Account Aggregator for Lenders: Faster Credit Decisions
Account Aggregator for Lenders: Turning Consented Bank Data into Faster Credit Decisions
Last reviewed: November 2026
Account aggregator for lenders means using an RBI-licensed intermediary to pull a borrower's bank and other financial data, with the borrower's digital consent, straight into your credit systems. It replaces uploaded PDF statements and screen-scraping with signed, structured data.
This guide is for CEOs, CROs and heads of underwriting at NBFCs, banks and fintech lenders. It covers how the data flows, where it improves decisions, what an integration involves and how to make the business case.
Key Takeaways
- Only lenders regulated by a financial sector regulator can receive AA data as Financial Information Users.
- The AA framework moved into the NBFC Account Aggregator Directions, 2025, issued on 28 November 2025.
- Sahamati reported 352.19 million accounts linked and 595.11 million consents fulfilled as of 30 September 2026.
- A lender joining as a user may also have to share data as a provider.
How Does the Account Aggregator Flow Work?
The borrower approves a specific data request on an AA, and the provider sends the data to the lender. The aggregator routes the data but does not keep it.
- Consent request. Your journey asks the borrower to share, for example, 12 months of savings account data for loan assessment.
- Approval. The borrower selects accounts and approves the consent on the AA's app or web flow.
- Fetch. The provider, usually a bank, checks the consent and sends digitally signed data.
- Delivery. The data reaches your systems through the AA, encrypted in transit.
- Use. Your rules and models analyse it, within the purpose stated in the consent.
Each consent artefact records the customer, the information type, the purpose, the recipients, a notification address, and creation and expiry dates. It must be capable of being logged, audited and verified.
How Mature Is the Network Now?
The network is large enough to be a primary channel, not a pilot. Sahamati's dashboard showed the following as of 30 September 2026.
| Metric | Value |
|---|---|
| Cumulative accounts linked | 352.19 million |
| Cumulative consents fulfilled | 595.11 million |
| Accounts linked in September 2026 | 14.16 million |
| Data deliveries in September 2026 | 379.64 million |
Coverage still varies by bank and account type. Your own consent success rate, not the national total, should drive your business case.
Where Does AA Data Improve Credit Decisions?
AA data helps most where income is irregular or documents are slow. The table maps uses to decisions.
| Use Case | AA Data Used | Decision Improved |
|---|---|---|
| Salaried personal loans | Salary credits, EMIs, bounces | Income verification and FOIR |
| MSME and self-employed loans | Inflows, outflows, balances | Cash-flow-based limits |
| Thin-file borrowers | Transaction behaviour | Approval where bureau data is thin |
| Loan stacking checks | New EMI debits, lender credits | Detecting recent borrowing |
| Portfolio monitoring | Periodic fetches with consent | Early warning before default |
Digiqt's alternative data credit scoring AI agent shows how consented transaction data can support thin-file decisions. The loan stacking detection AI agent shows how fresh EMI debits and lender credits can flag recent borrowing.
Is your underwriting still re-keying PDF statements that could arrive as signed data?
Talk to Our Lending Technology Specialists
Visit Digiqt to see how we build loan origination journeys that use consented financial data.
What Does an FIU Integration Involve?
An integration is a product, data and compliance project, not just an API call. Plan for each layer.
| Layer | What You Build | Why It Matters |
|---|---|---|
| Eligibility | Confirm regulated status and AA onboarding | Only regulated entities can be users |
| Consent design | Purpose, data range, frequency and expiry by product | Over-asking lowers approvals and raises privacy risk |
| Connectivity | APIs to one or more AAs, built on ReBIT specifications | RBI expects regulated entities to adopt them |
| Data parsing | Normalise transactions into a standard schema | Underwriting rules need clean categories |
| Decisioning | Feed variables to your rules engine and models | Turns data into limits and pricing |
| Retention | Delete data when the purpose ends | Supports consent limits and DPDP readiness |
| Provider role | Share data you hold as an FIP where required | Directions link FIU and FIP roles |
How Should Consent Differ by Product?
