How a NBFC Qualified 3× More Leads Without Hiring a Single Agent

In January 2026, a mid-sized NBFC based in Pune — offering personal loans and business credit to salaried and self-employed individuals — was running 400 outbound calls per day with an 8-person inside sales team. Their lead qualification cost was ₹1,480 per qualified lead. Their best month had produced 112 disbursals. They were stuck.

By March 2026, the same team was running 1,800 calls per day, managing the pipeline with 2 human agents, and producing 310 disbursals per month. Their lead qualification cost dropped to ₹390. This is the detailed story of how that happened.

The Problem Before Sreegen

The NBFC (name withheld at their request) was aggregating leads from three sources: a digital lending marketplace, Facebook/Instagram lead forms, and Google Ads landing pages. Lead volume wasn't the constraint — they were receiving 600–800 leads per day across these channels.

The constraint was qualification capacity. Their 8 callers could handle roughly 400 calls per day. That meant 200–400 leads per day went uncalled for 24+ hours. In personal loan intent, 24 hours is an eternity — the lead has already taken a loan from someone else or mentally moved on.

The second problem: script consistency. RBI's Fair Practices Code has specific requirements for how lending institutions can present loan terms. Manual callers were improvising, particularly under pressure to hit conversion targets. The compliance risk was real.

The third problem: night leads. Digital ads running past 9pm were generating leads that weren't touched until 9am the next morning. In some cases, these were customers with urgent liquidity needs who were already looking at three other options by morning.

The Decision to Try AI Calling

The CFO's initial concern was compliance. Could an AI agent accurately represent loan terms? What happens if it says something incorrect about interest rates or processing fees?

The answer was structured AI: the agent doesn't calculate or quote loan terms. It qualifies the lead (income, employment type, existing EMIs, requested loan amount) and books a callback with a human underwriter who gives the actual quote. The AI's job is gating, not selling. This is fully compliant and in fact more compliant than human callers who improvise.

The second concern was language. Their leads came from Maharashtra, Gujarat, and Karnataka in roughly equal proportions. A Marathi-speaking farmer in Nashik applying for a business loan needs to feel understood, not foreign. Multilingual capability was non-negotiable.

The Implementation

Week 1: Script development. The Sreegen onboarding team worked with the NBFC's compliance officer to build a script that met RBI Fair Practices Code requirements. Key qualification fields: monthly income, employment type (salaried/self-employed), existing loan EMIs, requested amount, and purpose. The script was reviewed and signed off by their in-house legal counsel before going live.

Week 2: Pilot with 500 leads across all three language groups (Hindi-primary, Marathi-primary, Gujarati-primary). Call recordings reviewed by compliance team. Minor script adjustments. Contact rate baseline established: 58% on Hindi leads, 62% on Marathi, 54% on Gujarati.

Week 3: Full go-live. Lead flow connected via webhook from all three lead sources to Sreegen API. Leads now called within 90 seconds of submission, 24/7. Campaign concurrency set to 40 simultaneous calls with auto-scaling based on queue depth.

Week 4–6: Iteration. Script optimised based on sentiment data. Discovered that Gujarati leads had significantly higher qualification rates but lower contact rates — identified that most Gujarati leads came from a specific marketplace that attracted business loan intent (higher value). Adjusted concurrency priority for this source.

The Numbers After 60 Days

What the Two Remaining Agents Do

The 6 agents who were previously doing cold first-touch calls weren't laid off — 4 moved to Sreegen-qualified callbacks (warm conversations with people who have already expressed intent and have been pre-qualified). Two moved into other roles within the organisation. The callbacks are fundamentally different conversations: shorter, higher intent, better conversion. The agents themselves report significantly lower call fatigue.

The two agents managing the AI campaign spend their time: reviewing flagged calls (edge cases the AI passes to human review), monitoring live dashboard metrics, and iterating on the script based on weekly data reviews. It's an analyst-type role, not a dialler role.

The Compliance Outcome

Before Sreegen, the NBFC's monthly audit flagged an average of 23 calls with some form of compliance deviation — incorrect interest rate framing, pressure tactics, missing mandatory disclosures. After switching to AI-first calling, that number dropped to zero. Every AI call includes the mandatory disclosures at the required points in the conversation. No improvisation, no omission.

Lessons for Other NBFCs

From this deployment and others like it, here's what makes financial services AI calling work:

1. Get compliance sign-off on the script first. Don't launch until your legal team has reviewed and approved. The compliance advantage of AI only realises if the script itself is compliant.
2. Separate qualification from quoting. The AI qualifies. Humans quote. This keeps the AI in its strength zone (structured conversation, consistent delivery) and keeps your underwriters in their strength zone (nuanced pricing, objection handling on specifics).
3. Invest in language quality. Financial conversations require trust. A Marathi speaker getting a Marathi-language call from an AI that sounds natural will engage longer and disclose more relevant financial information than one navigating a Hindi or English call.
4. The human team gets better jobs. Frame this internally as a team upgrade, not a headcount cut. The agents doing warm callbacks are doing more interesting, higher-value work with better outcomes. Retention actually improves.