We were wrong about that in more ways than one. We didn’t yet know we were terrible at project management or that we’d eventually need an actual QA process. You don’t know what you don’t know, they say, but what we did know was that the hours were long and we weren’t getting much sleep. Eventually the math stopped working. We needed help, badly, and that was a good problem to have.
But it was still a problem. Hiring someone meant something more than filling an open seat — it meant a person would rely on our decisions to make a living. My name would be on someone’s mortgage or car loan, in a roundabout way, whether I liked it or not. We hired anyway, because we needed the help more than we needed to avoid the fear. And in doing that, it felt like we became a 'real' company.
Twenty years later, we still feel a version of that same hesitation before every hire. Blend is still a small agency, with a headcount of 24. Every person we add changes the financial picture, and it changes the culture itself, so we think hard before we do it, every single time.
I used to assume this hesitation to hire was a startup problem, and we’d grow out of it. I don’t think that anymore. Talk to a director at a Fortune 500 company about adding headcount, and you’ll hear the same three questions:
- Do we actually have the budget to add someone?
- Where does this person fit into our established workflow?
- Who is responsible for training, managing, and covering for them while they ramp up?
For a small agency or a global enterprise, hiring is never just a budget line item. It’s a bet on capacity, fit, and time, all at once.
Our AI conversation is the wrong conversation
Most of the AI conversation in our industry is still about content and efficiency. Can it write a blog post? Can it draft some ad copy? Can it summarize this report? Those are useful questions, but they’re the wrong ones to build a strategy around.
Arnold Macauley made this same point in a recent LinkedIn piece: AI's best output isn't content. It's actually capacity. To take that a step further: for an agency owner who has spent over two decades weighing every single hire, capacity is the whole game. Content is just part of what that capacity produces.
That's the lens I use now when I look at Optimizely Agent Platform. Mark (formerly Opal) isn’t one more chatbot window. It’s a system directly connected to the workflows we already run in Optimizely CMS and CMP (Content Marketing Platform), and because it's aware of our brand standards and content pool, it is built to hand off real work rather than just answer questions.
This week, Optimizely introduced something new on top of that platform: Virtual Teammates. This isn't another name for an AI agent. Each Virtual Teammate gets its own identity, the same way a new hire walks in with a name and a personality, not just a job description. There’s an SEO and AI Search Analyst, a Marketing Analyst, a Personalization Strategist, a CRO Manager, a Chief of Staff, each built for a specific role rather than one generalist trying to do everything. It's less like adding a tool and more like making a hire, minus everything that used to make me hesitate.
In fact, Optimizely's Virtual Teammates and Agent Platform map almost exactly onto those three Fortune 500 questions I mentioned before.
-
Budget: A new hire's cost is real regardless of whether the work justifies it — they're not any cheaper while you're onboarding them. On the other hand, Mark and Virtual Teammates run on usage, so the capacity scales with the work in front of us instead of sitting on payroll during the slow months.
-
Fit: This is the one that used to worry me most about any new AI tool. Where does it actually live in the workflow? Mark doesn’t sit off to the side waiting for someone to remember to open it. He’s inside the CMS pages our team is already editing and the experiments our strategists are already running. There’s nothing new to bolt on.
-
Time: Every new hire costs the team time before it gives any back, usually months of it before they’re fully ramped. Virtual Teammates skipped that part. They showed up already knowing our brand and our workflows, and now they run on schedules and triggers in addition to prompts. We wake up to a weekly digest on web strategy, or to automated regression tests that follow our testing system's guidelines and requirements. While we sleep!
None of this replaces the decision to hire. We still add people to our team, and we will continue to do so. What's changed is the shape of the decision. We’re no longer weighing between “hire someone or don’t get the work done” — instead, we’re balancing hiring with extending the team you already have. That second option didn’t really exist a few years ago.
An extension, not a replacement
That’s the part I don’t want lost in any of this. The whole reason to hand off work to an agentic teammate is so the humans on our team can do the parts of the job that require human ingenuity. Things like strategy conversations with clients, or judgment calls on a design solution, or the actual relationships that have been built over years.
It’s been eighteen years since we became Optimizely partners, and after sitting on the partner advisory board and watching where this platform is headed, I keep coming back to why this new set of tools feels so personal and even life-changing. We didn’t build Blend so our team could spend their days banging their heads against repetitive work — we built it so we could do the work we loved. Optimizely makes this possible by taking on the repetitive work, letting our team get home at a decent hour while they’re at it.
Put Mark to work on the audits and the first drafts. Then use that saved time to be more present with your clients, your customers, and your people. That was the whole point of hiring that first person twenty years ago — to delegate. Turns out I love to delegate. Especially to Mark, who loves to work overtime.