Why Agentic AI still needs a Pilot

Listen to Steve read this post below (8 min audio)

There’s a popular fantasy floating around right now — that soon, Agentic AI will be able to do pretty much anything for us. It’ll manage projects, build businesses, design products, write reports, send emails, answer phones, hire freelancers, anticipate what we need, and just do it before we even ask.

It’s not going to happen.
Not soon. And maybe not ever.

It’s not because AI won’t be able to — but because humans won’t want it to.

We Love Control

Humans are fussy creatures. We change our minds. We contradict ourselves. We get bored, we’re irrational, and sentimental. Just because we did something a certain way once doesn’t mean we’ll ever do it that way again.

What we love — more than efficiency, more than productivity — is agency itself.

Agency is the ability to make decisions, to change our mind, to say “No, not that way — this way.” It’s the very essence of being human. We could even argue that capitalism itself is built on maximising personal agency – and Money, well that’s the proxy. We then use it to buy this independent and choice.

This is what I’m calling FAP – the Full Autonomy Paradox.

Your newest, most needy employee yet!

Instead – AI Agents will be like hiring an assistant — only this one has a PhD in every subject, works 24/7, and never forgets a task. Sounds brilliant, right?

Until you realise it’s also the world’s most annoyingly efficient employee — I’ve never been one, but I’ve had a few! (Mostly I preferred the lazy one who was a good hang! – seriously)

This employee will constantly ask for clarification, over-deliver on things you didn’t much care for, and ping you ten times a day for “quick approvals.”

You’ll still have to onboard it, brief it, give feedback, and steer it back on course when it takes your vague instruction as gospel. In short — you’ll still be needed, maybe more.

Even if it can run the project faster than any human alive, you’ll still need to define what matters, why it matters, and when it’s actually finished. The genius of the machine only works when guided by the messiness of the human.

The Takeoff and Landing Problem

One of the smartest people I know, Nic Hodges, describes this perfectly as “the takeoff and landing problem.”

It’s a killer metaphor.

A plane can fly itself for most of the journey — but it still needs a pilot for takeoff and landing. The human has to decide where to go, when to leave, how fast to fly, and what turbulence to avoid.

During the flight, the autopilot handles the routine stuff, but when it’s time to land — when conditions change, when to use more fuel to get there on schedule, and when something unexpected happens — the pilot steps back in. AI will be exactly the same.

Even Groceries Aren’t Predictable

If you think AI will autonomously handle your life, consider this: automated grocery shopping hasn’t even worked.

It should be simple — but it isn’t. One week you love Greek yoghurt, the next week you can’t stand it. Your calendar shifts, you travel, friends visit, or you suddenly go keto.

Imagine an AI “helpfully” increasing your yoghurt order right when you never want to see another spoonful again.

That’s humanity. Predictably unpredictable.

The same goes for shoes, holidays, and investments. Just because I bought gold Nikes last time doesn’t mean I want them again. (I probably do) – Just because I went to the tropics in the past three holidays, doesn’t mean I won’t crave snow this year, or a desert road trip next. We’re creatures of mood, not pattern. One week we’re minimalist, the next we’re maximalist. Our tastes shift with the weather, our emotions, even the algorithm which guides our scroll.

Now imagine an AI trying to keep up with your imagination. It would be like a perfectly obedient assistant who never gets the memo that you’ve changed your mind.

Our desires are contextual, and ever-changing — and that’s something AI will never fully automate. Because AI is a ‘history buff’ – based on data, which is the story of yesterday.

The Illusion of Autonomy

The dream of AI autonomously buying things, organising events, or investing money for us sounds efficient — but it misunderstands human satisfaction.

We love making choices. We need to feel ownership.

Sure, I’d love an AI to do my tax perfectly and legally minimise it while I sip coffee — that’s a task with clear inputs and outputs. But most of business and life doesn’t work that way. Most of what we do involves landscape shifts, nuance, discussion, and a desire for control.

So instead of AI replacing work, it will add another employee we need to manage — just extremely capable one.

The Real Productivity Shift

Yes, productivity will rise. We’ll get more done, faster. But we won’t work less.

We’ll become orchestrators instead of operators.
We’ll become pilots — setting the destination, authorising the flight plan, and landing safely on the runway of human intent.

Because in the end, AI doesn’t know what kind of future we want — it only knows where we’ve been.

And just because we flew somewhere once, doesn’t mean we ever want to go back.

Keep thinking,

Steve.