AI for Distance and Open Learning Institutions

An open university does not have a campus peak. It has a country-shaped peak. Learners sit in different states, on different networks, with study-centre staff who may see them twice a year. Identity is a photograph, an enrolment number, and a sometimes-shared mobile. Examinations are logistics plus integrity. If you treat that institution as a residential campus with more students, you will buy the wrong agent.

Indira Gandhi National Open University is the name people use as a shorthand for this shape. State open universities and dual-mode deemed or central campuses have their own statutes and volumes. This article uses IGNOU-like as a systems description — large enrolment, distributed study centres, distance materials, term-end exams — not as a claim about any one university's current stack. Read that university's Act and student handbook before you copy a control.

UGC's open and distance learning and online regulations exist and have been revised more than once. Hedge the live text. This is a field guide for controllers of examinations, student-services directors and CIOs. It is written on 17 August 2026. It is not legal advice and not an IGNOU manual.

Three problems that are not chat

Scale. A residential helpdesk can fail closed and open a counter. An open university's learner may be a working adult two states away. Self-service has to work in low bandwidth and in more than one language. An agent that only speaks fluent campus English is a Delhi toy.

Identity. The person on the phone may be a relative. The person on the portal may be a study-centre clerk. Enrolment numbers leak in WhatsApp groups. An agent that resets a password or changes an exam city because the voice knew the mother's name is an integrity incident. Bind sensitive actions to the same identity standard the university already uses for exam hall tickets — and if that standard is weak, fix it before you add a tool.

Exam integrity. Term-end and online exams are the institution's credibility. An agent that explains the exam calendar is useful. An agent that drafts answers to assignment questions from the current question bank is a leak. An agent that helps the invigilator by scoring suspicion on a webcam is a different product with a different rights impact. Do not bundle them.

Open-learning workflows versus what an agent may do in a first season.

Workflow	First-season agent	Never without a separate project
Programme / calendar / fee FAQ	Retrieve signed handbook; draft reply	Promise a seat or a waiver
Study-centre locator	Read the published list	Reassign a learner to a centre
Assignment submission status	Read authenticated status	Accept or grade the assignment
Exam city / hall ticket	Explain published rules	Change city or issue a hall ticket
Revaluation request	Draft the form for the learner	Decide the revaluation
Proctoring / suspicion	Out of scope	Any automated integrity finding that affects a result

Materials, assignments and the leak surface

Open universities sit on large content stores: SLM PDFs, recorded lectures, assignment banks, previous-year papers, some of it licensed, some of it examiner-confidential. A retrieval agent that indexes everything in the CMS will eventually retrieve next term's assignment. Corpus design is the product.

Pin three corpora: public prospectus, authenticated learner-status APIs, and a signed FAQ the student-services director owns. Keep assignment banks, examiner notes and unpublished results off the agent's index. If a study-centre clerk needs those, they use the existing staff system, not the learner-facing bot.

Translation and summarisation of study materials can be a legitimate aid. It is also a copyright and quality problem. Do not let a model silently rewrite a course that a board of studies approved. If you summarise, label it as a study aid, keep the canonical PDF one click away, and do not treat the summary as the syllabus.

Study centres are part of the system

A bot that tells a learner one thing while a study-centre coordinator tells them another creates hearings. Publish the same circular pack to both. If the agent is allowed to draft coordinator replies, those drafts must cite the same pack. Coordinators will ask for a staff agent that does the paperwork. Paperwork at a study centre includes identity documents and sometimes fee cash. Default deny on writes. A coordinator-facing retrieval tool over circulars is enough for year one.

Regional centres already have escalation paths. Put the agent at the bottom of that path, not as a bypass. A learner who has already opened a ticket should not get a contradictory chat.

DPDP at national scale

Open-university datasets are large and long-lived. Alumni of a twenty-year programme are still identifiable. Purpose limitation and retention actually hurt here: you cannot keep every chat forever for quality and also claim you minimise.

Processors multiply: print vendors, exam centres, online-exam vendors, SMS, the AI vendor. A single DPA with the model host is not an inventory. Cross-border study (overseas learners, foreign study centres if any) is extra-territorial processing in both directions. Take counsel. Do not let a generic campus agent become the overseas student office.

Two rooms you can walk into

Volume is not an excuse for a second registrar.

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.

Our volume means only AI can cope.

Volume means you need self-service on the three questions you can document, plus a ticket system that scales. It does not mean the agent may issue hall tickets. Scale the FAQ. Do not scale authority.

Learners already use ChatGPT on assignments.

That is a pedagogy and integrity policy for coursework. It is not a reason to give an institutional agent the unpublished bank. Those are different risks.

Study-centre staff will only adopt WhatsApp.

Then write WhatsApp as a channel with a staff identity, a log, and no write tools. A coordinator WhatsApp group without logs is already your incident.

UGC ODL rules require us to be innovative.

Read the live ODL/online regulations and do what they actually require. Innovation is not a clause that says index the CMS. Hedge the current PDF.

A five-week open-learning agent that does not sit the exam

If exam season is inside those five weeks, slip the go-live, not the integrity rule.

1. Week 1: name volumes — learners, study centres, peak ticket types. Pull last term-end's top twenty questions.
2. Week 2: pin three corpora. Explicitly list what is out: assignment banks, examiner notes, unpublished results, ID images.
3. Week 3: identity. Agent inherits the portal login. No password reset, no exam-city change, no mobile-number change as tools.
4. Week 4: DPA, training ban, retention for chats, CERT-In log floor. Study-centre staff get a separate retrieval view, still no writes.
5. Week 5: tabletop — a leaked TMA, a relative on the phone, a contradictory coordinator message. Fix the pack. Then open to 5 percent of learners.

How this shows up in the file

Subject: Distance / open-learning agent — scope. This is not a residential-campus bot. Scale, identity and exam integrity are the design constraints. The agent may retrieve the signed prospectus, calendar and FAQ. It may not change identity, exam city, or results. Assignment banks and unpublished assessments are outside the index. UGC ODL/online regulations are cited from a dated PDF for any specific duty; this note does not invent one.

Attach the corpus list, the identity rule, and the exam-season freeze.

What we will and will not claim

Prcept AI will run an on-prem or air-gapped retrieval agent on the three corpora you pin, at the volume your stack can hold, without training on learner chats. We will refuse exam-integrity and result tools unless a later, separate project with a rights impact assessment says otherwise. IGNOU-like scale is an engineering problem. It is not a reason to skip the pin.

What the next file must contain

“AI for Distance and Open Learning Institutions” earns a line in the noting only if a P3 University can attach proof of “open university AI support.” A heading is not proof. A vendor slide is not proof. A workshop photograph is not proof.

Write three dated sentences: what was decided, who owns it after the next posting order, and when it will be re-checked. If you cannot write the three sentences, you are not ready to buy, to sell, or to go live.

Leave unsourced percentages out of the note. DPDP is not a blanket localisation statute. The November 2025 AI governance text is guidance, not an Act. CERT-In’s 28 April 2022 directions still set specified incident and log clocks. A PAC, when lawful, lives in GFR Rule 166.

• Name the designation that owns “open university AI support.”
• Attach one artefact a stranger can open next year.
• Record the instrument you are actually using.
• Revisit when the model, the SI, the notice or the posting changes.

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.