--- title: "Around AI, not on top of it" description: "Bolting AI onto your existing org gets you a small speed-up and a bigger PowerPoint output. Reorganising around it is painful, breaks the org chart, and is the only thing that actually moves the needle." date: 2026-05-27 image: "/uploads/2026/05/the-orchestrator-or-conductor-of-ai.jpg" categories: - "AI" - "Future" - "Leadership" --- Most of the conversations I've been having lately keep landing in the same place. I've been working with a few companies on AI, giving design talks to a few others, and talking to a lot of people about what they're seeing inside their own organisations. Once you start hearing the same things from enough places, a pattern shows up. There are two ways a company can bring AI into the building, and only one of them actually changes anything. The first is to add AI on top of what you already do. The second is to organise everything around it. The first is easy and feels safe. The second is hard, sometimes painful, and is the only one that matters. ## What "on top" looks like You buy ChatGPT seats for everyone. Maybe Copilot. Maybe a Claude subscription for the more curious teams. You send around a Notion page with prompting tips and run a workshop. Leadership ticks the box and moves on to the next quarterly priority. Then people draft emails a bit faster. Someone in marketing knocks out a first pass of a blog post in twenty minutes instead of an hour. A developer autocompletes some boilerplate. A finance analyst gets ChatGPT to explain a regex. That's it. The meetings are still the meetings. PowerPoint is still PowerPoint. The same people sit in the same roles producing the same things, just with a slightly faster first draft. The org chart hasn't moved. No team is meaningfully smaller. The savings are real but they're rounding errors. This is where most companies sit today. They've bought the tools, declared the transformation, and gone back to working the way they did in 2023. ## What "around" looks like Organising around AI starts before anyone touches a keyboard. For every problem, every task, every new piece of work, the first question becomes: can AI do this faster, better, or differently? Not "can AI help me write the email faster," but "does this email need to exist at all if an agent can handle the conversation it's trying to set up?" Not "can AI help speed up this approval process," but "why is there a five-step approval process for something AI could have just done?" The workflow gets redesigned end to end. Roles change shape. One person owns a problem from definition to shipped solution, supported by a small fleet of agents doing the work that used to be spread across five people in three departments. The handoffs disappear because there's nothing to hand off. I've [written before about what this kind of organisation actually looks like](/thoughts/the-ai-organisation-why-most-companies-are-doing-it-wrong). The short version is that it's flatter, smaller, faster, and structurally incompatible with how most companies are arranged today. That's the part most leadership teams don't want to look at. ## Why this is so painful Adding AI on top is a tooling change. Organising around AI is an organisational rewrite. Most companies have only signed up for the first one. When you reorganise around AI, most management roles stop making sense. A lot of middle management exists to coordinate humans: route information between teams, manage dependencies, smooth handoffs, run the stand-up, write the status update. Once the coordination cost falls close to zero, the role does too. That's somebody's career. Approval chains collapse. Job titles stop describing what people do. Teams that used to feel safely staffed start to feel suspiciously large. People who built twenty-year careers on managing other people suddenly need a different answer to "what do you actually do here?" Some will find a great one. Many won't. This is why most companies will keep choosing the easy version. Handing out subscriptions doesn't threaten anyone's job. Rewriting the org chart threatens almost everyone's. Leadership teams are made up of the same people the rewrite is hardest on. You can see why the conversation tends to stall. ## The orchestrator What's needed now is a different kind of person. For every problem they encounter, they automatically ask: can this be solved faster or better with AI? They don't wait for someone to assign that question. It's just how they think. They build the systems, wire up the agents, redesign the workflow, ship the result. They're part product designer, part operator, part engineer, part strategist. They use AI to do the work of a whole team, and they keep going. I don't have a clean name for this role yet. Orchestrator. Conductor. Transformation navigator. Builder. Orchestrator and Conductor feel closest. The work really is about directing a system of moving parts more than building any single thing. The job is too new for the name to fully settle, but it's real, and it's going to be one of the most leveraged seats in any company over the next ten years. Most of these people won't look like traditional AI experts. They won't have "ML" in their job title or a PhD in anything. The strongest ones I've seen are CEOs who quietly built an internal AI system that lets them be everywhere at once. Designers and front-end developers who stepped up and absorbed an entire product organisation. Ops people who replaced their team with five well-tuned agents and a tight feedback loop. The common thread is hands-on experience using AI to solve real, tricky problems, and the willingness to keep doing it without waiting for a mandate. If you're trying to figure out what to be over the next ten years, this is the answer I keep coming back to. Get good at this. Whatever it ends up being called. ## Some people are getting superpowers You can already see this happening, and it's strange to watch. In the same meeting, on the same problem, one person is clicking through Excel and writing paragraphs by hand. They're working hard. They're doing what they've always done, and doing it well. Two seats away, someone else has an agent stack running in the background. By the time the meeting ends, they've shipped the thing the meeting was about. Not a draft of it. The thing. It looks like watching some people get superpowers. The gap isn't about job title or seniority. It's not about who's smarter or better educated. It's about who has rewired how they work and who is still doing the 2022 version of the job. Both are good at what they do. One is operating at ten times the leverage. The gap keeps widening. The people who have crossed over get better every week, because the tools improve every week and they're the ones actually using them. Everyone else is still on the other side, watching it happen, sometimes not even sure what they're looking at. ## If you're starting from scratch The flip side is that if you're starting a company today, or you're still small enough to make the call, this is the easiest moment in a generation to build differently. No legacy org chart to dismantle. No approval chains to unwind. No managers to redeploy or careers to protect. You can decide from day one that every task starts with the around-AI question and that anything that can be automated will be. That's a serious advantage. A two-person company structured around AI from the first hire can run circles around a 200-person incumbent still bolting tools on top of the old shape. The smaller and earlier you are, the lower the friction. The painful version of this change is for companies that already have everything to lose. For new and small ones, it isn't painful. It's just how you build. If you're in that position, force the question. Don't hire for a role until you've tried to solve it with AI first. Don't add a process until you've tried to automate it away. Don't copy the org chart from your last job just because it's familiar. The default shape of a company in 2026 should look very different from the default shape of a company in 2020. ## The 3 to 5 year window Every company that wants to still be around in 3 to 5 years has to go through some version of this transition. Not as a strategic initiative or a transformation programme. As a survival question. The competitor that organises around AI will eventually eat the one that didn't. Highly competitive industries will feel this first and hardest. Protected niches get a bit longer, but not much. Even there, the costs and capabilities are shifting fast enough that "we have time" turns out to be the wrong reading of the room. There isn't a clean recipe. Every company's path will look different depending on what they do, who they have, what the org looks like today. You have to look at your own work, your own structure, your own people, and start asking the around-AI question for real, with the willingness to take the answers seriously even when they're uncomfortable. The clock is shorter than most leadership teams want to admit. The companies that move now still have time to do this in a considered way. The ones that wait will have it done to them. I'm watching this play out across the companies I work with. Some are leaning in, painfully, honestly, and the change is starting to show. Most are still bolting tools on top of an unchanged organisation and hoping that's enough. I'm pretty sure it isn't. If you're reading this, the most useful thing you can probably do is start being one of those orchestrators inside whatever role you sit in today. Don't wait for permission. Don't wait for a job title. Don't wait for a clean recipe that nobody is going to write for you. The people who figure this out first will define what comes next. It really does look like watching some people get superpowers. I'd rather be one of the people getting them than one of the people still wondering where they went.