The Problem
For a long time, field collections ran almost entirely on cash. An agent would visit a delinquent customer, collect the overdue amount in hand, and submit it later. The process worked, but it came with all the risk and friction that cash handling brings on both sides.
The shift toward digital had been gradual. The first attempt was simple: agents would send a payment link over SMS or WhatsApp. The problem was obvious: even with an official Bajaj Finance SMS header, a link asking you to click and pay money mirrors one of the most common scam formats in India. Customers ignored it, and agents couldn't do much to convince them otherwise.
The next step was Collect Request. The agent would open DMS One App, pull up the customer's UPI ID, and trigger a collect request directly. That request would land as a push notification inside the customer's most-used UPI app, PhonePe or GPay, prompting them to approve the payment.
On paper, this was better than a link. In practice, the trust problem didn't go away. It just moved. The customer would be mid-conversation with a field agent and suddenly receive a random payment notification in their PhonePe. They hadn't initiated anything. They hadn't scanned anything. A notification had simply appeared asking them to authorize a transaction. For a large portion of customers, that still reads as suspicious. The agent standing in front of them couldn't override that instinct.
Digital payment adoption stayed stuck. We were collecting ₹1 Cr monthly through digital channels despite an active field presence that reached millions of customers.
What I Found
Collect Request had a trust problem, and it wasn't one you could fix by redesigning the notification or adding more information to the screen. The issue was structural: the customer was being asked to act on an incoming request with no visual confirmation they could validate themselves. The mental model it created was "someone I don't fully trust sent me a payment request."
The fix was in rethinking who initiates and who validates. In every payment interaction customers already trusted (scanning at a kirana store, paying a local vendor), the customer held the QR code. They scanned it. They chose to pay. The payee never pushed a request at them. That's what kept control with the payer.
If the agent could generate a QR code and the customer could scan it on whatever UPI app they already used, the dynamic shifts. The agent stops initiating a pull from the customer's account. The customer makes a payment they've decided to make, to a payee they can read on their own screen.
One more thing worth noting: NPCI later confirmed it would discontinue P2M Collect Request effective Feb 28, 2026. That made the QR shift a strategic necessity, not just a UX call. We didn't build Dynamic QR because of that ruling, but it was the right direction regardless.
What I Did
Designed the agent-side flow within DMS One App. The agent opens a case, enters the outstanding EMI amount, and generates a scannable QR code on their device. The QR page displayed everything the customer needed to verify before paying: the Bajaj Finance logo, UPI app logos, the customer's name, customer ID, loan account number, amount to be paid, and a unique transaction ID. No ambiguity about who the payment goes to, for how much, or for which loan.
Designed the customer-side experience to require nothing new. The customer scans with PhonePe, GPay, Paytm, or their bank app. They see "BAJAJ FINANCE LTD." as the payee and the pre-filled amount. They authorize. No app to download, no unfamiliar interface, no incoming request to accept or reject.
Ran Dynamic QR alongside Collect Request rather than replacing it. Agents could offer whichever mode worked for the customer in the moment. This also gave us a read on comparative adoption before Collect Request was wound down.
What Changed
| Metric | Before | After |
|---|---|---|
| Monthly digital payment volume | ₹1 Cr | ₹82 Cr |
| Timeframe | -- | 18 months post-launch |
Monthly digital payment volume went from ₹1 Cr to ₹82 Cr over 18 months, with no change to the customer base, the agent team, or the collection process. Only the payment format changed. The customer went from approving a request pushed to their phone to scanning a code on their own terms.
What I'd Do Differently
We had data showing Dynamic QR was outperforming Collect Request on adoption fairly early. We still ran both modes in parallel longer than we needed to. If I were sequencing this again, I'd have moved faster to sunset Collect Request for agent-assisted cases once the signal was clear. We waited for more certainty than the data actually required.