Most leaders I speak with are quietly certain their AI systems are neutral. The machine runs on data. Data is just numbers. Numbers don’t discriminate.
Tracey Spicer AM spent years finding out how wrong that assumption is.
Every AI tool running in your business carries within it the assumptions, blind spots, and opinions of the people who built it. And the people who built most of the AI tools now operating across global business are, overwhelmingly, a small, similar group writing code in a handful of cities in the Northern Hemisphere.
Tracey Spicer AM put it plainly when we spoke on The Wisdom Of … Show: “An algorithm is not just a mathematical equation. It’s an opinion written in code about how the world should work.”
That opinion is now running your hiring process, your credit assessments, your customer recommendations and your medical triage systems, in some organisations.
The question worth asking is not whether that opinion contains bias. It does. The question is whether your organisation has the integrity and the structure to do something about it.
Thirty Years Of Watching
Tracey Spicer AM has spent three decades in Australian public life. She’s a multiple Walkley Award-winning journalist across the ABC, Network 10, and Sky News. A Sydney Peace Prize recipient alongside Tarana Burke for her leadership of the MeToo movement in Australia. An Order of Australia recipient. The author of Man Made, winner of the Social Responsibility category at the Australian Business Book Awards. And the creator of a TEDx talk, “The Lady Stripped Bare,” that accumulated 7 million views by doing something almost no public figure in her position had done before: walking on stage in full broadcast makeup, hair, and heels, and taking it all off while making a case about what that invisible weight costs.
She has experienced a lot of trolling. She has also received thousands of messages from women saying they finally felt free to leave the house without blow-drying their hair. Both responses tell you something important about what happens when someone puts a genuine idea in front of a real audience.
That quality, stepping into the thing rather than describing it from a distance, is something I spend a lot of time on in my Masterclass. The leaders who create real change are rarely the ones with the most information. They’re the ones who can make an idea so visible, so concrete, that a room full of people who weren’t looking for it can suddenly see it.
Where The Book Began
Man Made did not start in a research library.
It started with Tracey’s 11-year-old son saying, “Mum, I want a robot slave.” He’d been watching South Park. Cartman was ordering his Amazon Alexa around. And in that moment, Tracey noticed something she couldn’t un-notice: the submissive, obliging voice assigned to female chatbots. The authoritative, strategic voice given to male ones. A deliberate design choice, because products that appeal to our deepest instincts sell better.
She started looking into it. What she found was not a fringe problem.
Self-driving cars whose light-sensor technology was calibrated for white skin, failing to recognise people of colour at pedestrian crossings. Hospital algorithms that give the available ventilator to the younger patient, because the system reads age as productivity. Amazon’s hiring tool, decommissioned in 2018, was sophisticated enough to infer from a woman’s listed hobbies, specifically that she’d played netball or basketball, that the CV came from a woman. Then move it aside. No human reviewed the output. The system ran.
“What I found very quickly was that it wasn’t just the gender factor. There were multifaceted biases built into the technology we use every day.”
These are the predictable outcomes of systems built without a representative range of human voices in the room when the building happened.
Performative Versus Effective
One of the sharpest observations Tracey makes is the distinction between organisations that express concern about bias and those that take structural action against it.
Three years ago, when Man Made first came out, the public commitments were there. The systemic changes largely were not.
What has shifted: regulation – Europe’s AI Act. New York legislation requires annual bias testing for hiring algorithms in companies with over 100 employees, covering women and people over 50, with financial penalties for non-compliance. When there are consequences, behaviour follows.
The organisations leading rather than just complying, Tracey says, don’t start with the regulation. They start with the strategy: where does AI genuinely serve the business, and what does building that with integrity actually require?
That reframe, from AI bias to AI integrity, is what shaped the model we built together in the episode.
Watch the full conversation and see the model built live
Join Simon’s Masterclass to apply frameworks like this in your own organisation
The Ai Integrity Model
In the episode, I built a visual model with Tracey across three components: data sets, algorithms, and machine learning. The structural principle is this – a triangle holds because all three sides carry the load. Remove the integrity from any one side, and the system doesn’t wobble … it fails.
