Imagine five new specialists joining your marketing team tomorrow morning. Each one already understands your brand and your data, and none of them need three months of onboarding.
Optimizely has introduced Virtual Teammates: five named, role-based agents, each built to own one job end-to end-rather than answer prompts one at a time.
This marks an important change in the way organisations use AI. As Optimizely explained in their recent 'Meet the Virtual Teammates' webinar, IDC and Lenovo put 88% of companies in what they call 'AI pilot purgatory', and BCG puts the number that have genuinely scaled AI use at just 4%. The pattern both figures is the same: a content writer uses one tool, an SEO lead uses another, an analyst uses a third, and none of it shares memory or results. Virtual Teammates are built to occupy one role, so the work is traceable and the output is cumulative.
Five roles, each owning a job end to end
Chief of Staff prepares for meetings, tracks follow-ups, monitors competitors and delivers recurring briefings.
SEO & AI Search Analyst monitors search and AI-search visibility, audits performance and reports on what it's changed.
Marketing Analyst turns campaign and conversion data into a readable brief, without anyone logging into GA4 to find it.
Personalisation Strategist finds audience opportunities and stages campaigns ready to launch.
CRO Manager plans, runs and reports on experiments end to end.
They run on Optimizely's own purpose-built marketing models. Every teammate holds its own identity and permissions, so activity is scoped and auditable, not a shared login with the keys to everything.
"A Virtual Teammate learns how your organisation works once, then keeps going."
- Alex Atzberger, CEO, Optimizely.
The distinction that matters here isn't speed, it's persistence and continuity. A prompted AI agent starts from zero every session: no memory of last week's conversation, no awareness that a meeting outcome just changed the whole approach. A Virtual Teammate carries that context forward, acts on standing instructions without being asked each time, and flags only the decisions you've instructed it to ask a human to make the call.
What's ready to automate now?
Start with work that's repetitive, well-defined and already has a clear owner. Recurring reporting is one obvious place to start: a weekly performance brief, a monthly stakeholder update, a competitor analysis. These already follow a template, so an agent that watches the data and writes the brief removes hours of manual pulling and formatting without changing what anyone decides.
SEO and AI-search monitoring is another strong candidate. AI referral traffic and crawlers now make up a meaningful share of visits to most sites, and that's something most teams check occasionally rather than continuously. An agent that checks it every day flags changes as they happen, instead of at the next scheduled review.
The same logic applies to experimentation, where a testing programme already exists. Designing a test, running it correctly and writing up the result in plain language is exactly the kind of end-to-end, rules-based work an agent does well, freeing the people who'd otherwise write that report to decide what to test next instead.
Getting the groundwork right first
An agent works from the content and data you already have. Getting the underlying structure right - clear tagging, a clear owner for what's current versus outdated - pays off directly in how useful the output is. That's the work to get right in the weeks before a wider rollout.
The same applies to workflow. Consolidate intake, approvals and sign-off into one clear path before automating what runs through it. Having a single clear path turns an agent's output straight into a decision.
Ensure governance is set per teammate. Each should have its own scoped permissions across the tools and data you authorise, with every action carrying a full audit trail back to that teammate. You decide where it needs sign-off before it acts, and where it can run on its own. That's usually the first question an IT lead will ask, and it should be answered at the point you bring a teammate onboard.
Test any specific recommendation an agent makes, especially in a fast-moving area like AI-search optimisation, against your own data. The field is moving fast enough that best practice is still being written, and testing it yourselves is part of the value.
Putting the first agent to work
Mando Group is an Optimizely Platinum Partner, recognised by Optimizely as an Agent Platform onboarding specialist, based in the UK. We help marketing and digital teams work out what to bring in first, how get the governance and data foundations right, and how to configure the agents you choose, so the output goes straight to someone who can act on it.
The best place to start is finding out where you stand today. Our agentic readiness assessment scores your organisation across agentic AI, agentic CMS and agentic experimentation, then gives you a clear set of next steps based on your result. It takes a few minutes, and you'll get a personalised report to work from.
