CAM Automation Software in India: A Buyer's Guide for Banks and NBFCs

Most platforms sold as CAM automation software in India do not write the credit appraisal memorandum. They produce CAM-ready data and leave the document to your analyst. Of the seven platforms here, three publish a claim to generate the memo narrative itself. Sort your shortlist on that question before you discuss price or integration.

Disclosure first: YuSight is our product and publishes 5x underwriting throughput at the same headcount. We have not ranked it first on this page. On India document breadth, Perfios is plainly ahead of us, and for two of the buyer situations below we recommend somebody else.

This page is deliberately narrower than our general India credit appraisal software guide. That one covers origination, extraction and analysis layers across thirteen vendors. This one asks a single question of seven: what does the platform hand you at the end — a dataset, a template, or a drafted and cited CAM?

Key facts

• The RBI's FREE-AI Committee report, published 13 August 2025, sets out 7 "sutras" and 26 recommendations across six strategic pillars for responsible AI in the Indian financial sector (RBI Press Release, 13 Aug 2025). Any AI-drafted CAM you deploy in 2026 will be read against that framework.
• Under the RBI (Digital Lending) Directions, 2025, outsourcing "shall in no manner dilute or absolve the RE of its obligations" (para 5(vii)), and "all data is stored only in servers located within India" (para 13(iv)) (RBI/2025-26/36, 8 May 2025). If a vendor drafts your memo, the appraisal is still yours to defend.
• Only three of the seven platforms here publish a claim to generate the CAM narrative. Two publish "CAM-ready" data outputs and explicitly leave the document to the lender. Two describe CAM as a step inside a lending workflow.
• Not one of the seven publishes a price. Any per-file rate you have been quoted is a specific deal, not a rate card.
• YuSight publishes 5x underwriting throughput at the same headcount, a ~30-minute end-to-end CAM draft, 95.2% extraction accuracy against a manual benchmark, and 100% of figures cited to source document and page.

How was this comparison built?

Stated plainly so you can discount it appropriately.

What was compared: publicly published capability for producing an Indian credit appraisal memorandum — document intake and India document coverage, spreading, ratio work, cross-source triangulation, whether the platform drafts the memo narrative, whether figures are cited, editability and audit trail.

Sources: each vendor's own website only, fetched 28 August 2026. Analyst grids, review-site rankings and third-party press were excluded, because they carry claims the vendor has not committed to in public. Where a vendor's own press release is the only public statement of a capability, it is cited as a press release and labelled as such.

What we could not verify, and did not guess:

• Pricing. None of the seven publishes it. Perfios publishes package names — three retail tiers and four SME tiers, including a Cashflow package — with no numbers attached.
• Accuracy methodology. Three vendors publish an accuracy percentage. None publishes the test set, document mix or benchmark behind it. Treat every accuracy figure on this page — ours included — as a claim to be tested on your own files.
• Citation depth. This is the sharpest gap in the category. Several vendors describe outputs as traceable or grounded; only some publish a per-figure citation claim. We have marked exactly what each one says rather than reading intent into it.
• Newgen's CAM position. Newgen's commercial lending page names underwriting, risk rating, DSCR, cash flow analysis and covenant tracking, but does not name credit memo or CAM generation as a discrete capability.

Conflict of interest: YuVerse publishes this page and sells YuSight.

What does "CAM automation" actually mean when a vendor says it?

Three different things, and the ambiguity is where money gets wasted. The Indian credit appraisal process runs proposal → document collection → verification → spreading and triangulation → appraisal note → sanction. "CAM automation" is sold against three different stretches of that.

None of the three is wrong. Buying the third when you needed the first is what leaves the appraisal note still being typed into Word at 9pm. If you are not yet sure which layer you are short of, the credit appraisal note format is a useful test: print your last five notes and count how many minutes went on data versus judgement.

Which platforms generate the CAM narrative itself?

1. Perfios CAM AI — the widest India data coverage on this list

We will say this plainly: Perfios is the closest direct competitor to us in India, and on breadth of Indian data sources it is ahead of us.