Data sets. The starting point is cleaning. Garbage in, garbage out. Most historical data sets lean toward men, toward white populations, toward city-dwellers, toward the young, toward people without disability. That lean reflects who got counted, photographed, surveyed, and quoted over the preceding decades. Cleaning it is painstaking, expensive, and usually skipped. Beyond cleaning, the second step is intentional: actively correcting for the gaps the existing data has created, leaning toward closing them rather than reproducing them. The third is managing size. Curated, controllable data is more useful than vast, unmanaged data with integrity you cannot guarantee.
Algorithms. Who was in the room when the algorithm was designed? If three-quarters of the human population were absent from that conversation, the solution would not serve three-quarters of the market. A commercial observation, not a political one. Tracey accepted the reframe I offered here – the goal is bias consciousness rather than bias training. A workshop produces a behaviour for a season. Consciousness permanently changes how people see and interrogate systems. Add to that continuous testing, because machine learning means an algorithm clean at the start of the year can drift by the end of it, and no one notices until the harm has already compounded.
Machine learning. Human oversight before any output reaches your client. The Amazon hiring tool story makes this concrete – the machine was sophisticated enough to infer gender from hobbies even after all explicit identifying information had been stripped. No human reviewed the output. It ran unchecked until someone finally examined the results.
What Metoo Taught Her About Changing Minds
The conversation moved at one point to something directly relevant for any leader trying to drive change inside an organisation comfortable with the status quo.
How do you open a conversation that people are primed to resist?
Tracey’s answer came from years of speaking about MeToo, first to rooms full of women, then to rooms full of men.
“When you are a journalist, you’ve got two ears and one mouth. It’s more important to listen than it is to speak.”
She didn’t lead with the argument. She led with the audience. Who’s in this room today? What are they carrying? How do you speak in their language rather than the language that comes naturally to you?
For male audiences, she shifted to examples of bias that landed differently: young men paying more for car insurance because historical data coded them as worse drivers; men in high-harassment workplaces also face elevated rates of bullying. Shared language, reached from both sides of the room, held longer than any argument.
“You can’t just talk about examples at one polarised side of the debate. You have to talk about it on both sides. So there’s some kind of shared language around it.”
Any leader trying to change a culture, a system, or a room should sit with that.
The Pause
Right at the end of our conversation, I asked Tracey for her personal philosophy. The one she’d pass on if she could.
She said she learned it the hard way.
“Take a breath before you say something, or do something, or decide something.”
That is something that I strongly believe in. In sport, the pause is a weapon. The head fake draws the defender. The stillness before the shot creates the space. But technology has systematically removed the natural pauses that used to exist – the walk between the desk and the phone, the overnight gap before the letter got a reply, and the absence of instant send. We receive the inflammatory email, hit caps lock, and fire.
Turn off automatic send-and-receive. Reread before you click the manual send button. You might stop a few you’ll later regret.
The same principle applies in AI governance. The human-in-the-loop system is, at its root, an institutionalised pause. A deliberate decision to put a human moment between machine output and real-world consequence.
It sounds simple. Almost no one builds it in from the start.
Why This Matters Now
The best leaders I know have stopped asking “How do I implement AI faster?” and started asking “How do I build AI I’m prepared to stand behind?”
Those produce very different systems.
Tracey Spicer AM has spent years documenting what happens when velocity wins over integrity in the design of AI systems. The costs are measurable – talent excluded from pipelines, clients poorly served by products not designed with them in mind, regulatory exposure growing in every major jurisdiction, and, in the most serious cases, people’s lives shaped by a machine running an opinion no one checked.
She is also, and this is what I find most striking about her thinking, genuinely optimistic about what is possible when the design decisions being made right now are made differently.
We raised a toast near the end of our conversation … to a future where our children and their children live in a world of integrity, fairness, and equity, where everyone has real opportunities, and no one is moved aside by code written a decade ago.
“It sounds like a pipe dream. But it would be wonderful to work towards that.”
That work starts in the decisions being made inside your organisation right now, around your data sets, your algorithms, and who is in the room when both are being built.
Watch the full conversation with Tracey Spicer AM on The Wisdom Of… Show now
Join Simon’s Masterclass to apply frameworks like this in your own business
