The Biggest Threat to the Future of Work Isn’t AI. It’s Unethical AI.

The biggest threat to the future of work isn’t artificial intelligence. It’s artificial intelligence (AI) that quietly learns our worst habits and convinces us they’re objective.

That’s exactly what happened when Amazon abandoned an experimental AI recruitment tool after discovering it had learned to penalise CVs containing the word “women’s”. The system hadn’t suddenly become biased. It had simply become an exceptionally efficient student of yesterday’s decisions. It was trained on ten years of historical hiring data in a male-dominated industry and concluded that male candidates were preferable because history appeared to say so.
That story should concern every CEO, HR leader and board member embracing AI.

Because Amazon’s algorithm wasn’t broken. It was doing exactly what it had been taught to do. And that raises a far more important question than whether AI will replace jobs.

What exactly are we teaching our algorithms?

For years, we’ve been told that data is the new oil. I see it differently. Data is memory. It remembers every hiring decision. Every promotion. Every overlooked candidate. Every unconscious bias. Every brilliant decision. Every flawed one.

Artificial intelligence doesn’t arrive with values of its own. It inherits ours. If yesterday’s workforce decisions contained bias, AI doesn’t eliminate it. It scales it.

As mathematician and data scientist Cathy O’Neil argues in Weapons of Math Destruction, algorithms are not neutral simply because they’re mathematical. When opaque models are deployed at scale to make high-impact decisions, they can quietly reinforce inequality while creating the illusion of objectivity.

That may be the greatest misconception surrounding AI today.

We assume machines are objective because they don’t have emotions. But algorithms don’t create culture. They preserve it. An unfair hiring manager might reject dozens of candidates. An unfair algorithm can reject millions before anyone notices.

AI isn’t creating bias. It is giving old bias a new level of speed, scale, and credibility.

That is why the conversation around artificial intelligence needs to change. For too long we’ve been asking whether AI is intelligent enough. We should be asking whether it is ethical enough.

The dangerous myth of objective algorithms

One of the biggest myths surrounding AI is that computers make unbiased decisions. They don’t. Every AI model reflects the data, assumptions, and objectives provided by humans. If those inputs contain historical inequality, AI learns inequality. If they contain fairness, AI can help reinforce fairness.

Technology is a mirror before it becomes a decision-maker.

The Amazon case demonstrated exactly that. The algorithm didn’t invent bias. It simply learned from historical recruitment patterns and concluded that they represented success.

The uncomfortable truth is that algorithms rarely create new problems. They expose existing ones. That’s precisely why governments around the world are paying closer attention to AI used in employment.

The European Union’s AI Act classifies AI systems used in recruitment, hiring, and workforce management as high-risk, recognising that automated decisions can significantly affect people’s careers and livelihoods. Organisations that are deploying these systems are expected to demonstrate transparency, human oversight, and robust risk management.

Similarly, the UK Government’s guidance on Responsible AI in Recruitment encourages employers to ensure AI supports fair, explainable and accountable hiring rather than replacing human judgement.

These developments are encouraging. But regulation alone won’t solve the problem. Ethics cannot simply be legislated. They must be designed.

Technology should challenge bias, not automate it

When we started building TFY more than a decade ago, we deliberately rejected one assumption that had become fashionable across much of the technology industry — that the purpose of AI is to replace human judgement. I’ve never believed that. Human judgement is imperfect. But removing humans from important decisions isn’t the answer. Helping humans make better decisions is.

Technology should challenge assumptions, highlight patterns, present evidence, reduce repetitive administration, and reveal opportunities people might otherwise miss. It should never become the unquestioned authority deciding someone’s career, livelihood, or future.

The future of workforce technology isn’t about replacing recruiters. It’s about giving recruiters better information so they can make better decisions. That’s a fundamentally different philosophy.

Opportunity is the metric that matters

Building TFY without venture capital gave us something I have always valued more than rapid growth: time. Time to question assumptions. Time to listen to customers. Time to optimise for trust rather than quarterly expectations. Time to ask a different question.

Not simply: “Can we automate this?”

But:“Should we?”

Those are two very different conversations.

Today’s AI race often celebrates larger models, faster inference, and greater automation. Those achievements matter. But I believe the companies that will define the next decade won’t necessarily build the most powerful AI.

They’ll build the most trusted AI.

Because trust is rapidly becoming the most valuable competitive advantage in business, customers increasingly want to understand not only whether organisations use AI, but how those systems influence decisions affecting people’s lives. That expectation will only grow.

Technology worthy of humanity

Over the years, I’ve come to believe that the real question isn’t whether technology is becoming more intelligent. The real question is whether we’re making it more worthy of the people it serves. Technology worthy of humanity doesn’t replace human potential.

It expands it. To me, that means three things:

First, AI should augment human judgement, not replace it.

Second, it should be transparent enough for people to understand, question and improve its recommendations.

Third, it should expand opportunity instead of reinforcing historical inequality.

Those principles apply equally to recruitment, workforce management, education, and entrepreneurship. They also apply to every founder who is building the next generation of AI. Because every algorithm inherits something from the past. Some inherit knowledge. Others inherit prejudice. The difference depends entirely on what we choose to teach them.

Every generation inherits a defining business challenge. The Industrial Revolution challenged us to protect workers. The Digital Revolution challenged us to protect data. The AI Revolution challenges us to protect human judgement.

The future doesn’t need more artificial intelligence. It needs more intelligence about how we use it. Because technology earns its place in society only when it leaves people with more opportunity than it found them with.