Perfios CAM ingests through "ePDF uploads, scanned documents, online data fetch, and AA frameworks", and cross-analyses bank, GST and ITR data across sources. Its published data set is unusually deep for the Indian market: bank statements, commercial and consumer bureau, PAN and OVDs, payslips, Form 16, Form 26AS, ITR and ITR-V, GST, financial statements, stock statements, EPFO data, MCA documents, litigation data and statutory compliance alerts. On the fraud side it identifies document metadata "to detect tampering" and runs multi-document reconciliation for discrepancies. Outputs include a profile summary note, a personal discussion sheet, repayment track record analysis, strength and risk indicators, and a lender policy check with eligibility calculation, customisable "aligned with various BREs and credit policies" (Perfios CAM).

Published outcome claims on that page: 80% reduction in CAM preparation time, 50% increase in credit manager efficiency, 90% reduction in dependency on CPA and 5% reduction in fraud losses.

Perfios launched CAM AI on 12 August 2025, claiming it "reduces underwriting turnaround time by up to 85%", compresses "multi-day credit assessment processes into just hours", and lets lenders "process up to 2X more loan applications with existing teams". Perfios states deployment across 50+ institutions in India, Southeast Asia and MENA, processing over 5 million transactions monthly, built on domain-specific LLMs, RAG pipelines and agentic algorithms. Its CTO is quoted: "Every insight is backed by source traceability, analysts can see exactly which document led to which conclusion" (Perfios press release, 14 Aug 2025).

Buy it if your gap is India document breadth — EPFO, MCA, litigation and statutory data alongside the usual GST/ITR/bureau set — and you want intake, fraud checks and CAM from one vendor. Press it on what "source traceability" resolves to on screen. Ask to click ten figures in a generated CAM and see where they land.

We compare the two products on the spreading layer specifically in Perfios vs YuSight for financial spreading, and cover the wider field in Perfios alternatives for credit analysis.

2. GLIB.ai — memo generation built outward from the financial statement

GLIB publishes three analysers — Bank Statement Analyzer, Financial Statement Analyzer and Invoice Analyzer — over a "Financial Intelligence Engine", with human-in-the-loop review workflows, escalation handling and audit trails. Published claims: 99%+ classification and processing accuracy, 99.4% extraction accuracy, 7x faster processing cycle and 95%+ reduction in average handling time, with named Indian institutions including ICICI Bank, Bank of Baroda and TATA Capital (GLIB.ai).

Its Financial Statement Analyzer turns "messy financial statements into spreads, scorecards, and memos in minutes", supports bring-your-own taxonomy for chart of accounts and scoring models, handles 25+ languages, and offers "1-Page Memo Generation" (GLIB.ai FSA). The dedicated credit memo product drafts an executive summary, financial analysis, risk assessment and credit recommendation, exports to Word or PDF, and states outputs are "grounded in accurate spreading data, not hallucinations", against a published manual baseline of "2-4 hours per credit memo" (GLIB.ai AI Credit Memo Generator).

Note precisely what is and is not claimed. GLIB publishes grounding in spread data; we did not find a published per-figure citation-to-page claim on those pages. That is a demo question, not a criticism.

Buy it if your appraisal turns mainly on audited and management financial statements, and you want a one-page memo fast. Press it on GST, ITR and Form 26AS depth for MSME files, where the Indian appraisal usually lives.

3. YuSight — the CAM layer alongside your existing LOS

Ours, listed third because on India data breadth Perfios is ahead and on statement-first memo speed GLIB has a strong case.

YuSight runs Document Intelligence (classification, validation and multi-entity mapping), Financial Spreading, modular analysers including a Repayment Tracker that compares bank statement debits against bureau-reported facilities, and CAM Generation with full version history. Published figures: ~30-minute end-to-end CAM draft, 95.2% extraction accuracy against a manual benchmark, 100% of figures cited to source document and page, 5x throughput at the same headcount, and a sample CAM completed in 28 minutes carrying 142 citations.

Buy it if you already own an LOS and an analyser, and the remaining bottleneck is the appraisal note itself — particularly on multi-entity group files where documents must land on the right borrower before anything else is true. Do not buy it as an LOS replacement. It is not one, and we cover why in does credit memo automation replace your LOS or sit alongside it.

Which platforms treat the CAM as a step inside a lending platform?

