The same organisations that rolled women out for photoshoots to prove their progressive credentials will be the first to make those same people redundant. This is a warning piece.

The same organisations that rolled women out for photoshoots to demonstrate how progressive they were — glossy annual reports, International Women's Day LinkedIn posts, diversity award submissions — will be the first to make those same people redundant. Not out of malice. Out of economic logic that was always going to arrive, and that the diversity industry spent a decade helping nobody prepare for.

I want to be precise about what this piece is. I have led recruitment drives. Coached at girl coding events. Managed engineers across every background for twenty years. I am not arguing against diversity in tech. I am arguing that the corporate implementation of diversity in tech has been a structural failure that has actively harmed the people it claimed to serve — and that the AI consolidation wave now bearing down on the industry is about to make that harm impossible to ignore.

The pipeline pointed the wrong way

The tech industry's diversity initiatives were not primarily designed to get women deeply into engineering. They were designed to get women into tech companies — and there is a significant difference. The roles that filled were product owner, scrum master, engineering manager, DEI programme lead. Roles adjacent to engineering rather than inside it. Roles that do not compound technically. Roles that do not build the craft knowledge that makes someone indispensable when an organisation restructures around a leaner, more capable core. A scrum master certification is not a technical foundation. In 2026 it is closer to a liability.

The men who followed the same non-engineering routes are equally exposed — and I want to be clear about that too. The one promoted sideways because he was too socially embedded to manage out. The one who learned enough process vocabulary to survive without ever shipping anything genuinely difficult. Steve Jobs called these people bozos and called them a cancer, and he was right on both counts. The bozo does not just fail individually. They replicate. They hire to their own level. They cannot recognise genuine excellence because they have never had it themselves. The organisation fills gradually with people selected by bozos selected by bozos, and the excellent people leave.

The difference between the male and female version of this failure is distributional, not categorical. More women were concentrated in vulnerable roles through specific initiatives. More men survived through social capital — the golf course relationship, the old boys network, the confident technical blagging that is harder to call out than agile-by-the-book process adherence. But neither group has what the coming era actually demands: years of genuine craft developed under real delivery pressure, the ability to sit across from an AI system and know exactly what to ask it, exactly where to verify its output, and exactly when to override it.

What organic diversity actually looked like

The modern corporate diversity movement has buried a history that demolishes its own assumptions. The teams that built foundational technology were meaningfully diverse — and that diversity happened without a single initiative. Ada Lovelace is the origin of programming. Grace Hopper invented the compiler and coined the term debugging. The women of ENIAC programmed the first general-purpose computer. Margaret Hamilton wrote the Apollo guidance software and coined the term software engineering itself. The Hidden Figures mathematicians were doing calculations NASA could not do without them. None of these were diversity hires. They were the best people for a job so new that social gatekeeping had not yet formed around it. The work was the only credential that mattered.

The early Macintosh team makes the point even more sharply. Susan Kare designed the visual language of personal computing — the icons, the typefaces, the desktop metaphor that made computers feel human for the first time. She was hired because she was the best person for a problem nobody had solved before, in a domain so new that no credentialing system had formed around it yet. Joanna Hoffman was kept on the original Mac team for a different but equally precise reason: she was one of the few people willing to argue with Jobs directly and make it stick. Not tolerated despite that quality. Retained because of it. The team needed someone whose judgment could not be overridden by force of personality. Neither Kare nor Hoffman arrived through a diversity programme. Both arrived through the only filter that produces lasting contribution: they were the right people for the problem and the work made that undeniable.

The early PC revolution had the same character. The Homebrew Computer Club was counter-cultural and deliberately unconventional. The hacker ethic explicitly rejected credentialism and background as qualifiers. What united those rooms was a shared intensity about what they were building. The diversity was a byproduct of selecting on that intensity wherever it existed.

The corporate version inverted the causality entirely. The original said: here is excellent work — remove the barriers stopping excellent people from doing it regardless of who they are. The corporate version said: here is an org structure — make its demographic profile look different. These are not the same project. They are almost opposite projects dressed in identical language. One produces durable, craft-grounded inclusion. The other produces the 2022-23 layoff statistics.

The CDO industrial complex

The Chief Diversity Officer role is almost self-indicting by design. No direct product responsibility. No engineering accountability. No revenue line. No delivery metric connecting to whether the company builds anything excellent. Success measured entirely by internal cultural optics and external perception scores. In the lean AI-era organisation, that role description is a definition of expendable — and everyone who designed those roles knew enough about corporate structures to know it.

Beyond the individual economics, the CDO function consumed the budget and credibility that genuine grassroots inclusion work needed. It immunised bad policy from honest critique by making any internal challenge feel like a personal attack on the person running it. And it produced representation metrics that looked like progress while the underlying pipeline — the one that would have taken twenty years and real investment to fix — went largely unaddressed.

The AI wave changes the arithmetic permanently

The pre-AI engineering organisation had enough process overhead that non-technical leadership could survive. Communication layers between business and engineering. Ceremony-heavy agile. Stakeholder coordination. Roadmap ownership. AI is eliminating the need for those functions — not partially, structurally — and what it leaves behind is a much smaller organisation where every person's judgment is load-bearing in ways it never was at scale. Deep product and technical intuition combined. The ability to make consequential architectural decisions. The knowledge to read AI-generated code and immediately identify the three places it will fail under production load. These are craft skills. They come from years in the trenches. They cannot be certified into existence.

