Clayton Christensen once said, “If we are to develop profound theory to solve the intractable problems in our societally-critical domains… we must learn to crawl into the life of what makes people tick.” I’ve been thinking about that as I contemplate how Generative AI (GenAI) will impact education, the workforce, and beyond in the year ahead.
As we enter 2025, we’re still in the throes of the AI hype cycle. Most conversations highlight its enormous breakthrough capabilities. But how AI is being used and what tools scale is also a mirror to what makes us tick: what we as a society prioritize, and where we spend our dollars, time, and attention. In other words, the AI tools that flourish this year will reveal the true motivations of both individuals and systems.
With that in mind, while AI is a common denominator to most of the trends I’m watching this year, it’s the motivations behind those trends that merit a closer look.
1. Breakthrough assessments will try to break through. Perhaps the most exciting developments in AI for teaching and learning could come from novel assessment models that can offer formative, dynamic performance task assessments at scale. We’ve long seen that underinvestment in diagnostics, assessments, and outdated policy has prevented schools from adopting truly personalized, competency-based systems. The question is not whether new technologies could overcome that but if our policies and practices can absorb the potential of those new technologies into an entrenched system anchored on high-stakes summative tests. What makes these systems “tick” is a tricky mix of inertia and the political challenges of implementing assessments for learning and accountability. The track record of overcoming those challenges is not great. While there’s long been dissatisfaction with K-12 testing and a total lack of transparency in postsecondary outcomes, federal testing and accountability pilots have largely floundered.
As AI ushers in new possibilities, federal and state accountability systems will wrestle with if and how to allow radically different assessment approaches to break through legacy systems. While there’s unlikely a single solution, I’ll watch for states that take new strides in piloting innovative assessment and accountability systems powered by AI.
2. Self-help bots will proliferate in consumer and education markets alike. A video of Sal Khan instructing GPT-4o to tutor his son was widely circulated last year. But while that portrait of a parent, bot, and student solving a problem together was compelling, I’m not convinced tools are being built with the core assumption or hope that users will be surrounded by human support.
In a new report publishing January 14, 2025, Anna Arsenault and I look at how navigation and guidance tools are evolving in the age of AI. The early market behind guidance bots is telling: it’s evolving to solve information and advice gaps–but is much less focused on connection gaps. In other words, there’s greater demand for tools that help students help themselves (with the help of cheerful, encouraging bots), rather than connecting them to more human help (or helpful humans) at scale.
With a host of new GenAI tools offering on-demand coaching and support, we’re entering an era where we can finally overcome the human capital constraints that have long plagued college and career guidance systems in high schools and higher ed. We studied this space partly because we think the domains in education and workforce where our ratios of staff to students are most broken are where AI supports will scale fastest.
While these individual tools may produce important gains in students’ postsecondary and career journeys, they could morph into architecting a system based on lone pursuits rather than collective help. In practice, I worry that the virtues of self-help–like empowerment, agency, and self-determination–are colliding with the realities of the market where bots are likely to become a replacement, rather than a supplement, to hard-to-fund human help.
The same is happening in a macro sense in the AI copilot market. While copilots promise massive productivity gains, they seem to cater to a long-standing tradition of rugged individualism and a dearth of human mentorship that AI is set to supercharge.
I’ll be watching how efficacy research in these new tools evolves in the coming year and advocating for schools to demand tools that tackle both information and connection gaps with a longer view toward the supports and networks students need to succeed.
3. AI will start to displace our weak ties. Self-help bots are one aspect of a much larger category of AI companions gaining steam in consumer markets.
As I’ve noted in the past, AI companions are on a clear path to disrupting human connection as we know it. They cater to one of the most fundamental things that make us all tick: our deeply wired need to connect. Companions can gain a market foothold in widespread loneliness, offering a quick and frictionless alternative to seeking out human connection. But disruption is about what happens beyond that foothold: individuals who may now see AI companionship as a foreign or even absurd context will not expect the ways the technology steadily improves to meet more complex relational and social needs, optimized much like its social media predecessors to create sticky, ongoing engagement. Before we know it, many of us will have relationships with bots.
