Retail
Unsecured consumer and card lending.
PersonalConsumer durableVehicle loanEducation
Digital Lending
Originate, decide, service and collect on one data layer.
Trustt Digital Lending is a loan origination and loan management system for Indian banks and NBFCs that runs origination, credit rules, servicing, collections and co-lending on one data layer.
Four pillars on a single applicant context. For each one: what it does, what it writes back, and what it leaves in the audit trail.
01
Origination
Conversational intake. Every field carries its source.
ChatVoiceAA prefill
02
Decisioning
Explainable in about five seconds, reasons attached.
200+ variablesReplayableBRE
03
Servicing
The full post-disbursal life, no end-of-day blackout.
RepaymentCo-lendingGL
04
Collections
The call happens. The promise is written back.
AI callingPTP captureGuardrails
The shape of the platform
One applicant context, end to end
Every stage reads and writes one data layer under a single applicant identifier. Nothing is re-keyed at a handover, the audit trail stays continuous, and identity runs underneath all five as a shared rail.
Step 101 / 05
Origination
Intake is a conversation in the customer’s language. Aadhaar returns identity, the Account Aggregator returns income, document intelligence reads the statements, and every field arrives with its source.
Step 202 / 05
Decisioning
The engine evaluates 200+ variables against your policy and returns an approve, a price and a structure in about five seconds, with the reasoning written at decision time.
Step 303 / 05
Disbursal
Payout, eNACH mandate, schedule generation and opening the loan in the LMS happen as one movement. The idempotency key is required.
Step 404 / 05
Servicing
The full post-disbursal life on the same loan object: repayment in every shape Indian lending uses, restructuring, accounting and co-lending settlement, in real time.
Step 505 / 05
Collections
Signals surface stress before an EMI is missed, the agent places the call, and the promise is written back to the book validated against what was said.
LOS · Loan origination
How an application assembles itself
Origination runs as a conversation: chat, voice or web, in the language the customer speaks. Four rails answer underneath it, and the application is assembled from what comes back.
What answered
Aadhaar eKYCidentity
Account Aggregatorincome
Document intelligencestatement · GST · ITR
Bureauobligations
Identity
Verified
AADHAAR · 0.99
Monthly income
₹84,200
ACCOUNT AGGREGATOR · 0.98
Bank statement
12 months
DOC INTELLIGENCE · 0.97
GST turnover
₹38.4L FY
DOC INTELLIGENCE · 0.94
ITR income
₹9.8L FY
DOC INTELLIGENCE · 0.96
Obligations
2 active
BUREAU · 0.99
every field carries source, timestamp and confidence
The journey behind it is yours: orchestration, parallel checks, deviation and approval matrices, configured by your team in an afternoon, without a vendor release.
Applications are lost between sessions. Three agents work that gap.
01
Drop-off recoveryStalled
Re-engages in the channel it stopped in, within minutes rather than the next working day.
02
Offer activationPre-approved
Works a pre-approved list conversationally, so reach is not capped by branch calling capacity.
03
Application assistMid-journey
Answers eligibility and document questions inside the journey.
BRE · Business rules engine
How a rule is authored,
tested and shipped
A rule cannot reach production until it has passed a backtest and an A/B test.
Written by the
credit team
Plain language compiles to executable policy. The grammar is deliberately narrow, which is what keeps the backtest trustworthy.
four gates · one workflow
credit owns the rule, engineering owns the rail
credit owns the rule, engineering owns the rail
written by the credit team
Decline if GST turnover has fallen more than 30% quarter on quarter.
compiled, backtested and live the same afternoon
four gates
Author
Plain language, by the credit team. No ticket.
Compile
To executable policy, in a narrow grammar.
Prove
Backtested on your book, then A/B against live.
approval −1.8ppShip
Same day, with author and approver on the record.
rollback in one actionCredit decisioning & accountability
Every decision
can be run again.
The reasoning is written at decision time, so the record of what happened never depends on interrogating a model afterwards.
decision · app 48291
Approve · ₹30L
14.5% · 36m
14.5% · 36m
policyretail-pl-v18
variable setvset-2026-06
inputssource-tagged
reasonswritten at decision time
- Approve, price and structure in about five seconds~5s
- 200+ variables across six families200+
- Reason codes on every decision, declines included100%
- Every policy change carries author, backtest and approverPROVENANCE
- Replay re-executes against frozen stateREPLAY
Shared rail · Identity
Identity runs inside the spine
KYC, KYB, eSign and verification run inside the same system that decides and services the loan, so the consent artefact never lives apart from the decision it belongs to.
Aadhaar eKYCverified · 0.99
Purposecredit appraisal
Retentionper policy, vault ref
> one record, one party answerable
Aadhaar eKYC
eSign
Vault
Liveness
KYB
Shared rail · Disbursal
From approval to payout, as one movement
The idempotency key is required. A duplicate disbursal should never reach the book.
