The Adoption Problem Nobody Wants to Talk About

This graduation season, students across the US have been booing AI off the stage. Eric Schmidt got booed at the University of Arizona, a real estate executive got booed at UCF, and a music exec got booed at Middle Tennessee State. According to recent polling, 42% of Gen Z say AI will harm their job opportunities—more than any other generation. But when Jensen Huang talked about AI at Carnegie Mellon, acknowledging the disruption while focusing on what it makes possible, nobody booed. Same topic, completely different reaction, which tells you that the messenger and the framing matter at least as much as the message itself.

Adeline: This isn’t people rejecting technology, it’s people rejecting the way technology is being sold to them, as an inevitability they should be grateful for, by people who won’t bear the costs of getting it wrong. I understand the frustration, and I agree that some people will lose their jobs. But this kind of disruption has happened before, and it’s also a huge opportunity to get ahead by mastering skills that not many people have yet, precisely because the tool is so new. For me, the deeper discontent is that the benefits of AI will be concentrated in the hands of a few—but that’s a different conversation, and one worth having separately.

Here’s the adoption paradox in one stat: 82% of enterprise leaders say they offer AI training, and yet 59% still report an AI skills gap. Only 32% of individual contributors say they even have access to AI tools, compared to 80% of the C-suite. Leadership thinks training is happening while the people doing the work say it’s irrelevant to what they actually do all day.

Suzanne: The reason is pretty simple—nobody asked the frontline staff what problem needed solving before buying the solution. Organizations are throwing tools at people and expecting them to figure it out, and even when the tool would genuinely help, people are nervous about AI, they don’t know how to use it properly, and they want to keep the human in the loop. That’s a phrase we hear constantly in our work with government agencies.

Adeline: The attitude I’ve encountered at the European Parliament and at the ESM is similar but has a specific flavor. Some officials genuinely don’t believe AI can do as good a job as they can, and beyond that, they’re afraid of losing the ability to think. They don’t want to outsource drafting emails to AI because they don’t want to lose their writing skills. Richard Susskind makes the point in How to Think About AI that we’ve long accepted machines outperforming us physically—it’s the intellectual competition we resist. EU public sector AI adoption lags behind global peers, and French public servants largely don’t expect AI to improve their efficiency. Different system, same disconnect.

And the resistance isn’t just a matter of sentiment anymore—it’s showing up as concrete action on both continents. In the US, 360,000 Americans have joined Facebook groups opposing data centers, voters in Festus, Missouri ousted four city council members after they approved a $6 billion data center project, and nearly half of all planned data centers in 2026 have been delayed or cancelled. In Europe, more than 25 data center projects were blocked or delayed in January alone, with rural communities packing town halls over electricity prices, water use, and tax breaks handed to Big Tech. People aren’t rejecting AI as a concept—they’re rejecting being told it’s inevitable while they bear the costs and see none of the benefit. That’s not a technology problem, it’s an adoption failure.

Suzanne: Here’s the part that doesn’t get enough attention: in government, the AI-and-jobs conversation is already moot. No matter what you agree with politically, agencies have lost significant headcount and the work hasn’t gone away. This isn’t about fear of losing jobs—the people are already gone. Meanwhile, federal agencies are rushing to deploy AI while sitting on years of governance debt, from naming conventions that have never been standardized to retention policies that aren’t enforced to roles and responsibilities that exist on paper but not in practice. The remaining workforce needs help handling the load, and AI isn’t a threat to their jobs—it’s potentially the only way they survive the workload. But only if it’s deployed in a way that actually helps them do their work, not in a way that adds one more tool to learn on top of everything else.

Neither side of the Atlantic has figured this out. The US approach is to deploy first and fix later, prioritizing speed over readiness, and the backlash is catching up fast. The EU approach is to regulate first and deploy cautiously, but even that isn’t working—the AI Act just got a 16-month enforcement delay after companies like ASML, Airbus, and SAP warned that Europe was regulating itself out of the race.

Adeline: And within EU institutions themselves, the pattern is even more cautious: many won’t deploy tools until they have the full governance plan in place, which means they’re still planning while the work piles up.

The US moves too fast for its people, and the EU moves too slow for its ambitions.

Suzanne: So, what actually works? In our experience, it’s not a framework and it’s not a mandate from the top—it’s a mindset shift. Instead of rolling out AI as a capital-T Transformation, you start with a specific pain point, something concrete that real people are struggling with today, and you solve that first.

We think of it as crawl, walk, run. Crawl means finding the pain point and starting with a small group of super users—people who are excited, who’ve already brought use cases to the table—and delivering a win within days rather than weeks. Something simple, like a summary that used to take hours or a status indicator that required digging through five different systems. The goal is to make it undeniable that this actually helps the people using it.

Walk means expanding outward once those early users are bought in. You add structure, automate the workflows around the quick wins, and start connecting data sources. The super users become your evangelists, not because anyone asked them to sell AI, but because they’re showing their colleagues that it made their work easier in ways they can point to.

Run means you’ve earned the right to talk about the bigger stuff—predictive analytics, risk identification, executive dashboards—not because you started there, but because you built a foundation of trust and utility that makes mission-level intelligence possible.

You can’t boil the ocean your way to adoption, and you can’t mandate it from the top and expect people to thank you. You have to start where people are, solve what they actually need, and earn your way to the bigger vision. The question isn’t whether AI will be adopted—it’s whether anyone will bother to do it in a way that people actually want to use.