Institutions adopt AI for many of the same reasons businesses do: efficiency, consistency, scale, cost.
But institutions are not just organisations.
They are interfaces between individuals and society.
When an institution changes how it works, the effects don’t stay inside. They ripple outward — into education, access to services, rights, obligations, and trust.
Like in business, AI rarely arrives as “AI”.
It arrives as help.
In education, AI assists with grading, feedback, curriculum planning, plagiarism detection, personalised learning paths. Used thoughtfully, it can support teachers and students. Used uncritically, it can reshape learning around what is easy to measure rather than what is hard to teach: curiosity, judgment, creativity.
In social services, AI helps triage cases, detect patterns, allocate resources. These systems promise fairness and efficiency. The risk is quiet but serious: turning complex human situations into scores, categories, or thresholds — and mistaking consistency for justice.
In security and policing, AI is used for prediction, surveillance, risk assessment. Here the stakes rise sharply. Errors are no longer just inconvenient; they can be life-altering. Bias, once automated, becomes harder to challenge — not because it is invisible, but because it appears objective.
In taxation and administration, AI sorts, flags, audits, optimises. It can reduce errors and speed up processes. It can also create systems that feel opaque, unappealable, and distant — especially when decisions are technically correct but socially incomprehensible.
In justice, AI assists with document analysis, precedent research, risk evaluation. Used as a tool, it can support human judgment. Used as an authority, it threatens something more fragile than efficiency: legitimacy. Justice depends not only on outcomes, but on the ability to understand and contest how those outcomes were reached.
Across these domains, the pattern repeats.
Institutions adopt AI to cope with complexity and overload.
AI delivers consistency and scale.
And slowly, quietly, decision-making shifts — not always in intent, but in practice.
What distinguishes institutions from businesses is not just their mission, but their power. Institutional decisions shape norms. They define what counts as acceptable, fair, normal, or suspicious. When AI enters these spaces, it doesn’t just optimise processes — it participates in governance.
This raises a question institutions are not used to asking explicitly:
When an AI system influences decisions that affect the public, who is it accountable to?
The developer?
The vendor?
The institution?
The citizen?
And just as importantly:
Who has the right — and the ability — to question it?
Institutions often move slowly, and for good reasons. But AI moves fast, and embeds itself quietly. By the time debates catch up, systems are already in place, habits formed, dependencies established.
The challenge is not whether institutions should use AI. Many already do, and will continue to.
The challenge is whether they can adopt it without surrendering what gives them legitimacy in the first place: human judgment, transparency, and the possibility of appeal.
If AI is becoming an institutional partner, the question is no longer only what works — but what remains contestable.