Editor’s note: This is the 17^th article since May 20, 2026 in an ongoing series by Dr. Andrew Maxwell, the Bergeron Chair in Technology Entrepreneurship in the Lassonde School of Engineering at York University. Every week – and occasionally every other week – we’ll present a new article by Maxwell, in a series whose wide-ranging and incisive themes encompass: Canada and innovation policy; productivity and industry; innovation frameworks; AI and higher education; research and intellectual property; technology adoption; entrepreneurship and commercialization; universities and higher education; entrepreneurship education; and AI and the future of work. 

 This is Part 5 of a five-part series looking at the future of universities: Why are universities so difficult to change? What does AI mean for learning? If AI can teach, what is the professor for? And if some of the constraints around which universities were designed are disappearing, what might we design instead? Part 1 was published on August 26, 2026, Part 2 on September 2, Part 3 on September 9, and Part 4 on September 16.

The university’s innovator’s dilemma and the need for “fast-second” universities

 Universities are not startups. We should stop pretending they are.

They are not designed to move fast, break things, pivot weekly, abandon legacy products, cannibalize their own revenue, bypass committees, ignore accreditation or build new markets from scratch.

Universities are designed for continuity. They are designed to confer legitimacy, preserve standards, develop expertise, maintain public trust, govern credentials and sustain complex communities of teaching, research, service, and professional formation.

Those strengths matter. In an AI-rich world, they may matter even more.

But they are not the strengths of disruptors.

That distinction is essential. If we ask universities to behave like startups, we will misunderstand both the problem and the opportunity. Most universities will not invent the AI-native future of learning from within. Their structures, incentives, assets, and governance systems were built for the current model. They are, in many ways, institutions of the status quo.

But that does not make them irrelevant. It means their strategy should be different.

Universities do not need to be first.

But they cannot afford to be slow.

Their opportunity is to become intelligent fast seconds: learning from disruptors, watching what students and markets actually adopt, avoiding what fails, and using their trust, scale, expertise and credentialing authority to integrate and legitimize the best new models of learning.

In an earlier Innovation Doctor article, We Helped Create AI. Why Are We Letting Others Design Its Future?, I argued that Ontario should create a new public, not-for-profit, AI-native learning institution: a trusted learning system designed around feedback, reflection, capability, lifelong learning, public purpose  and responsible AI adoption.

The central argument was not that AI should simply be added to existing universities, but that AI forces us to ask whether the existing university model remains the right institutional architecture for the world now emerging. That original paper explicitly framed the challenge as moving beyond “AI inside existing universities” toward asking what kind of public learning institution we would build today.

That argument built on two related Innovation Doctor themes. In my earlier article, The Student Will See You Now, I argued that AI shifts the learning relationship: students are no longer waiting passively for scarce feedback, scarce explanation or scarce access to expertise. In my work on AI-enhanced experiential learning, I argued that the real educational opportunity is not faster content production, but better learning design: deeper reflection, better feedback, stronger problem framing and more visible development of judgment.

This article takes the next step.

It asks not what AI makes possible for learning, but how existing universities are likely to respond — and why their best strategy may be fast second rather than first mover.

American academic and business consultant Clayton Christensen helps explain why universities are unlikely to disrupt themselves. Constantinos Markides, professor of strategy and entrepreneurship at the London Business School, and Paul Geroski, who was a leading economist in the United Kingdom, help explain what they might do instead.

The incumbent problem

In The Innovator’s Dilemma, Christensen’s central insight was not that great organizations fail because they are foolish, lazy or badly managed. His more uncomfortable argument was that successful organizations can fail because they are well managed according to the logic of their existing business. They listen to their best customers. They improve existing products. They protect familiar revenue streams. They allocate resources to opportunities that fit current metrics. They make sensible decisions.

And still they lose leadership.

That is the university’s AI dilemma.

The danger is not that universities will ignore AI. The danger is that they will use AI in exactly the ways that protect the existing university. AI will be welcomed, but translated into familiar institutional categories: a course issue, a cheating issue, a faculty development issue, a procurement issue, a research productivity issue, a student support issue, a policy issue, a risk issue.

All of these are real. None of them is enough.

The deeper issue is not AI use. It is institutional design.

Most university conversations about AI are still sustaining innovation conversations. They ask how AI can help students learn existing course material, help faculty grade more efficiently, detect misconduct, reduce administrative burden, support existing programs, or improve research productivity. These are reasonable questions, and they may produce useful improvements.

But they begin with the existing university. They begin with the course, the lecture, the assignment, the exam, the department, the semester, the degree, the transcript, the learning management system, the faculty workload model, the committee and the policy. AI is then asked to make that inherited architecture work better.

That is sustaining AI.

Disruptive AI asks a different question: what kind of learning system becomes possible when every learner can access continuous coaching, immediate feedback, adaptive practice, structured reflection, simulation, translation, mentoring support and guided experimentation?

That question does not begin with the course. It begins with the learner. It does not begin with the lecture. It begins with the learning journey. It does not begin with the assignment. It begins with the development of judgment, capability, confidence, creativity, ethical reasoning and agency.

