Culture Is Infrastructure: Why AI Won't Let It Run on Autopilot

What Amazon's Tonya White Hallett understands about scaling fast without losing your human core

Infrastructure is invisible. Until it fails.

You don't think about the wiring until the lights go out. You don't think about the foundations until something cracks. And most leaders don't think about their culture until a wave of resignations, a failed integration, or a brutally honest survey forces them to.

That was my first thought reading a recent ETHRWorld interview with Tonya White Hallett, who leads People eXperience and Technology for Amazon's worldwide consumer business. Her argument is deceptively simple and, I think, exactly right: in the age of AI, culture is no longer a soft layer you can bolt on. It's infrastructure. And it can no longer be left to run on autopilot.

I've spent my career making a version of this case. So let me tell you why she's right, and why it matters even more than most boardrooms have realised.

Culture doesn't hold itself up anymore

For years we filed culture under "soft." The values on the wall nobody can recite. The off-site. Something HR owns and the business tolerates.

Hallett's point is that this was always a category error, and AI has made it an expensive one. Companies inherit values from their founders, she notes, but in an environment reshaped by automation, hybrid work and post-pandemic disruption, culture stops sustaining itself automatically. What worries her most is the gap between the culture organisations say they want and the one their people actually live every day.

Here is the line that stayed with me. Culture, she says, is not what you put on a slide or even what you design. It is "what people consistently experience when no one is watching."

I couldn't agree more. In my book I call this the smell of the place: the thing people feel in the first ten minutes and can never quite unfeel. You can't see it on the org chart. But every decision, every system, every interaction runs through it.

AI doesn't create the cracks. It scales them.

Here's the uncomfortable part. AI is not a culture problem. It's a culture amplifier.

Hallett makes a sharp observation: employees increasingly experience their organisation through digital systems, the learning platform, the support tool, the feedback app. When those systems feel transactional or disconnected, the culture starts to feel that way too. At Amazon's scale, she says, complexity outruns clarity unless you actively simplify, and if you don't, you dilute both culture and performance.

Read that again, because it's the whole game. Drop powerful AI into a healthy organisation and it compounds what's good. Drop the same tools into one running on autopilot, with unclear purpose, absent leaders, and a workforce that stopped trusting the last survey, and AI doesn't fix any of it. It accelerates it, at machine speed. That is the human cost hiding inside the race to adopt.

Empathy is not "soft." It's a performance multiplier.

The part of the interview I'd put in front of every executive is her reframing of empathy: not as a nicety, but as a multiplier. When people feel genuinely seen and valued, Hallett argues, they invest differently. They take risks, cover for each other, and give the discretionary effort no incentive scheme can buy.

This is the EX to CX to Growth logic I keep coming back to. Engaged people create delighted customers, who create growth, in that order. AI changes the speed of every step. It does not reverse the sequence. Automate the customer experience while the employee experience quietly corrodes, and you're filling a leaking bucket. Very efficiently.

And crucially, she insists empathy can scale, but only when it is built into systems rather than left as an abstract value. Listening continuously. Giving leaders honest visibility into how their teams actually experience their leadership. Not as surveillance, but as a learning discipline.

From advisor to builder

There's a challenge in here for my own tribe, the people function. Hallett thinks the debate about HR "earning a seat at the table" is already over, and outdated. The real shift is from support function to business builder: people leaders who understand the operating model as deeply as any commercial executive and connect talent decisions straight to customer outcomes. Her framing is blunt: the people function shouldn't support the business; it should be the business. That takes courage, the willingness to carry uncomfortable truths into the room rather than observe from the side of it.

The future of work is still human

Asked to describe the future of HR in a single word, Hallett didn't reach for a technology. She said: human.

That's the right answer, and it's the one I'll keep defending. AI will transform how organisations operate. But culture, the infrastructure you can't see, will still decide whether your best people want to build the future with you, or somewhere else.

Infrastructure is invisible until it fails. The winners of the next few years will be the ones who maintained it on purpose, while it was still quiet.

Source: ETHRWorld EMEA, "Culture can no longer run on autopilot": Amazon's Tonya White Hallett on the human cost of the AI race, interview by Yasmin Taj.

This is the heart of my book, Employee Engagement, What Else?, because culture doesn't live in your AI roadmap or your values statement. It lives in the daily reality of work. And in the AI race, that's exactly where it's about to be tested.