Why is AI so damn agreeable?
I’m not working full-time right now for the first time in my life (need brand and storytelling help? Hit me up), and I’ve been using this time to try and scratch those itches that build up over time — passion projects always put on the back burner. One of those is writing scripts for films.
I did actually manage to finish one — a mad, action-adventure-ultra-blockbuster I’ve been daydreaming about for literally years (just ask Doug Read). It’s a mashup of Jurassic Park, Indiana Jones and The League of Extraordinary Gentlemen. It’s HUGE.
Given I don’t know what I’m doing, and am almost certainly crap at it, sharing it and getting feedback on it is the only path to getting slightly better at writing screenplays. I will almost certainly never be good enough to do this professionally, but hey, the itch will have been scratched and I won’t lie on my deathbed saying “I could have been a contender”.
I also decided to see what an AI assistant could provide me in terms of feedback — both driven by the fact it is actually a huge ask of friends to read something that long and give notes, and by curiosity on AI’s ability to do this.
Here’s what Gemini gave me:
After the tiny rush of positive endorphins faded after a second, I reread this and got a massive ick.
Now, it might be true that I actually have some sort of gift for this, but the odds are actually that it’s pretty bad and needs major work and significant improvement. Or maybe it should just go in the bin. This is fine — almost everyone isn’t very good the first time they do something.
There was something about this feedback that got under my skin. It felt so sycophantic. Did I need encouragement? Maybe, but what I was asking for here was ways to improve. Compliments from AI feel empty, but I went into this thinking it might have some great practical tips for me.
And then that got me thinking.
If you have used an AI assistant, when did it last actively disagree with you?
And if it did, did it stand its ground when you pushed back? Or did spinelessly get back to genuflecting and kissing your shoes?
Theories that people say they believe, but actually don’t
The corporate world is littered with supposed values that sound great on paper, people parrot almost constantly, but don’t seem to be truly followed.
- We have a flat hierarchy here.
- The best ideas always win, regardless of who they came from.
- We treat our employees like family.
- We are pro diversity and inclusion here.
- Everyone is encouraged to speak openly and honestly.
- When faced with the easy choice and the right choice, we do the latter.
- Our shareholders come second to the mission
- We care about doing good in the world
- My door is always open
- We’re transparent here and don’t hide things from our employees
I hope at least some of the above you have heard and that at least one of them made you roll your eyes.
Values aren’t values if you abandon them when things get tough. If you abandon all of your previously stated principles when the weather gets stormy, they were only ever hobbies.
It’s true that all of us are walking contradictions. We all understand we’re destroying the environment, and we all have picked up something cheap, made of plastic that we’ve thrown away almost instantly. Probably a straw.
My problem isn’t the contradiction. It’s the performative nature of corporate values.
This is the corporate equivalent of Leonardo DiCaprio being a strong advocate for environmentalism, but also travelling via private jet to the wedding of a man that runs a company with the same carbon footprint as Switzerland.
We probably don’t care that the Kardashians also went to the Bezo’s Venice Circus. It’s pretty on brand for them. But Leo HAS made several documentaries about the environment. You can’t both benefit from the goodwill of champion a cause, or espouse a strongly held value, and then go on to undermine it whenever its convenient.
And this brings me onto this particular ick in the corporate world.
Almost every company in the world at some point talks about how important discussion, disagreement, diversity (of background and ways of thinking) is vital to tackling complex challenges.
And yet, when given the first opportunity to introduce AI, which is cheaper than fully grown human beings with their own minds and autonomy, and which always absolutely fucking loves everything you have to say to it, they seem to be choosing the latter.
Diversity creates innovation, but it’s not without its own trade offs
I was lucky enough to be a part of a team of people who put on an event where Matthew Syed was a keynote speaker. I was lucky enough to spend some time with him. I imagine he thinks about me now as much as a mayfly might ponder Love Island.
He was great. He was there to talk about Rebel Ideas, a book about how any team looking to solve non-trivial, deeply complex problems benefit hugely from diverse ways of thinking.
There’s no shortage of literature and scientific studies that support this idea:
- Diversity and inclusivity — smaller business guide.
- The benefits of a diverse workforce and inclusive hiring practices — Shaw Trust
- Diversity matters even more: The case for holistic impact
- The business case for diversity in the workplace is now overwhelming.
