Redefining progress in the age of AI
In the last part of this series, I shared four ideas that could help reshape how AI is built, from a Fellowship to an advocacy network. Each of them was about doing something different in practice: changing who is in the room, how harm is named, and where agency sits.
But as the workshops unfolded, another theme kept coming back, one that sat underneath everything else we talked about. It was less about what we were building and more about how we would know if it was working. In other words, how we define progress in the first place.
Because even the most inclusive systems will fail if we keep measuring success with the same metrics that created the problem.
That realisation pushed this project into deeper territory. It wasn’t just about how we design AI, it was about what we call “good”, whose experience counts as evidence of that, and who gets to decide when progress has been made.
The limits of traditional measures
One of the most revealing moments in the workshops was when we began to map out how AI systems are currently assessed. The conversation kept circling back to the same set of benchmarks: speed, accuracy, efficiency, scale. They were treated as self-evident indicators of progress, the default markers of whether something was working.
But the more we talked, the clearer it became that these metrics often say more about the priorities of the people building the system than about the people living with its consequences.
Speed tells you nothing about safety. Efficiency tells you nothing about dignity. Scale tells you nothing about fairness. And yet these are the measures that dominate our understanding of what successful AI looks like.
Participants shared stories that made this gap visible. Someone described how an AI system used by their employer reduced processing time but created new barriers for anyone who didn’t fit neatly into predefined categories. Someone else spoke about safety tools that could flag content quickly, but rarely understood context, leaving certain forms of abuse untouched.
When we rely on metrics that prioritise speed or scale above all else, we end up validating systems that reproduce harm more efficiently. We optimise the wrong thing.
And this isn’t unique to AI. It mirrors what happens in workplaces too: when leaders measure output without context, performance without conditions, results without equity. The numbers improve, but the experience does not.
The question isn’t whether these traditional metrics are useful – sometimes they are. The question is whether they are enough. And increasingly, the answer is no.
What fairness looks like when we measure it differently
As the workshops deepened, another set of measures began to surface, not as abstract ideals, but as reflections of what people actually needed from the systems shaping their lives. These weren’t framed as replacements for traditional metrics, but as the parts we overlook when efficiency becomes the only story we tell.
Equity came up first a practical question: does this system work fairly for people with different backgrounds, needs, and identities? And if not, who is bearing the cost of that gap?
Safety followed closely behind. Participants spoke about wanting systems that didn’t just avoid harm in theory but recognised harm in practice. Safety wasn’t about eliminating all risk but not amplifying the risks that already exist.
Dignity was harder to define, but everyone understood it. It showed up in the desire to be spoken to in human language. To be recognised by systems without having to contort yourself into categories that don’t fit. To not have to prove your legitimacy just to be seen.
And then there was power – who holds it, who feels it, and who is quietly excluded from it. Many participants spoke about moments where technology made them feel small or invisible. Redistributing power wasn’t about giving everyone control over everything; it was about ensuring that the people affected by a decision weren’t the last to know about it.
And when you apply these principles to leadership, the parallel becomes almost impossible to ignore. Organisations also measure speed, efficiency, and scale. They track outputs and deadlines. They optimise for predictability. And in doing so, they often miss the quieter signals: who feels safe to speak, who feels valued, who feels visible, and who feels unsupported.
The metrics we choose shape the cultures we build. And cultures built on narrow definitions of success eventually reveal their limitations through the subtle erosion of trust, belonging, and potential.
What leaders can learn from this
As we explored these alternative measures of progress, it became clear that the questions we were asking of AI were the same questions leaders should be asking of their organisations. Systems don’t become extractive or exclusionary by accident. They get that way when speed is rewarded over understanding, when efficiency becomes more important than impact, when the people affected by decisions aren’t the people making them.
And that pattern isn’t unique to technology, it shows up everywhere decisions are made without the full context of the people they affect.
In the workshops, participants kept returning to the same tension: "What happens to the people at the edges when progress is defined only by those at the centre?"
It’s a leadership question as much as it is a design one.
When leaders focus only on outputs, they inevitably miss the quieter indicators that something isn’t working: the hesitation in a meeting, the idea that never gets voiced, the team member who has learned to adapt rather than to contribute. These moments rarely appear in dashboards, but they shape performance more than any KPI.
The alternative we explored in the project – equity, safety, dignity, power – offers a different way of seeing leadership. Not as something performed from the top, but as something co-created with the people who live the consequences of the work. And it’s not abstract.
A system that respects dignity functions differently. A team that feels safe will take risks they wouldn’t take otherwise. A culture that distributes power gives you better decisions, not just faster ones.
These principles aren’t “soft”, they’re structural. They determine whose ideas surface, whose concerns are named early, and whose insights shape the outcome. They reveal whether people are participating because they believe their voice matters, or because they’ve learned that resistance will just slow things down.
What this project surfaced – again and again – is that inclusion works best when it isn’t treated as an initiative but as an operating condition. When it shows up not just in who is in the room, but in how decisions are made once they’re there.
Leadership, at its core, is a design discipline. And the systems we design will always reflect the assumptions we start with.
A future built with intention
As the project came to an end, one question stayed with me more than any other:
“If we had the chance to rebuild our systems from the ground up – not just technologically, but structurally – what would we choose to protect, and what would we choose to let go?”
The future of AI isn’t only a technical question. It’s a values question. A power question. A leadership question.
What struck me throughout this work wasn’t the scale of the problem – we know the problems – but the clarity of what’s possible when different voices shape the answers. Bias became visible, gaps became design opportunities, and people who had been positioned as “users” became co-creators of the future.
When you see that happen in a room full of diverse perspectives, it’s hard not to imagine what it could look like at scale: systems that don’t just avoid harm but actively expand possibility; technology that doesn’t flatten people into categories; leadership cultures that prioritise understanding before optimisation.
A future like that starts with decisions made in the earliest stages of design. It starts with who gets to ask the questions, who gets to challenge the assumptions, and who gets to define what “good” looks like.
And it starts with leaders who are willing to slow down long enough to notice the cost of moving fast, who understand that progress measured only by speed will always leave someone behind.
If there’s one thing this project has made clear, it’s that the systems we inherit are not the systems we have to keep. We can choose different starting points like fairness, dignity, agency, accountability, and build from there. We can design organisations and technologies that don’t require people to fit into narrow definitions to be recognised or protected.
The future of AI will shape the future of leadership. And the future of leadership will shape the systems we all live in.
The real opportunity – the one that sits underneath every workshop, every insight, every difficult conversation – is this: we get to decide what we build next. Not by perfecting systems, but by widening who gets to define them. Not by chasing efficiency, but by designing for humanity. Not by treating fairness as an outcome, but by making it the foundation.
This series may end here, but the work doesn’t. If anything, this is the beginning, the point where awareness turns into responsibility, and responsibility turns into the decisions that shape the future all of us will inherit.
Until next time,
Tania