Tempo conflict at the heart of AI in higher education

This blog was kindly authored by Somayeh Aghnia, Cofounder, London School of Innovation.

The conversation about AI often moves too fast. Faster models. Faster outputs. Faster decisions. And universities, institutions built to move through deliberation, governance, and public trust, are being dragged into a tempo they did not design for.

This is not a minor operational irritation. It is a structural tension.

Universities are ‘slow’ for reasons that are not inefficiency: fairness needs process; standards need consistency; legitimate authority needs accountability; education needs continuity across cohorts; public trust needs explainability. Yet AI is ‘fast’ in ways that are not simply fashion: capability changes in months; norms move in weeks; expectations reset in days; learners and staff adopt tools with or without permission.

The strategic question, then, is not ‘How do we adopt AI?’ It is more fundamental:

How do you move at the right speed?

Too slow, and you become irrelevant in a very specific way: not that universities disappear, but that they lose their organising power over practice. AI use becomes unofficial, uneven, and privately optimised. Teaching teams improvise. Students develop parallel norms. What the institution says and what happens in reality drift apart. The university remains the credentialing authority, but it is no longer the site where the rules of the game are collectively set.

Too fast, and a different failure mode appears. In the rush to ‘keep up’, decisions get made through convenience and urgency rather than legitimacy. Policy becomes reactive. Implementation becomes inconsistent. Equity becomes accidental. And because universities operate on trust, ‘breaking things’ is not a learning cost; it is a reputational and moral cost. In an institution that grades, selects, and certifies, speed can quietly become a way of exercising power without adequate scrutiny.

This is a tempo challenge: navigating exponential technology within systems designed to move slower and cyclically, where the cost of being wrong is not only financial or technical, but civic.

What makes this challenge slippery is that ‘speed’ is rarely one thing. Inside any university, there are multiple clocks running at once: academic cycles, committee cycles, regulatory cycles, procurement cycles, student cycles and staff workload cycles. AI collapses these clocks into direct conflict. It forces a confrontation between the tempo of technological change and the tempo of institutional legitimacy.

A useful analogy appears in Anna Wood’s recent HEPI Weekend Reading piece on whether universities should build or buy their online education capability. The surface question is about online delivery, but underneath sits a capability-and-pace dilemma: build internal capacity slowly and retain agency, or buy speed and risk dependency and misalignment. AI has brought that same dilemma into the centre of the university itself, not just its delivery channels.

So what futures might unfold, depending on how this tempo conflict is handled?

One possible future is the compliance university: a thickening layer of policy, detection, and guidance designed to restore an earlier sense of control. This produces visible activity and institutional reassurance. But it also tends to create a widening gap between formal rules and lived behaviour, because the technology’s use continues off-platform. Over time, the institution risks becoming a regulatory shell around an informal reality, students and staff adapting in private while governance tries to catch up in public.

Another future is the platform university: rapid adoption, rapid procurement, rapid rollout. The institution moves at the speed of the vendor roadmap, hoping momentum will substitute for coherence. This can generate genuine innovation, but it can also create whiplash: uneven practice across departments, unclear standards across cohorts, and a growing sense that educational norms are being set by tool capability rather than academic judgement. When backlash arrives, it tends to be moral rather than technical: concerns about surveillance, workload, fairness, and the quiet reshaping of assessment and academic labour.

A third future is more paradoxical: the split-tempo university. Parts of the institution move fast, experimenting, learning, iterating, while other parts move slowly, holding standards, rights, consistency, and trust. This can look like a pragmatic coexistence of tempos. It can also look like internal fragmentation: one culture speaking the language of agility and innovation, another speaking the language of deliberation and protection, each viewing the other as the obstacle. In that scenario, the problem is no longer AI. The problem is institutional coherence.

A useful contrast comes from smaller and specialist providers that are not trying to retrofit AI into an inherited operating system. The London School of Innovation, for example, has been designed as AI-native from inception, using AI as a lens for curriculum, assessment and institutional operations rather than ‘implementing’ tools into existing structures. That does not remove the tension between speed and trust, but it shifts where the tension sits: the question becomes less about catching up and more about what kind of institutional tempo is built into the model from day one. That is still a hard question; trust, standards and legitimacy do not come pre-installed,  but it surfaces the design choice that older institutions rarely get to make explicitly. For most universities, the operating system already exists. The question is whether AI requires patching it, or replacing it, or learning to run both in parallel.

These futures are not predictions. They are ways the tension expresses itself when pace is treated as a reaction rather than a design variable.

And that is why the ‘right speed’ question is not merely managerial. It is philosophical and political. It forces universities to confront what they are willing to trade for relevance, and what they are unwilling to sacrifice even under pressure. There is a harder recognition underneath all of this: in universities, pace is never neutral. Moving quickly tends to empower those already equipped to act quickly, staff with lighter workloads, students with stronger digital capital and departments with more resources. Moving slowly tends to protect those already empowered by process, established curricula, existing hierarchies and dominant assessment cultures. Neither speed nor caution is inherently equitable. The tempo question is therefore also a distributional question: who gains and who loses depending on how fast the institution moves, and whether any of that is being consciously considered.

This is where the Tai Chi metaphor becomes more than aesthetics. Tai Chi is not slow because it is timid. It is slow because it is controlled. It is a discipline of timing, balance, and sensitivity to force. Power comes from knowing when to yield and when to move decisively, not from constant acceleration.

In the AI era, universities may discover that the hardest part is not choosing between speed and caution, but learning to inhabit the tension without collapsing into either paralysis or panic.

Because AI will not wait for committee schedules. And universities cannot afford to abandon the very qualities that make them socially meaningful: legitimacy, standards, and trust.

So the tempo question around AI in Higher Education remains,  and it is a demanding one which will not wait for the next committee cycle:

How to move fast enough to remain relevant, without moving so carelessly that they cease to be trustworthy. These two demands do not resolve. They have to be held together under pressure.