Transport Department Use Cases, Ranked by ROI

Transport is where vendors bring drones. The commissioner has a different problem. Permit files sit for weeks because a circular is missing from the noting. Fancy-number auctions generate PDFs nobody can search. Enforcement wants a magic camera. The driving-test yard is short of examiners. Citizen calls ask the same three questions the portal already answers badly. If you rank those jobs by demo charm, you will buy the camera. If you rank them by reconstructable hours and legal risk, you will buy a clerk.

This data study is a method, not a national statistic. We have not found a MoRTH publication that ranks state-level agentic use cases by rupee ROI for FY 2025-26. Anyone who quotes a single ROI percentage for 'transport AI in India' is selling a press note. What follows is how a state CIO can build a ranked list in four weeks, with a starter ranking you must overwrite with your own timings.

It is written on 17 August 2026. It is not legal advice. Vahan, Sarathi and e-challan stacks are living systems with their own access regimes. Describe integrations as connectors and purpose tags. Do not invent a write API because the RFP would be prettier.

How to score a transport use case

Give each candidate four marks out of five, then multiply. Volume: how many times a year the work happens. Time: minutes a competent person spends when the file is ordinary. Risk: what happens if the machine is wrong — a delayed answer, a wrong permit, a wrongful challan, a licence photograph in a foreign log. Control: whether you own the system of record or are a guest on a national platform. High volume times high time times low risk times high control is where you start.

Refuse to score 'AI traffic management' as one row. Split it. Signal-timing advice to an engineer is not the same object as an automated penalty on a citizen. The second row almost always loses on risk even if the first row is worth a pilot.

Starter ranking for a composite state transport department. Replace the hour ranges with your own sample of fifty cases. These are planning figures, not a census.

Use case	Why it scores	Why it fails	Starter rank
Permit-file completeness and circular retrieval	High volume, high clerk time, you own the noting, low write risk	If you let the agent sanction the permit	1
Citizen FAQ on already-published fees and forms	High volume, low risk if it refuses when the circular is missing	If it invents a fee	2
Auction and fancy-number document search	Painful internal search, reconstructable, no citizen right changed	If records leave the VLAN	3
Assisted extraction on licence or permit scans, on-prem	Saves keying if confidence is shown and an officer accepts	Public OCR API; silent overwrite of Vahan / Sarathi	4
Examiner aid in a notified test process	Only if the state instrument already allows the aid	A new test invented by a camera vendor	5 — conditional
Automated e-challan from a vision model	Looks like ROI in a slide	Process, notice, evidence, appellate path, camera calibration	Do not rank until legal process exists

National platforms are not your database

Vahan and Sarathi are not a playground. They are national applications many states use through notified access. Your agent may, where you have a lawful connector, read what you are allowed to read, with a purpose tag. It may not scrape a public portal and call that an integration. It may not write a licence field because a model is confident. If the only interface you have is a clerk with a login, say so in the RFP. Honesty here prevents a six-month SI fantasy.

Personal data in this department is not abstract. Photographs, medical fitness, addresses, phone numbers, and sometimes Aadhaar-seeded KYC sit in the same drawer as vehicle numbers. DPDP will treat much of that as digital personal data. CERT-In already wants ICT logs for a rolling 180 days in Indian jurisdiction. Design the agent as if both sentences are true, because they are.

Two commissionerates, same budget line

Objections you will hear — and what to do with them

Leadership wants a visible camera

Give them a visible FAQ and a visible drop in permit pendency first. Cameras can come when the legal process exists. A gantry is not a strategy. It is a procurement category.

We cannot rank without a data-science cell

You need a stopwatch and fifty files. A data-science cell can come later. Ranking is administrative, not glamorous.

The vendor already has a Vahan integration

Ask for the letter that authorises it, the purpose tag, and whether it writes. 'Integration' that is a stored clerk password is a security incident, not a use case.

Challans will fund the whole AI programme

Do not finance a governance stack on contested penalties. That incentive is how exclusion errors are born in another department and how unlawful challans are born in this one.

A four-week playbook to publish your own ranking

1. Week 1: list ten candidate jobs. Kill any job that requires a write to a national platform you do not control, or a penalty process you do not have.
2. Week 2: time fifty ordinary cases on the top three remaining jobs. Write the minutes down. Do not let the vendor time them.
3. Week 3: score risk with counsel in the room — wrongful permit, leaked photograph, invented fee. Drop any job counsel will not initial.
4. Week 4: publish a one-page ranking as an internal note. Fund rank 1 only. Put rank 2 in the next change-request envelope, which this cluster also tells you to budget.

File note you can paste

Subject: Ranking of agentic use cases in the Transport Department — method, not a national figure.

This department has not relied on a single national ROI statistic for transport AI. Use cases will be ranked by sampled clerk or officer time, legal risk, and whether we control the system of record. Agents will not write to Vahan, Sarathi or any national register without a documented authorisation. Personal images and medical certificates will not be sent to public model APIs. Automated penalties will not be piloted until notice, evidence and appeal packets exist. The first funded job will be permit-file completeness and circular retrieval unless the sample of fifty cases shows a better rank 1.

This note is an internal aid. It is not legal advice.

Prcept AI will help you run the clerk on your rack. We will not invent a Parivahan API. We will not train on licence photographs. If you want a camera programme, hire a camera programme with its own legal design. Do not hide it inside an agent RFP.

Fees the agent must not invent

Transport citizen FAQs die on invented fees. A model that is 'usually right' about a learner-licence charge is still a corruption story when it is wrong. Bind every rupee amount to a dated circular hash. If the circular is missing, the agent says it does not know and points at the official page. It does not split the difference between last year's PDF and a newspaper clip.

The same rule applies to eligibility for permits. 'Usually granted in this district' is not a rule. If the notification lists grounds, retrieve the list. If the officer has discretion, say so and stop. A helpful paragraph that promises a permit is how you manufacture a legitimate expectation you cannot defend.

Re-time the FAQ after every fee circular. That re-timing is a change-request unit, not a goodwill patch. If your envelope cannot pay for a fee-table update, you should not have offered a fee-speaking agent.

This article is informational field guidance for Indian public institutions, not legal, transport-regulatory or procurement advice. Confirm against the Motor Vehicles framework, your state notifications, Parivahan access rules, DPDP, CERT-In directions, and counsel before you file it.

How to sequence this in a state, not a slide

“Transport Department Use Cases, Ranked by ROI” is a department problem. A P1 CIO/CTO should name the legacy system, the officer who owns the file, and the citizen charter clock before buying “transport department automation”.

Transport departments do not need a vision of 'AI traffic'. They need a ranked list of clerical hours they can recover without writing to Vahan, inventing an e-challan court, or sending licence images to a public API. Do not invent league tables of states. Read tenders and policies. Election Model Code of Conduct can freeze a rollout. NIC is a partner, not a villain. SDC readiness is GPU, power, ops and egress — not a logo.

• Audit the legacy store first.
• Keep mutation and money as officer actions.
• Map SLAs to the citizen charter.
• Budget change requests after go-live.

Close this loop before the next CAB

Put “Transport Department Use Cases, Ranked by ROI” 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 P1 CIO/CTO, not “the vendor.”

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