Approach

How I sequence AI adoption in an institution.

The enterprise AI playbook doesn't transfer cleanly to higher education. This is the four-phase maturity model I use instead — one that lets different departments move at different speeds without the whole program stalling behind the most cautious stakeholder in the room.

01

Assisted

AI supports individual tasks — drafting, summarizing, search — with a human reviewing every output. Low risk, fast to deploy, builds literacy.

02

Advisory

AI recommends decisions across a defined workflow — admissions triage, resource planning — but a person still approves the action.

03

Adaptive

Systems begin to act within guardrails, learning from outcomes and adjusting — student support routing, dynamic scheduling — under active oversight.

04

Autonomous

Agentic systems execute defined processes end to end, with governance built into the architecture rather than bolted on after.

I built this model running a structured AI discovery across fifteen departments at Shiv Nadar Institution of Eminence — mapping more than forty interconnected systems across admissions, registrar, finance, research administration, and learning management. What became clear was that treating "AI strategy" as a single institution-wide initiative was the wrong frame. Different departments landed at different phases by design: student-facing support workflows moved to Advisory quickly, because the downside of a bad recommendation was bounded and reviewable. Anything touching academic records or accreditation-relevant decisions stayed in Assisted far longer, because the cost of an ungoverned error there isn't just operational — it's regulatory and reputational.

That unevenness isn't a failure of the program. It's the discipline working as intended — and it's the same instinct that made the ServiceNow/ITSM transformations I ran in enterprise telecom actually stick: sequence the change so the governance keeps pace with the ambition, not the other way round.

Read the full essay on this model →