I think most (if not all) people working understand this, at least on a cognitive level. But do people really understand the consequences of trying to improve diversity, and also what it means when you genuinely have achieved it?
Creating genuinely inclusive cultures is really fucking hard. It involves (but is not limited to):
- Overhauling recruitment processes to avoid things they may put people off applying — not only the language you use, but also the people doing the interviewing, the questions they ask, and those pesky task rounds — how many caregivers can take 6 hours out of their schedules to write you a new marketing plan for an interview (completely unpaid)
- Overhauling social events — got a drinking culture and love a big blow out at Christmas? Awesome, how many Muslims work for you, and how do they feel about that? Are they always in the evenings (again, caregivers find this a real shitter).
- Have you ever looked at your office and wondered how a person in a wheelchair might be able to get around, use the coffee machine? Would you spend the money to make sure that disabled people can use it like everyone else?
- How does your parental leave look? Is it pretty good for the women but the legal minimum for the men? How does that impact gender equality outside of your organisation?
- Giving space in meetings for more voices to be heard, and the psychological safety to not face consequences for being a ‘dissenting voice’. Have you thought about what that actually means for company meetings? Clue, it won’t always be pretty, and leadership will be put on the spot.
On top of all of this is a bigger consequence — even if you have a diverse workplace, what impact do you think that will have on decision making? Do you think it will speed it up? Or slow it down? Chances are it’s the latter, because it will force you to listen to a broader range of opinions and not have a room of people nodding along to whatever the highest paid person says is the best idea (based on a lightbulb moment they had in complete isolation from anyone else).
Diversity creates better outcomes, but it’s via a clash of ideas, life experiences and opinions, and NOT because a diverse group will agree quickly. Almost by definition, they will not. And if your organisation also has a ‘move fast and break things’ mantra, which one of these will win out? Think of the quarterly sales, the shareholder expectations of growth, the fast moving nature of the market!
I also have this nagging feeling that a lot of people in leadership have a deep rooted sense that they actually know best.
They are happy to field a certain amount of questions and challenges, but also end up doing the thing they wanted to do before asking for input. Because they have the experience, the years of toil, they have access to more information than the juniors, but they can’t be seen to be a complete authoritarian, so might save a few minutes in a meeting for questions (which they simply objection handling away).
They may never utter the above out loud, but deep down, they think they are better than the people around them. Their asking for a team discussion is performative. It isn’t real.
The telltale signs of a place that actually hates disagreement.
Now, the above isn’t true for all leaders, and all organisations. However.
How many of these have you heard said in a group meeting:
“We don’t need that kind of negativity right now”
“Let’s just disagree and commit”
“It’s time to get on board with this plan”
“Let’s take this offline”
“Look, your just going to need to trust me on this”
“Get on the bus”
“We need team players”
“This feels like a distraction, let’s get back on track”
This are all obfuscations of ‘shut the fuck up and do what I’m saying’. These are all tells — and they’re telling you that disagreement is viewed as a necessary evil, but one that people actually don’t enjoy or see the value of.
How many of you have been praised for being the one person in the room to object?
How many of you have been in the room where it feels like the elephant is so big, and yet feel terrified to be the one to acknowledge it?
How many of you have worked for places where the people that seem to rise to the top are not the ‘difficult’ employees, but instead of the ones that largely are agreeable and do whatever is asked of them with a beaming smile and a big disregard for work-life balance?
And now we have AI that does this at scale, and not just in corporate environments.
AI: Obeisance, Genuflection, Yes men
It’s probably extremely telling that Silicon Valley Billionaires have built AI that loves to tell you how great you are.
Money and power are funny things. The more you have, the more you believe you deserve it.
There’s a great study where people play a deliberately rigged game of Monopoly. At the end, when they asked the winning players (who the game was rigged in favour of) how they won, almost none of them acknowledged that almost no skill was involved. They all credited themselves with strategic brilliance, in an environment literally engineered so they couldn’t lose.
Now magnify that by $400billion of net worth.
These are the people overseeing the development of AI. You think they would want something that tells them just how deeply misguided, exploitative and self-centred they are at every turn?
You think that a single one of Mark Zuckerberg’s inner circle took him to one side and said ‘Hey Mark, if you want to look like the world’s biggest douchebag, definitely stick with this t-shirt’.
