From Classroom to Workforce: The AI Question That Gets Overlooked

Why culture and agency matter as much as efficiency in AI implementation

Ask anyone what they want from AI, and you’ll hear the same answer: efficiency. Do more with less. Free-up time.

It’s a reasonable ambition. Yet across our work with funders, non-profits, operators, and researchers, a harder question keeps surfacing: what is AI implementation actually doing to the people, relationships, and culture inside these organizations and the systems they work within? Their sense of agency, relationships, sense of value at work? That question, we’d argue, is the one that will determine whether your AI strategy succeeds or quietly undermines itself. And the answers may dramatically affect the longer-term productivity of the organizations and initiatives you fund or invest in, too.

Well-intentioned AI implementation can quietly erode agency, relationships, and sense of purpose. Consider what’s already happening in classrooms. As AI plagiarism detection tools proliferate, teachers and students are mutually suspicious of each other. Students are working “defensively” anticipating accusations, while teachers are second-guessing work they’d otherwise trust. Researchers are beginning to document this pattern. Nobody designed that outcome. It emerged from implementation choices made without asking what they would signal.

In an education ecosystem where technology use is sometimes being banned, we need to work even harder to justify why it’s worth it; and where technology is imposed, feels bad, or compromises important human connection, the chances of success are low. The lesson is not to slow down AI adoption. It’s to be deliberate about how it’s introduced and what it’s asked to replace. As Harvard Business Review argues, when AI intermediates the small, often messy interactions that make up work relationships, we aren’t just saving time. We are giving up the moments that build connection in the first place.

This pattern isn’t confined to classrooms. Across our work in education and workforce,  we see the same dynamic surfacing in different forms.

In our research around AI in K-12, the themes are consistent. The conditions that build relational trust are human attention, presence and a sense that someone is genuinely invested in your progress. Those are the first to erode with poorly designed AI implementation. An AI grading tool saves teachers time while simultaneously signaling to a student that their work wasn’t worth a human’s attention. A workflow designed to make teachers more efficient can leave them feeling like a robot rather than the center of their students’ learning. These are patterns, and they’re largely invisible until the damage is done.

In supporting a non-profit organization to design its AI strategy, efficiency was the natural starting point. Mission-driven organizations are perpetually resource-constrained, and the case for doing more with less writes itself. But as we worked through where AI could play a role, we pushed the conversation to a different question: not what can we automate or make more efficient, but what does AI now make possible that wasn’t before? And getting an organization to genuinely operate from that reframe is not a technology challenge. It’s a behavior change.

None of this is an argument against AI adoption.

The efficiency gains are real, and education and workforce organizations that figure out how to implement AI well will have more time, capacity, and resources to direct toward the students, communities, and people they serve. But “well” has to mean more than just “efficiently.” ADP’s recent research found that the heaviest AI users reported the highest engagement, but also weaker connections to colleagues and a diminished sense of their own productivity. The two turn out to be more linked than most implementation plans acknowledge. Implementing AI intentionally means keeping a clear eye on what it signals, what it preserves, and what it might be quietly taking away.

A few places to start are…

1. Reframe the question: Instead of asking only, “Does this shrink workload?”, ask “Does this stretch what we can do?”. Be deliberate about the “burdens” you are aiming to remove because some are relational, not just operational. Think about this both in terms of task productivity and organizational culture. How can we use the technology to improve working approaches?
2. Interrogate the signal: Your AI strategy communicates something about your organization whether you intend it to or not — what you value, who you trust, how you see the people doing the work. Backing a reactive plagiarism detection rollout is one example of a signal gone wrong. You may have solved an efficiency problem while simultaneously amplifying distrust. Being intentional about that signal isn’t a soft consideration. It’s a strategic one.
3. Push beyond productivity: Human experience and organizational culture are not considerations to layer on after the efficiency case is made. They should shape the strategy, not follow it. For funders, that means asking grantees not just what AI will save, but what it will do to the people doing the work.

We often return to the obvious – but often overlooked – fundamentals that technology is just a tool, and its implementation is what matters. AI is no different, and we are already seeing the ways in which well-intentioned adoption can have unintended consequences on people, organizational cultures, and the systems in which they operate. Funders, investors and operating organizations all need to think about how this affects the people they serve, and the people they employ.

If you’d like to think about how you address these issues in your strategy from the start, please get in touch.