Measuring ROI on a Campus AI Deployment

A vendor closed a campus demo with a slide titled 14x ROI. The multiplier came from minutes saved on 40,000 hypothetical tickets, priced at a full professor's hourly cost, with no plan to reduce a single contract seat. The finance officer asked when the saving would appear in the revised estimate. Nobody answered. The demo was fluent. The return was imaginary.

Campus AI can pay for itself. It pays when a named queue shrinks, a named error rate falls, a named overtime bill drops, or a named compliance failure stops happening. It does not pay because a dashboard counted tokens or because a student said the bot was helpful in a three-day trial.

This framework is for registrars, finance officers and CIOs who have to write a completion report. It is written on 17 August 2026. It is not an accounting standard and not legal advice. If your campus follows GFR-like utilisation certificates or a private trust's audit manual, use that manual's idea of a saving. Do not use a vendor's.

Pick a denominator before the pilot

A denominator is a count you already trust: helpdesk tickets of a tagged type, average first-response hours, registration-week overtime hours, percentage of fee queries reopened, days to issue a bona fide certificate, revaluation form defects. If the count does not already exist, spend two weeks creating it. Do not let the vendor's telemetry become the only number in the file.

Write the baseline from a clean period — last corresponding semester, not the week after a system outage. Write who owns the count (registrar, finance, exam). Write the target as a range, not a single heroic percentage. Write the cost side as five-year cash: licence or AMC, hardware you will buy, power, integration, evaluation labour, parallel-run staff, exit cost. Ignore value of innovation. Ignore avoided hypothetical lawsuits unless counsel will sign the probability.

What counts as a return versus what only counts as a story.

Claim	Counts if	Does not count if
Ticket deflection	Tagged volume falls and you redeploy or cut a seat, or overtime falls on the ledger	The bot closed tickets students immediately reopened at the counter
Faster first response	SLA clock you already publish moves, sampled weekly	The bot replies instantly with please visit the portal
Fewer form defects	Rejections of a named form fall, officer-counted	The agent drafts forms nobody files
Staff satisfaction	Useful as colour, never as the sole ROI	A survey of six early adopters
Student NPS	Secondary, after integrity and correctness	A QR code on the demo booth
Token spend down	Never a campus outcome	Always

The three tests of a saving

Attribution. Did the agent cause the change, or did you also simplify the form, hire two people, and shift the calendar? If you changed three things, say so. A joint cause can still justify the stack if the finance note is honest.

Cash or capacity. A saving is cash that left the revised estimate, or hours you reassigned to named work that was previously not done. Hours returned to faculty with no change in load is a speech.

Quality floor. A cheaper queue that answers the fee rule wrong is a negative return. Sample correctness against the signed circular pack every week. One wrong waiver explanation can cost more than a year of AMC.

What not to monetise

Do not monetise brand as an AI campus. Trustees can value brand; they should not let a vendor price it. Do not monetise DPDP compliance. Compliance is a duty. An agent that makes compliance harder is a cost. Do not monetise exam integrity as a positive ROI of a general helpdesk bot. If you buy a separate integrity product, measure integrity failures, not chat ratings. Do not annualise a two-week admissions spike as if it were fifty-two weeks of saving.

Report shape for the finance committee

One page. Baseline, observed, cost, quality sample, decision: continue, shrink, or exit. Annex the method. If the observed change is inside the noise of last year's variance, say inconclusive and exit or extend the measurement — do not declare victory.

Public universities should expect utilisation language. Private trusts should expect a trustee who has sat on a manufacturing board. Both will smell a multiplier that uses the Vice-Chancellor's notional hourly rate.

• State the window and the matching baseline window.
• State every other change in the process during the window.
• State the quality sample size and the error types found.
• State whether any staff cost actually moved.
• State the exit cost if you stop now versus in twelve months.

Two rooms you can walk into

Both campuses bought a helpdesk agent. Only one could find the return.

Objections you will hear — and what to do with them

These are the lines that stall the file. Answer them in the room, then put the answer in the note.

Some benefits are intangible.

Then keep them in a paragraph labelled colour. Do not add them to the rupee line. Intangible plus a fake multiplier is how you lose the next honest project.

We cannot get a clean baseline.

Then you are not ready to claim ROI. You may still run a time-boxed pilot to learn. Call it a learning cost. Do not call it a return.

The vendor's success manager will produce the ROI pack.

They may draft. You sign. If their pack uses professor-hour fiction, strike it before it reaches audit.

Quality sampling is too expensive.

Then you cannot safely deflect. An unsampled deflection is how wrong fee rules become folklore.

A six-week ROI frame you can attach to the work order

Write this before go-live. Retrofit ROI is fan fiction.

1. Week 1: choose one workflow and one primary metric you already count. Name the owner.
2. Week 2: lock the baseline window and pull the numbers. List confounding changes planned in the same term.
3. Week 3: lock the five-year cost model. Separate pilot from production.
4. Week 4: design the quality sample — size, scorer, fail types, stop rule if error rate exceeds X.
5. Week 5: write the decision rule in the work order: continue / shrink / exit.
6. Week 6: finance and registrar countersign. Only then switch the agent on.

How this shows up in the file

Subject: ROI frame for the named workflow agent. Primary metric owned by a named post. Baseline window and value stated. Target range stated. Five-year stack cost stated. Quality sample size and stop rule stated. A saving will be recognised only if a ledger line or a named redeployment moves. Vendor dashboards are annexures, not the return.

If the primary metric cannot be named, do not sign the work order.

What we will and will not claim

Prcept AI will give you exportable counts of conversations and retrievals you own. We will not write a 14x slide. If your overtime line does not move, exit us. DPIIT recognition is not a return.

This article is informational field guidance for Indian universities and public institutions, not legal, procurement, audit or engineering advice. Confirm against the live Gazette, GFR, state financial rules, GeM terms, UGC text, GIGW, DPDP commencement, departmental manual and your counsel before you file it.

How this works on an Indian campus

A P3 University should be able to run “Measuring ROI on a Campus AI Deployment” without importing a US playbook. “campus AI ROI measurement” hits UGC/AICTE/NAAC clocks, exam secrecy, reservation rules, and students who may be minors.

ROI on a campus agent is a change in a named workflow, priced against a five-year stack cost. Token charts are not a return. Savings you will not take to the ledger are not a return either. DPDP applies to student personal data. Chatbots are not a strategy. Exam and admissions writes stay with officers. Affiliated colleges need isolation, not one shared index.

• No production student data in a vendor SaaS sandbox.
• Write the academic integrity policy before the tool.
• Consent and purpose tags on student-facing agents.
• Budget for staff training, not only licences.

Close this loop before the next CAB

Put “Measuring ROI on a Campus AI Deployment” on the next change-advisory or bid-opening agenda as a single line item with an owner. If it cannot earn a line item, it will not earn a control. The owner should be a P3 University, not “the vendor.”

Revisit the item when the model, the GeM term, the region, or the SI changes. “campus AI ROI measurement” is not a one-time workshop. It is a watch item. Date the last check. Unsigned watch items are souvenirs.