A designer could then preview the draft campaign site within the “Visual Builder” editor.
• Opal also automatically suggested and crated any metadata for SEO and GEO purposes.
• By hitting “publish” the site went live.

Overall, this was a slick demo that shows the potential of AI agents to save vast amounts of time, while successfully keeping humans in the loop.

4. The agents also handle site optimisation

The session also included another impressive demo of an Opal agent that deals with the optimisation process, covering the following stages:

• Experiment ideation
• Experiment planning
• Experiment review
• Variation development
• Results summary.

Specifically, in the demo:

• The marketer selects the “Optimize webpage” prompt.
• Again, the agent already knows the full brand and design context and can operate within these appropriate guardrails.
• The agents present some prompts including “Embrace visual impact and value,” “Drive action with urgency and proof” and “Optimise use experience”
• On clicking on one of these (“Drive action.”) more granular prompts are displayed, including “New aspirational hero image,” “Registration headline”, “Updated headline” and “Live registration counters”.
• By selecting “Live registration counters” the AI comes up with two choices for the hero area, one with a registration counter, and one without.
• It’s possible then to ask the AI to change the design using the prompt, for example making the registration counter slightly larger.
• The user can then instruct Opal to automatically run tests with the two approaches.
• After two weeks Opal can notify you with a report that shows the variation with the best results, and the marketer can further dig into other analytics.
• You can then get Opal to automatically implement the recommended design.
• Additionally, an agent also wraps up any learning and findings in an email which can be shared with others.

5. You can build your own agents

Finally, some of the other capabilities of Opal were also covered including:

• The ability for customer to build their own AI agents through a no code interface, empowering marketing teams to do so without assistance from IT.
• Tools for extensibility.
• An agent directory with pre-built agents from both Optimizely and partners, as well as those built by the client
• Workflow capabilities that can sequence different agents together to drive more complex end-to-end processes, all defined via an easy-to-use authoring canvas.

Welcome to the agentic AI era

Overall, it is clear that Optimizely is pushing Opal’s agentic AI capabilities hard, claiming that they can produce a 79% increase in velocity and a 10% increase in win rates. But the demos were genuinely impressive and show the agents have the potential to make a real difference to the process of creating and optimising digital experience. 2026 year may just be the year of the AI agent.

If you’d like to discuss how AI agents might impact your CMS, or want to discuss Optimizely Opal, then get in touch!