But the deeper threat to process roles is not automation. It is architecture. The most important thing to understand about the scrum master, the product owner, the agile coach — is why those roles existed in the first place. They were not invented to build better products. They were invented to manage the coordination friction created by large teams with poor tooling and thick communication layers between engineering and reality. They were a solution to an organisational problem. The new generation of companies has eliminated the problem rather than managed it.

Linear replaced the standup. Cursor replaced the sprint review. Direct user contact replaced the product owner intermediary. CI/CD and feature flags replaced the release coordinator. Small teams with shared context and first-class tooling do not need a human layer to translate between business and engineering — because the engineers are talking directly to users, shipping daily, and seeing the results in real time. The coordination overhead that generated an entire job category has been engineered out of the system. These roles are not being automated. They are being architecturally eliminated. That is a harder problem for the people in them. There is no reskilling path from scrum master to the kind of engineer these companies actually hire. The path required starting earlier and going deeper.

The GenZ signal — and who is most at risk

GenZ engineers are widely criticised as fragile and entitled. My direct experience contradicts that almost entirely. What I observe is a generation that is genuinely eager to learn, refreshingly honest about the limits of what they know, and — critically — has a natural relationship with AI tooling as a thinking aid rather than a threat. They use it to interrogate their own code, not just to generate it. They ask why the AI suggested a particular pattern. They push back on its output. That sceptical, curious engagement is exactly the disposition that compounds in value in the coming era.

A GenZ woman with that disposition and genuine technical curiosity is precisely the person the industry should be fighting to develop right now. She is also precisely the person most at risk of being routed into a product owner role — because that is what the visible pipeline offers, and nobody has told her clearly enough what the alternative trajectory looks like in ten years, or what it looks like when consolidation hits and the product owner role evaporates with it.

The companies being built right now

The organisations that are defining the next era of software are not running scrum. Cursor reached two billion dollars in annual recurring revenue with roughly 150 people. Linear — which builds the project management tool that replaced the sprint ritual at thousands of engineering teams — is profitable, opinionated, and deliberately small. Stripe's engineering culture is built around leaders clearing their schedules to sit with engineers and complete real work alongside them. 37signals has been bootstrapped, profitable, and vocally anti-bloat for two decades, running Basecamp and Hey on a team that would fit in a single open-plan floor. None of them have a Chief Diversity Officer. None of them run a diversity award submission process. None of them have been photographed at an International Women's Day event.

What they share is selection on obsession. Cursor bans AI tools in first-round technical interviews — because programming without AI remains, in their CEO's words, a great time-boxed test of skill and intelligence. They invite finalists for two days on-site to build real things alongside the core team. Linear was built by people who found every existing project management tool an affront to craft. Stripe's operating principle — users first, always — is not a poster on a wall. It is the filter through which every engineering decision passes. These are not diverse companies in the corporate metrics sense. They are companies that will produce the next generation of technically excellent people regardless of background — if the pipeline that feeds them is honest about what they actually require.

This is the fork in the road. The consolidation wave eliminates the bloated process-heavy organisation. The people it releases either have craft depth or they do not. The ones who do get absorbed by the companies being built right now. The ones who do not face a market where the roles they held no longer exist and the roles that do exist require foundations they were never given the chance to build. The diversity question for the next decade is not whether large enterprises achieve balanced headcount in their restructuring. It is whether the craft-obsessed companies that are building everything important end up with diverse teams — or whether they default to demographic homogeneity because nobody fixed the grassroots pipeline in time.

The pipeline that was always the answer

Genuine technical workforce diversity was never going to happen in a five-year corporate initiative cycle. The journey from a twelve-year-old who takes apart her computer to understand it, through serious STEM education, through a technical degree, through five to eight years of craft development under real pressure — that is a fifteen to twenty year pipeline minimum. Every shortcut produces someone who looks like the destination but has not made the journey. The grassroots work — Raspberry Pi, Code Club, the best of Girls Who Code — planted real seeds. But it was underfunded, inconsistent, and overshadowed by a corporate version that wanted the harvest without the patience the growing required. Those seeds are still coming through. They are entering the industry now and over the next five years. They deserve honesty about what the companies being built right now actually require. Not another leadership programme. Not another scrum certification. Not another photoshoot.

Smaller organisations make genuine diversity more valuable, not less. A small, technically excellent, genuinely diverse team is objectively stronger than a homogeneous one of equal quality. Cognitive diversity at small scale is a meaningful competitive advantage — different problem framing, different pattern recognition, different blind spots corrected before they become expensive. The diversity argument was always correct. The implementation was catastrophically wrong, and the people who paid for that wrongness were not the ones running the initiatives.

The warning is this. The roles the corporate diversity movement celebrated as progress are not just being cut in the current wave. They are being designed out of the next generation of organisations entirely. The photoshoot appointments are not coming back in a different org. They are not coming back at all. And unless someone says this clearly — now, before the people currently in those roles run out of time to change course — the industry will have failed them twice. Once when it misdirected them into roles that carried false career security. And again when it watched the architecture change and said nothing.

The window has not closed. The grassroots pipeline is still maturing. The new companies are still small enough that one technically excellent hire changes the composition of the team meaningfully. The craft filter these companies run is genuinely indifferent to background — it cares only about curiosity, depth, and the inability to stop thinking about the problem. That is the diversity mechanism that has always worked. The question is whether enough people who went through the right pipeline arrive at the right moment. The corporate diversity machine spent a decade making that less likely. There is still time to make it more likely. But not much.