The queue in Munich
I was in the immigration queue at Munich airport (an oddly regular feature in my life) and I found myself talking to an American who’d ended up beside me. He asked what I do. I gave my usual generic response of corporate training, specifically working with teams. His reply came straight back, “That’s a difficult business now. AI is taking over everything. Do you even need a team?!” I don’t think I gave a particularly good answer at the time, but his question stayed with me. Then, later that afternoon I was looking at our project management software and a notice popped up inviting me to give ‘my agent’ “a spot at the table”. Built an agent in Claude? Bring it in “as a real teammate the whole team can use” The software industry, it turns out, has already settled the question I was still mulling: AI is a member of your team now, whether you asked for one or not. I felt compelled to challenge this! So, I put this question to some people in my network:
Do you consider AI to be a member of your team?
Their opinions were strong, their arguments surprised me, and in one case the answer was found in a famous yellow cartoon character…
Why do we need a team?
Before I share their answers, I think it’s worth asking what a team is actually for. The logic behind it is simple. An individual can only accomplish so much alone. But when you bring people together to solve the same problem, the collective output far exceeds the input. As the proverb goes, ‘if you want to go fast, go alone; if you want to go far, go together’ (which you may notice is roughly the reason we’re called Go Slow to Go Fast). Anyway, somewhere around 1911, Frederick Winslow Taylor published The Principles of Scientific Management and ‘the team’ became something else: a way of dividing labour so that each person does their narrow piece of work as efficiently as possible. It worked so well that the next century of organisational design was built on it. But while that model made sense for factories, applying it to modern office work has had a strange side effect. This Taylorist mindset has in part contributed to many of the knowledge-work teams I support spending their time churning through data and process administration. Instead of solving complex problems, they often sit together in meetings just managing the system, without doing the thing they were originally assembled for, which is ‘thinking together’.
The definitive no
Some answers were unequivocal. Ewen MacPherson, Group Chief People Officer at Havas UK, wouldn’t entertain the idea one bit: “In the same way a builder wouldn’t consider a shovel a member of theirs. AI is a sophisticated and highly advanced tool, but it’s just a tool.” He added a warning about language that is worth reminding ourselves of: “Claude is not a ‘he’ but an ‘it’. Words matter”.
Rahim Hirji, author of the new book SuperSkills, which explores how to avoid "algorithmic drift" and use conscious design to stay in charge of your own decisions, went further. For him the danger is precisely that words shape behaviour. Team members earn things, he told me: trust, the benefit of the doubt, the right to be believed when they sound sure of themselves. “Call AI a member of the team and you hand it all of that on day one, unearned.” People then stop checking its work, defer to it in meetings, and let its confident answer stand. He has a term for this: “the outsourcing of judgement”.
The recanting believer
Lianne Bell, a social media strategist, used to say yes but has since changed her mind. She was the classic early adopter: she built a chatbot for her Teams channels and wrote ChatGPT into learning materials, convinced we had to embrace AI or be left behind. Four years on, she’s not so sure. She avoids certain models on ethical grounds and worries that AI is homogenising how we all think and write. But what really niggles at her is the enthusiasm. Electrician, pilot, business development manager, stay-at-home parent: everyone she asks says the same thing. They love it, they use it for everything. “Humanity is placing a lot of trust in these experimental platforms when we don’t really know what the endgame will be.” Her conclusion: if anything, a tool. And now definitely not a member of her team.
What we say and what we do
Then the answers started to be less certain, and in an interesting way. Katie Glew, Global Head of People at a software company we work with, opened with “a quick and definitive no from me!” and then, in the very next breath, described ‘AI-den’, the AI in their support platform, who handles first-line customer queries and “needs a lot of attention and support”. He is, she conceded, “very likely considered a key team player” (note the pronoun).
Paul, Managing Director of Immediate Live and an advisor to Go Slow to Go Fast, was more reflective about the gap: “No I don’t, but maybe I should!” He’d recently been introduced to a business whose organisational chart includes its AI agents, complete with job descriptions and performance measures. Like people, he noted, AI needs training and clear objectives, and gets more valuable with experience. Jess Nicastri, our Chief Project Architect, was typically matter-of-fact about it: a capable assistant that frees up her time, not a colleague.