That is why AI is not simply another educational technology. Used seriously, AI becomes a new layer of cognitive infrastructure. It changes the cost, speed, scale, personalization, feedback intensity and accessibility of learning. And when those things change, the institution itself must be questioned.

When assets become constraints

Christensen’s later work with Michael Raynor at the Ivey Business School in The Innovator’s Solution adds an important layer. Organizations are not defined only by their visible assets. They are shaped by their resources, processes and values: what they have, how they work, and how they decide what matters.

This is critical for universities because their greatest strengths can also constrain their imagination.

Universities have extraordinary assets: faculty expertise, disciplinary depth, degree-granting authority, campuses, libraries, laboratories, research infrastructure, student services, academic governance, public legitimacy, alumni networks, employer relationships and community trust. These should not be dismissed. They are part of why universities still matter.

But in a disruptive transition, the question is not whether these assets are valuable in the abstract. The question is whether they help or hinder the institution’s ability to respond to the new context.

A university with classrooms naturally asks how AI can improve classroom learning. A university with departments asks how AI fits into departmental curricula. A university with degrees asks how AI supports degree completion. A university with exams asks how AI affects academic integrity. A university with research incentives asks how AI can increase publication output. A university with committees asks what policies should govern AI use.

None of these questions is wrong. But together they reveal the incumbent trap. The institution sees AI through the assets it already has. It asks how this disruptive technology can preserve, improve, or defend the inherited model.

The harder question is different: which university assets become more valuable in an AI-rich world, and which become less central?

That is the question universities are avoiding.

The fast-second university

This is where Fast Second by Constantinos Markides and Paul Geroski becomes especially useful. Their argument challenges the mythology of first movers. In many new markets, the pioneers who create the market are not the firms that ultimately capture the greatest value. The skills required to create a new market are not always the same as the skills required to scale, standardize, consolidate, and dominate it.

Markides and Geroski argue that established firms are often better suited to being “fast seconds” than radical pioneers. Their advantage is not discovering the market from scratch. Their advantage is entering once the market begins to take shape, helping establish the dominant design, and using scale, resources, distribution, brand, trust and execution capability to grow the market.

That is the lesson universities need.

Universities should stop pretending they are natural colonizers of the AI-learning frontier. They are not. The AI-native future of learning is already being prototyped by startups, platforms, employers, students, professional bodies, informal learning communities, AI tutor companies, bootcamps, open-source communities and global education experiments.

Some of those experiments will be shallow. Some will be dangerous. Some will be hype. Some will fail.

But some will reveal real changes in how people learn, practice, receive feedback, build capability, demonstrate competence and move through careers.

The university’s job is not to dismiss these experiments because they did not originate inside the academy. The university’s job is to learn from them faster than it currently learns.

A fast-second university would not rush to copy every AI tool. That would be foolish. Fast second does not mean frantic second. It means watching carefully, learning systematically, and moving decisively when evidence begins to emerge.

A fast-second university would study what disruptors are proving. Which AI tutoring models actually improve learning? Which forms of feedback increase motivation and capability? Which assessment models employers trust? Which credentials gain traction? Which student behaviours deepen learning, and which create dependency? Which AI-supported project models help learners develop judgment, creativity, and evidence-based reasoning?

Then it would do what universities can still do better than most disruptors. It would bring trust, standards, disciplinary expertise, public legitimacy, quality assurance, research methods, convening power, credentials and scale.

That is the university’s AI advantage.

Not speed alone. Not disruption for its own sake.

Trusted fast adoption.

The danger is slow absorption

The problem is that universities are not currently organized to be fast seconds. They are organized to be cautious incumbents.

This is how AI gets domesticated. A teaching and learning centre turns it into a workshop. An academic integrity office turns it into a misconduct category. A course committee turns it into a syllabus statement. An IT department turns it into a procurement question. A research office turns it into a productivity tool. A strategic plan turns it into a paragraph.

The institution appears active, but the architecture barely changes.

That is not fast second. That is slow absorption.

Raynor’s Strategy Paradox adds another important warning. High-impact strategies require commitment under uncertainty, but the same commitments that make a strategy powerful can also make it dangerously wrong if the future unfolds differently.

No one knows exactly what AI-enabled higher education will look like in 10 years. We do not know which AI tutoring models will create lasting learning gains, how assessment will evolve, which credentials employers will trust, or how students will use AI when it becomes embedded in everyday work.

But uncertainty is not an excuse for waiting. It is a reason to build options.

A university that wants to be fast second needs protected experimentation. It needs spaces where AI-supported reflection, project-based learning, assessment redesign, modular credentials, lifelong learning, student advising, faculty development, research translation, public-sector problem solving and AI-supported experiential education can be tested without every experiment being forced through the full machinery of the current system.

Some experiments will fail. Some will produce modest improvements. Some will reveal the architecture of the next university.

That is the point.

The goal is not to predict the future. The goal is to learn which futures are worth committing to.

Connected, but protected

If AI-native learning is developed entirely inside existing university structures, it will be judged by existing university priorities. Does it fit the semester? Does it belong to a department? Does it preserve the degree structure? Does it protect enrolment? Does it align with workload rules? Does it comply with existing assessment practices? Does it survive the committee process?

