Leah MessengerEvery company has a Kevin.
Ask a question in Slack or teams — any question, about any product, any edge case, anything that happened three reorgs ago — and someone will say "just ask Kevin." He'll know. He just always knows.
Which is great... until Kevin's out.
Because when people like Kevin take a week off, three things can happen simultaneously:
A campaign launch parks mid-flight because nobody's confident about the brand guidelines exception Kevin explained once, in a Notion comment, eighteen months ago.
A message thread about which experiment variant to call reaches seventeen replies, two conflicting reads, and one link to a doc that's been stale since Q1 before someone says "I'll just ask Kevin when he's back".
And Kevin himself answers a "so sorry, just a really quick one" message from his airport gate, because he always does, because the alternative is watching someone get it wrong.
None of this is Kevin's fault. It's just what happens when one person becomes the answer layer for an entire team; the OOO becomes a waiting room, the holiday becomes conditional, and work learns to be patient in ways it really shouldn't have to be.
"Just ask [insert your Kevin equivalent here]" is the highest praise a team can give someone. But on a deeper level, it's also a structural problem.
When one person's expertise becomes the routing layer for an entire team's decisions, that expertise doesn't scale. Instead, you just get queues on queues on queues.
Every questions waits for Kevin to get off calls, finish priority tasks, or get back from holiday. Every answer lives in Kevin's head, not in a doc or system... or anywhere Kevin himself could hand off if he wanted to.
And of course, we've all tried to fix this. We've all built the wiki. We've all watched it go stale within a quarter, because the useful part of Kevin's knowledge was never the process. Instead, it was the judgment, the "well, it depends", the three exceptions he applies without noticing he's applying them.
Because, this is the thing: documentation can easily capture what the process is — the step 1, step 2, and step 3 — but it's much harder to capture how Kevin decides.
The AI era was supposed to solve this, and mostly hasn't. In Optimizely's 2026 Global AI and Marketing Study, 76% of marketing leaders said they spend three or more hours every week fixing, editing, or fact-checking AI output — and only 4% said AI genuinely saves them time at every stage.
Access to AI was never the bottleneck, but getting AI to actually hold context, the way a colleague does, was. It's not that Kevin is a bottleneck on purpose. It's that nobody ever built a way to copy what's in his head.
A Virtual Teammate isn't a chatbot that answers when you prompt it. It doesn't wait around for someone to type a question into a box.
It's a member of your marketing team, manned by the real humans in your marketing team. Complete with a name of your choice, a place in the team roster, and access to the tools where the work actually happens. When it does something — like updates a campaign, flags a content issue, ships an analysis — there's a record of exactly who did it, the same as there would be for anyone else on the team.
It's called accountability, folks, and we all need it to take accountability (even AI).
And the best bit is that it's already working, already learning. It sits in on calls, it watches how Kevin works — not just what he says the process is, but the shortcuts he takes and the judgment calls he makes without thinking twice.
Over time it doesn't just store Kevin's answers. It actually picks up on how Kevin works. Memory persists across sessions: brand voice, team conventions, past decisions, previous campaign performance, the feedback Kevin gave in March that everyone else had forgotten. That all adds up, and is memorized.
That's the difference between a tool you have to operate and an AI teammate you can actually rely on.
Worth being honest about: none of this happens because you turned something on.
The teams getting real value, properly onboard their Virtual Teammates. But the good news is, while ramping up a real-life human teammate can take 8-12 months, Virtual Teammates take a whole lot less than that with these best practices. You don't need to worry about those weekly 1/1s either.
Brief it on day one with brand guidelines, style docs, work you're proud of, jargon that you can't stand, jargon that you can't live without. Then set them off to let them work for you and learn from you. Their self-learning and memorizing capabilities — along with your reviews, corrections, and occasional redirections — help them develop how they work for you.
With Virtual Teammates, making corrections isn't time wasted... it's the mind, knowledge, and expertise transfer happening to make the outcomes more valuable to you. đź§
Some examples of Optimizely Virtual Teammates? Sure, you got it:
The SEO & AI Search Analyst: Your SEO lead sets the strategy and decides what to prioritize, while this Virtual Teammate monitors, audits, and acts on findings daily so that person isn't spending their week in dashboards.
The CRO Manager: Your experimentation lead owns the hypotheses, reads the results, and makes the final calls. This AI teammate runs the mechanics — sample sizes, experiment staging, daily anomaly checks — so the rest of the team spends time on the hard stuff, not the experimentation admin.
The Personalization Strategist: Your personalization team decides what's worth personalizing and why, while this VT gets you from that decision to a stage, monitored campaign faster, with every campaign decision staying firmly with the human marketers.
The Marketing Analyst: Your analysts interpret what matters and set priorities, but sets this VT off to pull the data, flag anomalies, and build reports so they can spend more time on insight, not data extraction.
The Web Manager: This AI teammate helps your (human) web manager by proactively sweeping for redirects, tracking technical errors, drafting brand-compliant CMS templates, and screening pages for accessibility issues before any human reviewer touches them.
Here's where it gets fun. Once a Virtual Teammate has learned how Kevin genuinely thinks, a few things become possible that weren't before (and I think you're going to like this bit):
A Virtual Teammate can keep things moving overnight: Not just answering when asked. Running on schedules and triggers, a VT can flag the content that's gone stale, and another one can catch the anomaly in the experiment data before it becomes someone's problem on Tuesday.
A Virtual Teammate can brief whoever's covering: The person filling in for Kevin usually inherits the tasks and next-to-no context, but this is the first time the context comes with them.
A Virtual Teammate makes his OOO email more honest than ever: Keeping the promise of "I'll have limited access to email" or "I'm out and unreachable until the 14th" is always the hard part, but with Virtual Teammates, nothing needs to be piling up or waiting around.
It reflects Kevin's judgment and patterns. It is not a replica or replacement, because (let us shout this from the rooftops) the human is still on the hook.
Kevin still decides which content goes out, which experiment gets run, which brief is actually ready. His Virtual Teammate handles the work that shouldn't require Kevin's calendar to clear first — and it does it inside the same platform his team already uses, configured in plain English, without an engineering ticket in sight (yay for your developer team).
A Virtual Teammate doesn't make Kevin's expertise redundant. Absolutely not. If anything, it protects it — takes what's only ever lived in his head and makes sure it doesn't disappear the moment he's out sick, on PTO, or just heads-down on something else. The idea of VTs was never to make Kevin, or anyone, less essential; it's to make sure the thing that makes them essential (their judgement) doesn't have a single point of failure.
His day gets to contain the work only Kevin can do, instead of the forty interruptions that route through him because nobody built a better way.