So far the theme is: almost nobody says yes. Yet the man in Munich thinks AI will replace teams. My software thinks it has already joined one. My network says it's just a tool, then gives it pronouns and names, and a job description.
The sycophant problem
Hannah Alexander-Wright, my co-director (and wife, in case you ever pick up on a vibe), answered with a warning couched in a Simpsons reference. Consider AI a member of your team, she argues, and you’ll turn into Homer Simpson at best, or Mr Burns at worst. She notes that pop culture is full of charismatic main characters propped up by cleverer, endlessly loyal sidekicks: Smithers behind Burns, Sergeant Wilson behind Captain Mainwaring, and Dwight behind Michael Scott in the US version of The Office. The sidekick never challenges, so the main character never learns. As Homer puts it at the end of one especially memorable episode: “Marge, my friend, I haven’t learned a thing.” Rely on a team member with greater knowledge who never teachers, challenges or pushes back and you’re not leading a team, you’re just sitting in your own echo chamber.
What a team is for, now
So after all that, here’s where I’ve landed. In my opinion the man in Munich was asking the right question about the wrong team. If a team is what Taylor built, a machine for dividing labour and processing work, then no, you soon won’t need one. AI is faster at that and it doesn’t book in meetings you don’t need. But my argument is: that was never the version of a team worth having. The teams I see doing remarkable work aren’t remarkable for their throughput. They’re remarkable for their judgement: deciding what matters, noticing when the plan is wrong, disrupting each other’s autopilot behaviour, saying so out loud (ref: brain-friendly feedback), and holding each other to what they agreed.
AI is stripping away the production layer, and what’s left underneath is exactly that judgement work, which no amount of processing power does for you. In our terms, the blank sheet of paper is the go slow; the keyboard is the go fast. Skip the first and the second just produces mediocrity at speed. Teams are - thankfully - shifting from engines of production to engines of judgement, and AI is forcing that shift.
The intern
Nick Villani, CEO of Edifai, gave me the most practical answer: “Treat AI as if it was an intern. Full of energy and optimism. Eager to please. But sometimes disastrously wrong. If you are going to pass off the intern’s work as your own, you better be damn sure you’ve checked it.” That’s close to how I actually work. In my doctoral research I use Claude in deliberate stages, checking and re-prompting at each one. And when the analysis is done, I bring the emerging themes to a session with human colleagues, because meaning gets made between people. The machine drafts; the humans decide. Rahim’s closing line was that instruments don’t join teams, they get wielded by them. On the days I wield mine well, the difference is hopefully obvious: every judgement still has a human in the loop. Me.
The answer for the man in Munich
So here’s what I should have said in that queue. Yes, you still need a team, possibly more than ever. Just not for what teams were built for in 1911. The production work is leaving; the judgement work is staying, and it can’t be done alone. Whether you call the AI in the room a teammate, an intern, a shovel or a sycophantic sidekick matters less than whether the humans around it are still doing the thing they were assembled for: thinking together. The world we’re all operating in is too volatile and too ambiguous for any one person to hold the answer. The way through isn’t a better tool, it’s harnessing the different perspectives and experiences around the table until a decision emerges that is worth pursuing and which nobody could have reached on their own. If your team spends its time on anything else, that’s the part AI is coming for, and quite frankly, I think it is welcome to it.
PS: this is Claude’s response to the question (I restricted it to one sentence only): “No: a team member carries accountability, and when I’m confidently wrong, which happens, the consequences land on whoever stopped checking, which is a strange definition of ‘colleague’.”
PPS: do you get the feeling Claude was just telling me what I wanted to hear?
This is really interesting and I agree 100% with the personification issue. Because we interact with AI via more or less normal sentences and it replies in kind we treat it like a conversation, even though it isn't. Ai could be triggered in exactly the same way via a succession of drops down menus and tick boxes just like Word or Excel, the reason it isn't is that the options would need to be so incredibly complex that the menu would baffle the mind.
However if it was, even if it did exactly the same thing, would anyone ever suggest it was a team member? I agree that it takes over some of the production, but so have many products on the past, Excel can do in microseconds what statisticians used to spend all day doing but no one's ever given it a name badge, because it's framed as a tool rather than a companion. To me what we think of as AI is really a single front end that triggers a lot of different tools using a single human like inter face, and it's not more or less clever for that