4. Intellect Design Arena (eMACH.ai Commercial Loan Originations)

The most explicit CAM claim from an enterprise lending vendor on this list. Intellect's commercial loan origination page names "standardizing Credit Assessment Memos with auto-generated insights and recommendations", alongside AI-first financial analysis that "auto-ingests and reads data from audited financials, tax systems (GST), and bank statements to perform ratio and trend analysis", a 360-degree group exposure view, automated covenant validation and deviation handling. Published metrics: 60% manual effort reduction in financial analysis, 40% reduction in origination TAT in a YES Bank case study, 20% reduction in past-due loans, and a composition of 10 packaged business capabilities, 20 microservices and 50 APIs (Intellect Commercial Loan Originations). At group level Intellect publishes 500+ financial institutions worldwide across 61+ countries (Intellect).

Buy it if you are replacing commercial origination rails and want the CAM standardised inside the same platform. This is an enterprise programme, not a quarter's project.

5. Newgen

Newgen's commercial lending solution claims to "automate up to 70% of commercial lending journey", with "intelligent underwriting & risk rating with stress testing, DSCR, cash flow analysis", "collateral and covenant management integrated into workflows", and portfolio monitoring with covenant tracking and ratio analysis. Published outcomes: 30% faster loan processing, 40% lower operational costs, 25% increase in customer insights (Newgen Commercial Lending).

Credit memo or CAM generation is not named as a discrete capability on that page. Newgen may well do it inside its workflow product; it does not publish it there, so we will not write the claim for them. If a drafted CAM is what you are buying, ask for that specific demo.

6. Lentra

Lentra publishes GoNoGo® origination, 1LMS, MultiBureau®, BREx, AAConnect®, CoLending, Collateral Management, Document Management and a Bank Reconciliation System, with platform claims of "End-to-end decisions in 10 secs", "90%+ straight through processing", 3x faster turnaround and 25% savings in bureau spend (Lentra).

CAM generation is not mentioned. Those figures describe rule-driven, small-ticket journeys — and Lentra does not claim otherwise. A ₹35-crore working capital appraisal is not a ten-second decision, and no honest reading of that page suggests it is.

Which platforms produce CAM-ready data but leave the memo to you?

7. ScoreMe

ScoreMe publishes eight analysers — Bank Statement, Financial Statement, GSTR, ITR & 26AS, Bureau Data, Legal Data, KYC & Compliance, and an Integrated Analysis Solution — with stated scale of 200+ financial institutions, 30L+ Cr AUM managed and 500+ Cr transactions processed (ScoreMe). Its Financial Statement Analyzer publishes 300+ financial data points, 600+ trusted data sources, 100+ industries and 2L+ companies analysed (ScoreMe FSA).

ScoreMe is candid about its own position. Its own writing states that "by triangulating critical data streams, ScoreMe automates the heavy lifting of CAM generation" and provides "CAM ready, structured outputs" that "plug into a financial institution's LOS" (ScoreMe). The human underwriter still writes the CAM.

Buy it if your data layer is the weak link — this is a strong Indian extraction stack, and the three-way triangulation of GST, ITR and bank turnover is exactly what it is built for. Do not buy it expecting a drafted memo, because it does not claim to write one.

Capability comparison

A worked example: what "5x throughput" survives contact with arithmetic

Take an eight-analyst credit team writing MSME and mid-market working capital appraisals. Assume 140 productive hours per analyst per month — 1,120 hours — and 390 minutes (6.5 hours) per file end to end.

Current capacity: 1,120 × 60 ÷ 390 = 172 files a month.

Decompose the 390 minutes:

• Document collation, classification, entity mapping: 55 min
• Extraction and keying into the spread: 130 min
• Ratio computation and GST/ITR/bank cross-checks: 70 min
• Drafting the CAM: 95 min
• Reviewer tie-out: 40 min
• Total: 390 min

Now apply a platform that touches all five stages:

• Collation and mapping: 55 → 5
• Extraction and keying: 130 → 15 (residual exception handling)
• Ratios and cross-checks: 70 → 12
• Drafting: 95 → 45 (analyst edits the draft and writes the judgement sections)
• Reviewer tie-out, with one-click citations: 40 → 20
• New total: 97 min

Throughput multiple = 390 ÷ 97 = 4.0x. New capacity = 1,120 × 60 ÷ 97 = 692 files a month.

Note what that says about our own headline. On this decomposition the honest answer is 4.0x, not 5x — a published 5x implies a file with a larger extraction share than this one, or a shorter reviewer step. Ask us, and every vendor here, which file mix the multiple was measured on.

