Two months at Apoco

It’s been a long time since I wrote here, and I have a couple of things I’d like to share with you. Let’s start with work, as it’s now been two months since I joined Apoco as an AI lead.

To be very honest, I was looking for a job and I needed the money, so I was pretty happy to finally have work again, not that I’d been desperate, but the timing was impeccable :). I knew the guys behind Apoco from my IBM times, and I’d always thought they were really good at what they do with quite a nice non-bs attitude. They’d built an interesting team around them, and with their work being close to research, it meant there would be some nice, hard challenges to get into.

When we were talking about me joining, we discussed their product Brainio and the idea of developing it into a platform where people and agents work together in a team. Ondra offered me the chance to lead the product development, which would also mean doing research along the way and finding out what the hell this thing should be. What exactly we wanted to build wasn’t clear from the start, and only just about now is it starting to have more chiselled contours.

A lot of it comes down to how people and agents can work together. We talk about agents as coworkers, but what does that actually mean? How do you communicate with them, what do you let them do, and how much do they need to know? And of course, how do you make all of that work without spending a fucking fortune on it? Because you can’t just use the best models for everything and let them burn through tokens, right? So how do we become more productive when working with AI is becoming part of the job?

In the first few weeks, I think I fell into a bit of a trap. I started working a lot on the daily development and operations, and I didn’t give myself enough time to step back and experiment. There were discussions along the way, of course, but much of my time went into working with AI on the product, iterating, documenting, researching... and getting quite a lot of slop back.

Preparing for our status and strategy meetings was also quite a struggle. I’d prepare with AI agents in Claude Code or Pi, working through what the product could do for people, which problems it could help with and what that would be worth to them. And quite often, what came back sounded like rubbish. It really pained my mind, because with software architecture I could get somewhere useful, but these more open product discussions just weren’t working very well for me.

A couple of days ago, that changed. With Opus 5.5, I’m getting somewhere with those product questions that I couldn’t with Astra, another AI model I’d been using. We’ve put together an awesome blueprint for Brainio, and when I ask it to help me prepare for a strategy or status meeting, it gets pretty close to what I want in one go. Honestly, it’s a bit of a gamechanger for me so far.

And there’s been quite a lot to discuss. Over the past week, we’ve changed direction quite a few times. I’ve been working on a framework for AI-native productivity, trying to work out where agents can help a team and how to measure whether they do. We’ve also been giving Brainio more shape, getting into building it and exploring whether we can use our digital asset management platform as the foundation.

Working on all this with agents, I can have three or four things going at once, which sounds pretty good until I have to keep up with all of them. An agent has a plan or a proposal for me, and I need to read through it and figure out what I think. Each time I switch, I need to get back into what we were doing and why, before I can even consider what the agent is telling me. And after a while, I’m really tired from all that reading and switching between things, even when the work coming back is good.

I’ve been wondering whether some of that could be conveyed visually, or through audio. I still need to understand what’s going on, but reading through another long document when my attention is already somewhere else doesn’t seem like a particularly good way to do it.

I’m also trying to give my working day a bit more rhythm. I still use Brain.fm for those twenty-five minutes of work, followed by five minutes of care. And by care I also mean exercise, because I really need to move more, to keep my mind and blood flowing. Sometimes I need a bit of hype and energy too, as I can get quite sluggish. This helps me focus and reminds me to take care of myself along the way.

I can still get carried away with the work, partly because it’s fun and partly because I feel that push to get things done. But I’m getting better at it, and with things at work settling down a little too, I can finally calm down a bit.

What I’m looking forward to now is exploring how we can help teams and organisations become more productive with agents, and how we can measure the difference. My own experience makes me want to include the time and attention it takes to work with them in that picture. If agents let us do more things at once, but we spend the day catching up with them, how much have we really gained? That’s one of the questions I want to explore with the framework.

There are still a lot of discussions to have at Apoco about how we get there. I’m looking forward to working through those with the team and with the partners we find along the way, trying things and finding out what actually helps.

Does any of this sound familiar? If you’re trying to figure it out too, reach out. I’d love to hear how it’s going for you.