At Opticon in New York this August, Optimizely launched Virtual Teammates: AI coworkers built around specific marketing roles, which sit within Agent Platform and take on ongoing responsibilities rather than responding to one prompt at a time. It is the clearest sign yet that AI agents are becoming a core part of how Optimizely works.
Agent Platform brings together specialised agents, tools, data, integrations and workflows to carry out tasks across marketing and digital operations. Agents can analyse information, create outputs, interact with other systems, trigger actions and work together as part of a wider process.
That opens up a lot of possibilities. It also raises an obvious question for any organisation that is AI driven and already using Optimizely: where would any of this make a meaningful difference?
Rather than starting from the idea that you need an AI agent, it makes more sense to start with the work your teams already do. Look for where time is being lost, where processes are more manual than they need to be, where information is hard to use, and where worthwhile activity simply isn't happening because nobody has the capacity.
Here are the places we see where Agent Platform is most likely to earn its keep.
Where repetitive work is taking up valuable time
Some of the clearest opportunities are in activities people already do on repeat: compiling regular performance reports, checking for changes in customer behaviour, reviewing search performance, monitoring experiments, looking for personalisation opportunities, checking content against agreed standards, or pulling information together from several sources. None of this is necessarily difficult, but the cumulative time adds up.
This is where Virtual Teammates are worth a closer look, because they are designed around ongoing roles rather than individual prompts. You can think of these team mates as specialised virtual assistants who can also be shared across team members based on specific tasks and roles. The launch lineup includes a Chief of Staff, an SEO & AI Search Analyst, a Marketing Analyst, a Personalization Strategist and a CRO Manager. The Marketing Analyst, for example, is built to turn marketing data into insight by monitoring acquisition, content performance and conversion trends, while the CRO Manager analyses experiments and performance and recommends where to act. Virtual Teammates can also work on a schedule, so a weekly briefing arrives without anyone having to ask for it.
In short, when useful analysis happens consistently, the people who make decisions can spend more of their time interpreting the findings, deciding what matters and acting on it.
Where data isn't turning into action
Most organisations aren't short of data. The hard part is acting on it. Someone has to find the data, interpret it, decide whether it matters, work out what should happen next, create or configure the response, get it approved and then put it live. Several systems and several people may be involved along the way.
This is where orchestration becomes more interesting than any single agent. Agent Platform lets agents use tools, connect with other systems and be chained together into workflows, so an agent becomes part of a process rather than producing an answer and stopping there. Those workflows can be designed so that a person reviews the output at the points that matter.
For example, one agent could spot an unusual change in website performance, a second step could investigate the pages or journey affected, and the workflow could finish by preparing a recommended response for someone to review. An experimentation team might use agents to move from identifying a problem to a testable hypothesis, while a personalisation team could use them to surface potential audiences or opportunities.
Many of these processes don't begin and end inside Optimizely. Customer information might sit in a CRM, performance data in an analytics platform, and product, commercial or operational information somewhere else entirely. Agent Platform includes connectors and tools for working with other systems, and custom tools can be built where an agent needs to reach information or take an action that isn't available out of the box.
So a single workflow could retrieve information from one system, analyse it alongside data from another, apply your business rules, create an output and pass it on for approval or action. If your teams regularly export data, copy it between systems, reconcile different sources or send routine updates, there is likely an opportunity to connect more of that process.
Where valuable work isn't happening at all
Saving time on existing processes is only part of the value. Some of the biggest opportunities may be in work your organisation doesn't do today, because there simply isn't time.
You might want to monitor hundreds of pages for content quality, analyse customer behaviour more often, keep a closer eye on search visibility, find personalisation opportunities across a large site, or review digital performance continuously rather than once a month. Each of these can be valuable. At scale, they need more time than one person, or even a whole team, has available.
Agents change the economics of that work. So as well as asking where Agent Platform could save time on an existing task, it's worth asking what it could make possible at a frequency or scale that wasn't practical before.
Automating a poor process just makes it faster
There is little value in using Agent Platform to speed up a poor process. If something is overcomplicated, poorly defined or produces little value today, adding an agent won't fix the underlying problem.
Before building an agent or workflow, it's worth understanding why the work exists, who uses the output, what decision it supports, what systems and information it relies on, and what a good result looks like. That exercise may confirm a strong use case. It could equally show that the better answer is to simplify the process, remove a step, improve the underlying data or stop doing the work altogether.
Just because something can be automated doesn't mean it should be.
How much should you let an agent do?
Giving an agent the ability to take action raises different questions from using AI to draft copy. You need to decide what information it can access, which systems it can interact with, what it can do on its own, and where a person still needs to approve the outcome.
Optimizely has built control into the platform. Virtual Teammates have their own identity and permissions, so their access can be managed much like that of any member of the team. Agent outputs can be scored automatically against quality criteria you define, and execution guardrails flag runs that drift from an agent's normal behaviour.
That means autonomy can be increased step by step rather than as an all or nothing decision. For most businesses, a sensible first use case is one where an agent does the time consuming analysis or preparation and a person still makes the decision. Once you've seen how reliably that works, you can decide whether more of the process should be automated. The right level of human involvement depends on the task, the quality of the information available and the consequences of getting it wrong.
Start with a problem worth solving
With a growing number of agents, Virtual Teammates, tools and workflows available, it would be easy to produce a long list of things you could automate. That list won't necessarily tell you where the value is. However for some organisations if you have the capacity for change across complex processes, then making time to identify where the value lies can still be useful exercise.
A better starting point for many however, is a recognisable business problem: too much manual effort, information that isn't acted on, repetitive work across systems, inconsistent execution, or valuable activity that doesn't happen often enough. Understand what happens today and where the friction sits, look at the information and systems involved, and then decide what an agent could reasonably take on.
From there you can judge whether solving that problem creates enough value to justify the effort, how you'll measure the improvement, and what role human judgement still needs to play.
How we can help
DotCentric has worked with Optimizely for more than 15 years, and we are one of the few Optimizely Agent Platform Specialized Partners, recognised for our expertise in helping Optimizely customers implement their AI agent strategy and technology. That covers developing agents, custom tools and workflows within Agent Platform. Just as importantly, it means helping you work out where the technology will genuinely improve the way something works, what it will take to get there, and which use cases aren't worth the investment.
If you're an Optimizely customer trying to work out where Agent Platform could make a difference, talk to us about the problems and opportunities you're seeing. An Agent Platform workshop is often the quickest way to find the use cases worth pursuing, and we can help you decide whether one would be right for your team.