Now the pessimistic case. Suppose only drafting is automated and everything else stays manual: 390 → 340 minutes. Multiple = 390 ÷ 340 = 1.15x. Same headline capability, one-quarter of the benefit.

The entire difference is how many of the five stages the platform actually touches. That, and not the percentage on the homepage, is what your proof of concept must settle.

What to put in the contract, not the sales deck

• India data residency, in writing, per para 13(iv) of the Digital Lending Directions, 2025.
• Audit rights over the vendor and its sub-contractors — the RBI (Outsourcing of Information Technology Services) Directions, 2023 require the agreement to carry the "right to conduct audit of the service provider (including its sub-contractors) by the RE, whether by its internal or external auditors" (clause 16(k)), and a "clear exit strategy" for the outsourced service (clause 22) (RBI/2023-24/102).
• Model documentation you can hand an inspecting officer, in line with the FREE-AI framework's governance pillar.
• Version history on the memo — who changed which figure, when, and from what. Our fuller demo checklist is in what questions to ask a credit memo automation vendor, and the RBI compliance ground is in the RBI digital lending guidelines checklist.

FAQ

Which platforms automate CAM preparation for Indian banks and NBFCs?

Perfios CAM AI, GLIB.ai and YuSight publish claims to draft the memo itself. Intellect standardises the CAM inside its commercial origination platform. ScoreMe and Lentra produce structured data that a human then writes the CAM from. Newgen does not name CAM generation on its commercial lending page.

What is credit memo software?

Software that turns a borrower's document set into the written appraisal a credit committee reads. The good ones spread the financials, compute your ratios to your definitions, reconcile bureau against bank data, and draft the narrative with every figure traceable to a source page. The weaker ones just populate a template.

Does CAM automation replace our LOS?

No. An LOS owns application capture, the approval matrix, sanction and disbursement. A CAM layer owns spreading, ratio work and the memo. They are different budgets and different implementation timelines, and any vendor telling you one substitutes for the other is describing a two-year programme as a three-month one.

How much does CAM automation software cost in India?

Nobody on this list publishes a price. Pricing in this market is typically per file, per seat, or a platform fee plus consumption, and it moves a long way on volume commitment. Ask for the pricing model first — a per-file rate behaves very differently from a platform fee when your book grows.

Is an AI-drafted CAM acceptable to the RBI?

The RBI has not prohibited it, and the FREE-AI framework published in August 2025 is broadly enabling. But the Digital Lending Directions are clear that outsourcing does not dilute the regulated entity's obligations. In practice that means you must be able to explain and evidence every figure in the memo yourself.

Can these platforms handle multi-entity group borrowers?

This is the sharpest differentiator and the one most demos skip. Bring a real group file — a holding company, two operating entities and a guarantor — and check whether every document landed on the correct entity without a human quietly fixing it, and whether intercompany sales were eliminated before consolidation.

Do we still need a bank statement analyser if we buy CAM automation?

Sometimes not, but check what the CAM tool is actually built around. A memo generator built outward from financial statements is usually shallower on transaction-level bank analysis than a dedicated analyser. Run the same borrower through both and compare the cash flow output line by line.

What should we test in a proof of concept?

Five of your own files, not the vendor's demo set. Include one group structure, one photocopied scan, one file with a deliberately missing month of bank statements, and one where you already know a facility is invisible in the bureau. Then click ten random figures in the output and see where they land.

Key takeaways

• Ask one question first: does the platform hand you a dataset, a populated template, or a drafted memo? Three of the seven here draft.
• Perfios has the widest published India data coverage — EPFO, MCA, litigation and statutory data alongside GST, ITR and bureau — and is the closest direct competitor to us in this market.
• GLIB.ai is strong where the appraisal turns on financial statements; ScoreMe is strong feedstock and honest that it does not write the memo.
• Intellect names CAM standardisation explicitly; Newgen and Lentra do not publish a CAM generation claim, so do not assume one.
• The published throughput multiple depends entirely on how many workflow stages the tool touches — 4.0x when it touches five, 1.15x when it touches one.
• No vendor publishes pricing. Contract for India data residency, audit rights over sub-contractors and an exit strategy at signature.

See your first CAM in 30 minutes — [book a live demo](https://yuverse.ai/yusight). Bring your hardest multi-entity MSME file and check the citations yourself.