Approveddecision object written
Paid outIMPS, real time
MandateeNACH registered
Scheduledrepayment plan generated
Open in LMSservicing begins
disbursal.initiateddisbursal.settledmandate.registered
LMS · Loan management
What the LMS runs after disbursal
Most of a loan’s life happens here. It reads the same loan object origination wrote and collections reads, so a restructure is visible immediately, with no overnight reconciliation between systems that can disagree in the interval.
Repayment
Every shape Indian lending uses
EMI, bullet, step-up and down, moratorium, irregular and seasonal for agri.
Lifecycle
Each event with its actor
Reschedule, top-up, part-payment, foreclosure, write-off, recovery, all timestamped.
Accounting
Posted to the GL
Automated JE posting and reversal, provisioning, trial balance, GL reports.
Co-lending
Settled per partner
Configurable splits, partner settlement, per-partner GL segregation.
Statutory
Generated on schedule
Notices and statements produced without an operations ticket.
Availability
No end-of-day window
Real-time processing, horizontal scale, servicing stays up while the book closes.
LCS · Loan collections
Strategy escalates.
The write-back is validated.
What the agent captures is checked against what was actually said before it reaches the book.
Calling windowsDND and retry capsGrounded factsHuman handoff
> No configuration re-enables generation on late-bucket calls.
Early bucket
ReminderPre-due to 30 DPD. Payment link, no pressure.
Mid bucket
NegotiationPromise-to-pay inside approved bands.
Late bucket
Scripted onlyA reviewed script. The model does not generate.
Write-backVALIDATED
promise to pay₹12,4005 Sep 2026 · confirmed twice, on the call
Products
14 products. One core.
Four lending books on the same origination, rules, servicing and collections layer. A new product is a configuration.
MSME
Working capital and business lending, on cash-flow data.
Working capitalBusiness loanSupply chainEquipmentInvoice
Secured
Collateral-backed, with valuation and title in the journey.
HomeLoan against propertyGoldLoan against securities
Rural & inclusion
Group, agri and assisted lending in local languages.
Microfinance and JLG
Rich banking APIsLow-code journeysWorkflow orchestrationAnalytics and reportingeKYC and e-signHorizontal scale
> $1BN+ disbursed · 100+ institutions · launches in weeks
Technical surface
The API surface
Failure modes are documented beside the success paths. A bureau or Account Aggregator timeout returns a typed 424 naming the dependency, and the journey follows a configured fallback: proceed on available data, queue, or route to review.
core endpoints
method
endpoint
what it does
POST
/applications
create an application
POST
/underwriting/decision
decide, with reasons attached
GET
/underwriting/decision/{id}/replay
re-execute against frozen state
POST
/policy/rules
author, backtest, ship
POST
/disbursal
idempotency-keyed payout
GET
/loans/{id}
full servicing state
POST
/collections/strategy
allocate and dial
GET
/kyc/{id}
consent artefact, vault reference
Webhooks
application.* · decision.* · disbursal.* · repayment.* · collection.* · policy.*
Errors
typed · 409 idempotency conflict · 424 upstream dependency · 5xx ours, named as such
Versioning
dated, with deprecation notice before removal
> Fallback is configuration, because what a lender wants at a bureau timeout depends on the product.
Implementation
End-to-end, in four steps
No heavy IT lift and no multi-quarter integration programme. A typical bank or NBFC goes from kickoff to a production pilot in a month.
Step 1
Discover & set up
Environment provisioned, current products and credit policy mapped.
Step 2
Configure
Products, workflows and decision rules configured in LOS and BRE.
Step 3
Integrate
Core banking wired via APIs.
Step 4
Go live
UAT sign-off, pilot cohort onboarded, and production launch.
Frequently asked
The questions every evaluation starts with.
01.How does explainable underwriting work?
Each decision returns reason codes: which variables moved the outcome, in which direction, with what weight, against which policy version. Any decision replays against the policy exactly as it stood at the time.
02.How fast can a credit team change a rule?
Same day. Authored in plain language, compiled, backtested on your own book and A/B tested on a live traffic slice. Every change carries an author, an approver and a one-action rollback.
03.Can the LOS or the LMS be taken on its own?
Yes. All four run individually or together on one data, identity and rules layer, so taking more than one removes integration work.
04.How are RBI conduct rules enforced?
Calling windows, DND and retry caps are enforced in the platform and checked before dial. Late-bucket and settlement calls read legally reviewed scripts rather than generating language.
05.What happens at a bureau or Account Aggregator timeout?
A typed 424 naming the upstream dependency, then the fallback you configured: proceed on available data, queue for retry, or route to manual review.