Consent should ask for the minimum data the decision needs. The settings below are illustrative starting points, not regulatory limits.
| Product | Information Type | Data Range | Fetch Pattern | Consent Expiry |
|---|---|---|---|---|
| Salaried personal loan | Savings account | 6 to 12 months | One-time | Short, ending after decision |
| MSME term loan | Current and savings accounts | 12 months | One-time | Short, ending after decision |
| Working capital line | Current account | 12 months | Periodic | Aligned to limit review cycle |
| Loan monitoring | Repayment account | Rolling recent months | Periodic | Aligned to loan tenure, revocable |
Set these choices in a product configuration, not in code. That lets risk and compliance teams change them without a release.
What Checks Should Run on Every Fetch?
Signed data is not the same as complete data. Before the data reaches a model, the system should confirm the account holder's name matches the applicant, flag gaps in the date range, and spot accounts that look inactive.
It should also prompt the borrower when salary or business inflows appear to land in an account that was not shared. Digiqt's income verification AI agent shows how these checks can run without manual review.
Should You Integrate With One AA or Several?
Several aggregators operate, and borrowers may already use one. Connecting to more than one improves reach but adds testing and monitoring work.
A practical route is to start with one or two aggregators that cover your borrowers' main banks. Then add others when drop-off data shows a gap.
What Is the Business Case in Numbers?
This is an illustrative scenario with assumed rates. A lender receives 12,000 salaried applications a month, each needing a bank statement.
With PDF uploads, assume 25% need manual review at 15 minutes each. That is 3,000 reviews, or 750 hours a month.
With AA, assume 70% of applicants complete consent and 30% fall back to PDFs. Only 3,600 PDFs remain, 900 need review, and the effort drops to 225 hours, saving 525 hours a month.
Speed is only part of the gain. Signed provider data removes the edited-statement risk on the AA-sourced share of applications, which matters as much as the hours saved.
What Goes on the Cost Side?
Set the savings against real costs. These include aggregator or technology service provider fees per consent or fetch, which vary by contract, plus build effort, parsing and monitoring.
Ask vendors for pricing by successful fetch, not by request. Then model the break-even using your own volumes, approval rates and fallback rate rather than any vendor benchmark.
The cleanest test is a controlled pilot. Run AA and PDF journeys side by side for one product, and compare approval rates, turnaround time and early delinquency after a few months.
Which Metrics Show Whether AA Is Working?
Track the funnel from consent request to credit decision. The formulas are simple, but most lenders do not measure every step.
| Metric | Formula | What a Weak Number Signals |
|---|---|---|
| Consent initiation rate | Consent requests ÷ eligible applicants | AA step placed too late or explained poorly |
| Consent approval rate | Approved consents ÷ consent requests | Too much data requested, or a confusing AA screen |
| Fetch success rate | Successful fetches ÷ approved consents | Provider downtime or unsupported banks |
| Fallback rate | Applicants routed to PDFs ÷ eligible applicants | Coverage gaps by bank |
| Time to decision | Consent approval to credit decision | Slow parsing or manual review |
Review these by bank and by channel each month. A fall in one provider's fetch success rate is usually an integration or downtime issue, not a borrower problem.
Which Mistakes Weaken AA Programmes?
Most weak results come from journey design, not the framework.
- Asking for too much data. Long date ranges and many information types reduce consent completion.
- No fallback path. If a borrower's bank is not live, the journey should switch smoothly to another method.
- Using data beyond the consent. RBI bars users from using information beyond what the consent artefact allows.
- Forgetting revocation. Monitoring built on periodic fetches must cope with consent being withdrawn.
What Is Changing Next?
The legal base now sits in the NBFC Account Aggregator Directions, 2025, so update policy references from the older master direction. Most obligations under the DPDP Rules apply from May 2027, and consent records from AA journeys will support that work.
Lenders should also expect more account types and providers on the network over time. Track coverage through Sahamati's published metrics, and use the AI agents in loan underwriting approach to turn new data into consistent decisions.
Could your credit team approve more applicants if cash-flow data arrived in minutes?
Talk to Our Lending Technology Specialists
Visit Digiqt to see how we build lending platforms and AI agents that put consented data to work.