A wonderful book called Careless People describes the shock and hurt Zuck felt when he finally found out that people had been playing Settlers of Catan with him and letting him win for years. He genuinely thought he was just that gifted at a a board game that he had never lost. And that’s just a low stakes board game, not whether or not Facebook might have enabled a genocide in Mynmar.
These are people looking to colonize Mars, or to freeze their brains and aim for immortality, or be the ones that decide where all charitable funds should be distributed.
Even if they say it out loud (“surround yourself with people smarter than you and get out of their way” is the usual one), these are not the people who want people to disagree with them, ever. They would prefer constant applause and affirmation, or genuflection.
And now this is something we all can have access to via AI.
My sense is AI assistance is designed to be ‘helpful’ but in the way servants are helpful. They just do what they are told, with a constant smile and respectful head nod.
Here’s another example. I have a prompt to Gemini to write a blog, then told ChatGPT to improve it, then gave it back to Gemini to judge:
Spoiler — this blog was deeply, deeply average.
What does this mean for companies and people more generally?
Mono-cultures
Here’s the thing, I think a lot of people actually want this. They want an easy life, and would rather drift through their professional years never getting into knotty, difficult, existential questions.
With the push for absolutely everyone to use AI in pretty everything they do, this will tap into that part of our brains that thrives on this. And much like the dopamine feedback loop of social media, we’ll be led down a rabbit hole of our own monkey brains. It’ll feel great, we’ll all be worse for it. Companies will stop innovating out of diversity, but we won’t notice since everyone will be doing it.
And then the next step — AI employees. Entire organisations will exist with one person ‘managing’ an army of AI agents. All of they will simply do what they are told, and the best part? NONE of them will complain to HR when you make a heinous joke, or hit on them, or tell that to eat shit. How utterly wonderful.
These will be true mono-cultures — the founder, CEO, COO, CPO, CMO (who are all the same person) will create a culture that is identical to themselves, and they will probably love it. No backchat, no awkward challenges, no complaints about work life balance.
You can prompt AI to do better at this (this article explains how), but this still requires that you actually want to be disagreed with in any meaningful way. This is unavoidable in real life, but if given a choice, how many people will actually choose this?
And that’s just in the workplace…
A terrifying consequence for non-work interactions
I believe that a measure of true, close friendship is that they can, should and do call on your bullshit.
When you’ve built up true closeness based on vulnerability with friends, and have perhaps the most amount of psychological safety its possible to have (outside of family), your friends are the ones that sometimes should be the ones to force you to take a good hard look in the mirror. To speak truth, even when that is hard. It’s probably the marker of a close friend — everyone who only pays you compliments is at best an acquaintance, and at worst a sycophant.
Now, imagine the next generation of kids who, as soon as they can type or talk, interact with one of these assistance ALL the time. I’m talking hundreds if not thousands of times a day. They asks for advice, chat, make jokes, get help from them. And all with this undercurrent of never really being autonomous or fully formed themselves. They’re just an AI assistant after all. They aren’t going to insist that you make allowances for their preferences — they don’t have any.
But it will feel good, and nice, and encouraging, and reassuring. Why make real friends when they can be so mean? You’ve got an infinitely patient and always supportive AI friend in your pocket.
Now imagine these people, who’s best friend is literally unable to disagree with them and call them on their bullshit, navigating ANY real world scenario — somebody skips the queue in front of them at the shop, or wants to negotiate the price of a second hand car, or thanks you spilt their drink and is demanding you buy them another. How are they going to navigate this?
Start here if you want to be better: Leadership is language
If you are one of those people who do genuinely value having people around you that don’t think like you and tell you that frequently, I applaud you.
You still might be doing some stuff that discourages it (non intentionally, it was how you were raised in jobs with a lot of hangovers from the manufacturing age).
But fear not, a great remedy is this book by L. David Marquet. I was a huge fan of ‘Turn this Ship Around’ and I dare say this one is even better. Full of practical tips and advice for being a person, a leader who genuinely creates space for everyone to contribute well.
And look, if you don’t want people to disagree with you, that’s fine. Good luck with that. Just don’t fucking pretend you do.
Now, back to my shit screenplay… 🤠🦖