As more time is spent conversing with bots, people will invest less time in human conversation. Although that could eventually disrupt their closest friend and family networks, it’s more likely to dramatically shrink people’s weaker-tie acquaintanceship networks. Research has shown that a strong parasocial relationship can outcompete a weak human connection on dimensions like emotional support.
Again, this makes for a classic disruption story because people may not notice or miss a decline in their weak ties in the near term. But in the long term, a decline in weak ties spells less access to opportunities for individuals and less resilient communities on the whole.
With this subtle but troubling trend in mind, I’ll continue to track how tools further evolve in this direction–and vigorously advocate for more transparency and attention to safeguarding all human relationships, even our looser, seemingly inconsequential connections.
4. Colleges will be forced to grapple with the looming experience gap. While AI stands to revolutionize how schools approach assessment and support, it also puts greater pressure on what schools are expected to do to prepare students for the labor market. As education and workforce analyst and investor Ryan Craig has pointed out, entry-level jobs increasingly (and ironically) require multiple years of experience. Craig argues that even if recent graduates have the skills required for a job, employers today are looking for more. “Experience” is a vague term but is a proxy of skills applied in context. It’s also a proxy, I’d argue, for skills applied in a social context with professionals who can vouch for you.
With GenAI tools that can outcompete new graduates on a wide array of tasks, there will be even less slack in the system for early talent to perform basic tasks while they learn and are mentored on the job. In turn, we’ll likely see companies invest less in early talent and focus their mentoring energy on fewer and fewer hand-picked individuals. Whole sectors will be diligently using AI to unlock breakthrough efficiencies, all while shrinking their long-term leadership pipeline.
That’s producing an immense experience gap that employers aren’t willing to fill and colleges are ill-equipped to address, given half of graduates end up in jobs that don’t require a degree at all. I’m worried about what this means for graduates hitting the job market in the coming years, but I’m optimistic that this dynamic could force more higher ed institutions to adapt. The ROI of a degree was already in question, and these dynamics will further fuel students’ and families’ questions about the value of higher education.
Adapting won’t be easy. Even with fierce advocates and a clear need, “earn and learn” hasn’t scaled, and demand for internships far outpaces supply. Scalable work-integrated learning will hinge on colleges putting more effort into scaffolding the student experience and third-party providers willing to take on the tasks that neither education nor employers are willing or able to absorb. I’ll watch models like Colabl, Backrs, Codeplay, Mentors in Tech, and others working to build the ‘middleware’ between schools and employers to scale access to meaningful work projects and resume-worthy work experience.
5. Districts will renew their focus on whole child and economic mobility efforts. This is part prediction, part hope. As AI tools start to support districts’ immediate challenges, like absenteeism and proficiency, I’m waiting to see if schools start to demand tools that reach higher.
What makes school districts tick in 2025? Before the pandemic, many districts were operating under an ambitious north star and harnessing a broader aperture in terms of the scope of their impact–namely, supporting whole child development and working to ensure they were planting the seeds for students further from opportunity to be on a path to upward mobility. The pandemic made those broader and higher ambitions seem lofty at best and absurd at worst.
Schools–rightfully so–had to refocus on brass tack indicators like attendance, proficiency, and graduation. While the data shows we’re still far from a full recovery, the tides may be turning such that leaders can step into bolder visions of the future and move districts from surviving to thriving.
While I’m mildly optimistic that these key priorities can come into greater focus, I’m less optimistic that we’re on a path to pursuing them in the effective, integrated manner they deserve. Both whole child and economic mobility efforts aren’t entirely absent from district strategy–they are just highly fragmented, in turn diluting their reach and efficacy. If districts start to take these goals more seriously, they should follow the research: integrated student support models–that provide individualized supports to each and every student and keep educators in the wraparound services loop–are much more effective than more ad hoc supports delivered through federated, departmental structures. And in the case of promoting economic mobility, while there’s certainly enthusiasm for more career exposure and exploration, a smattering of career-related activities won’t lead to more choice-filled lives. Integrative life design, a more popular concept in higher education, should inform how schools build these pathways, allowing students to experience the world of work, reflect on those experiences, and continuously refine their sense of purpose and future possible selves. This year I’ll be watching for districts that take these more integrated strategies seriously in service of higher–and longer lasting–outcomes.