These questions are understandable. They are also exactly why universities struggle to respond to disruption.

A fast-second strategy requires a different organizational home. Not necessarily a university startup. Not necessarily a separate corporation. Not necessarily a replacement for existing institutions. But something protected enough to learn.

It could be a new public AI-native learning institution. It could be an AI learning trust, a university-system innovation studio, a living laboratory for AI-enabled public learning, a cross-institutional experimental credential body, or a public-interest AI education commons.

The form matters less than the design principle.

It must connect to university assets without being captured by university routines.

Connected, but protected.

That is the strategic design challenge.

The real fast-second test

Universities should stop asking whether they can become AI startups. They cannot. And they should not.

The better question is whether universities can become trusted fast adopters of what the disruptors reveal. Can they identify which AI-enabled learning models actually work? Can they integrate those models into credible education? Can they redesign assessment around judgment, evidence, reasoning, creativity, collaboration, and ethical awareness? Can they use AI to support struggle rather than eliminate it? Can they serve learners who do not fit the traditional full-time degree model? Can they build new pathways without destroying trust? Can they move before the market moves around them?

That is the fast-second test.

The greatest risk is not that universities fail to invent the next model. That may not be their role. The greater risk is that they fail to learn from those who do.

Disruptors will experiment with AI tutors, personalized learning, skills-based credentials, adaptive coaching, employer-linked learning, simulated practice, project-based portfolios, conversational assessment and new forms of learner support. If universities dismiss these developments because they do not look like traditional education, they will repeat the classic incumbent mistake.

At first, the new models may look weaker: less rigorous, less prestigious, less complete, less academic, less controlled. But disruptive models often begin that way. They improve. They find underserved learners. They build new habits. They reduce friction. They offer convenience, affordability, personalization and speed. They create new expectations.

By the time the traditional institution decides the model is serious, the learners may already have moved.

Universities do not need to be first.

But if they are slow second, they may become irrelevant third.

Policy is not strategy

Universities need AI policies. But policy cannot be the centre of the strategy.

Policy defines boundaries; experimentation discovers possibilities. Policy manages risk; experimentation creates learning. Policy asks what is allowed; experimentation asks what is valuable. Policy protects the institution; experimentation helps the institution adapt.

The danger is that policy may move faster than imagination. We may end up governing futures we have not yet explored.

That is backwards.

A fast-second university needs disciplined experimentation before premature closure. It needs evidence before orthodoxy. It needs learning before standardization. It needs protected spaces where faculty, students, staff, employers and communities can test new models without every experiment being forced through the inherited machinery of the current system.

This is not a call for reckless AI adoption. It is a call for rigorous institutional learning.

The Innovation Doctor diagnosis is simple: universities are treating AI as an implementation problem before they have treated it as a learning problem.

The university still matters

This is not an anti-university argument. It is the opposite.

In an AI-rich world, trust may become more important, not less. Learners will need credible pathways through an ocean of content, tools, claims, credentials and noise. Employers will need signals they can trust. Governments will need public-interest institutions that can protect equity, ethics, privacy and accountability. Communities will need partners who can help translate AI capability into human value.

Universities can play that role. But not if they confuse historic legitimacy with future relevance.

Trust must be earned again in a new context. Scale must be redesigned. Expertise must be reorganized. Credentials must be revalidated. Learning must be rebuilt around what AI now makes possible.

That is why the fast-second strategy matters. It allows universities to use their strengths without pretending those strengths are enough.

Stop pretending unoversities are startups

The university does not need another AI committee that supervises experimentation from a safe distance.

It needs new organizational forms with the authority to learn from disruptors, test what works, gather evidence, protect trust and scale new models of learning before the market moves around it.

Christensen warns us that incumbents often fail because they use new technologies to sustain old models. Raynor warns us that high-impact strategy requires commitment under uncertainty, and that uncertainty must be managed through options rather than denial. Markides and Geroski remind us that established organizations may not be the best pioneers of new markets, but they can be powerful fast seconds if they know when and how to move.

Together, their message for universities is clear.

Stop pretending universities are startups. Stop treating AI as another tool to preserve the old model. Stop confusing policy activity with institutional learning. Stop assuming that assets built for the pre-AI university will automatically remain assets in the AI-rich university.

Learn from the disruptors. Build protected experiments. Create evidence. Move fast when the pattern becomes clear. Use trust, scale and legitimacy to make the best new models better.

Universities are not built to be first.

But they cannot afford to be slow.

References

Christensen, C. M. (1997). The Innovator’s Dilemma: When New Technologies Cause Great Firms to Fail. Harvard Business Review Press.

Christensen, C. M., & Raynor, M. E. (2003). The Innovator’s Solution: Creating and Sustaining Successful Growth. Harvard Business School Press.

Markides, C. C., & Geroski, P. A. (2004). Fast Second: How Smart Companies Bypass Radical Innovation to Enter and Dominate New Markets. Jossey-Bass.

Raynor, M. E. (2007). The Strategy Paradox: Why Committing to Success Leads to Failure. Currency/Doubleday.

 R$