The engine underneath the DARP framework, FSCP and everything that follows
Not a separate system for every problem. Each one runs the same framework: Discover, Assess and Recover to plug the leakage first, then the same rules go into Prevent to stop it coming back. The engine wraps the systems you already run. Nothing migrated, nothing switched off.
End to end
The DARP framework and FSCP do not have separate building blocks. They are different questions asked of the same engine. Every problem is configuration on that engine.
The shared engineShared by every product.
One engine, two views of the same object. DataTwin sits between the systems you already run and the ledger they feed. Everything in navy is DataTwin; everything outside it is yours.
your systems around the outside · DataTwin between · your ledger at the centre
your systems at the top · DataTwin between · your ledger at the bottom
Inside DataTwin
Six modules, one shared engine
Accounts Payable
Duplicates, price and quantity variance, tax eligibility, approval limits
Accounts Receivable
Many-to-many receipt matching; withheld tax split from genuine shortfall
Partner Payouts
Payout and settlement tested against the partner agreement
Sales Commissions & Incentives
Plan-based calculation, validated before the payout is released
Channel & Distributor Rebates
Claims re-priced against the terms in force, tested before the credit
Financial Statement Close
8 domains · 56 sub-processes · 204 close-blocker indicators
The shared engine
every module runs the same three stages — the rings in this view, the bands in the other
DARP
How every module is delivered
Discover
Read the whole population out of the systems you run. Read-only.
Assess
Value every finding and rank it by size and by likelihood of collection.
Recover
Package the evidence into claims. The cash comes back.
Prevent
The same rules now run at entry, on every transaction.
Exceptions routed to a named owner — during the period, not at the close.
Systems around the outside
ISO 27001 certified · SOC 2 attested · runs on top of SAP, NetSuite or any ERP, with no migration
Custom build, SaaS or spreadsheet, it does not matter. We plug in, clean up and lock down control. A first pass through the engine returns your number in about two minutes.
Stage 01
Get every source in, then model it so the pieces can actually be compared. Most reconciliation projects suffer here, mapping based on the system's data model. This is the part we automated first.
Start in minutes
Drag a file in and go. CSVs, PDFs or screenshots. Nothing to configure and connect, every Discover engagement starts here.
For the ongoing run
Live connections into ERP, banking, PGs and portals. Read-only by default; write-back only where enabled, on validated data.
For recurring volume
Scheduled feeds on a cadence: distributor POS files, bank statements, PG settlements, vendor returns. Processed automatically.
AI-native
AUDERE, our Autonomous Data Engineering & Recon Engine, reads a source it has never seen, infers what each field means and builds the model to acquire it, then keeps that model current as the source changes. No mapping spreadsheet, no integration ticket.
Stage 02
Turn raw records into a valued finding during discovery, a KPI score and an anomaly for financial close, a pass-or-fail at entry for Prevent. Transformation, testing and routing are all one continuous pass.
Schema Controlled Data Processing
The schema controls what happens to the data. Every reshaping step is declared once and applied everywhere. So are the rules.
Part of the DARP framework
Where the matching happens. Two-way through N-way, at whatever granularity the question needs.
Any shape you need
Serial, parallel or any combination. Approvals, reviews, escalations, branches and rework loops, configured to your organisation's needs.
AI-native
Plenty of reconciliations do not resolve on rules: identifiers disagree, or the fact that settles it is a clause inside a contract. DataTwin applies reasoning to exactly those cases, and builds the schema itself from what you describe. Everything that posts stays deterministic.
Stage 03
Each product asks a different question of the same engine, so each gets its own reporting. Every number drills through to the transaction and to the document that supports it.
What is recoverable, and where each finding stands
Recovery position
Current engagement
₹4.82 Cr
Total identified across 1.2M transactions tested
Where each finding stands
By value
₹1.64 Cr
Cash returned to date
21 days
Median claim to settlement
Illustrative dashboards. Every figure shown is an example, not a customer result.
Reports are configured to the process, not to a template. Tell us what you need to see.
AI-native
Not a chatbot bolted onto a reporting tool. Three points where the platform would not function the same way without it, and one place we deliberately keep it out. The full architecture, from the agents to the models and the reasoning engine, is on the How AI is used page.
01
A source we have never seen is read, understood and modelled on arrival: fields inferred, relationships worked out, the acquisition model built without a mapping exercise.
02
Describe what needs reconciling against what. The engine elicits the detail and constructs the schema, with the rules and checks embedded inside it, for entities it has never handled.
03
Where identifiers disagree, evidence sits in a document, or the answer depends on context across several records, reasoning resolves it, faster and more accurately than a manual queue.
04
Anything that becomes an entry in your books is governed by explicit, inspectable rules. Nothing posts because a model was confident. That boundary is deliberate and it is not moving.
Across the platform
Read-only by default. ISO 27001 certified and SOC 2 attested, with role-based access, residency options and an immutable audit trail. AWS, Azure, GCP or your own cloud.
Read moreThe full account of where AI earns its place: document understanding, source modelling, schema construction, reasoning on hard matches, natural-language query. And where deterministic rules run instead.
Read morePrefer to see the engine working rather than described?
