Aurionpro Solutions

A diversified group whose largest share survives the removal of every model. Strip the AI and SmartLender still runs the corporate credit lifecycle for the Asian banks that have licensed it for twenty years, and the cash management, treasury, digital banking, payments and transit hardware businesses are untouched. The Euromoney citation states the ordering plainly, describing a broad lending suite with a serious push into AI.

Applying the share test rather than the survival test does not change the answer: the enterprise AI subsidiary and the AurionAI platform would go empty, but the lending suite, treasury, cash management and the entire non financial half of the group would not. Recorded so a later pass can revisit it: the AI subsidiary is genuinely AI native, is separately branded, and sells API services directly to banks and insurers who are buying inference rather than a lending platform. If that line continues to grow as a distinct business, the intellectai precedent applies, where a group's AI arm carries its own index entry separate from the parent.

Oversight is a stated design principle with a real mechanism attached, and the control path is still not described. The AI platform is presented as built specifically for regulated institutions with controls, workflows and domain knowledge embedded, and the explainability engine exists precisely so that model output can be interrogated and audited rather than accepted, which is a genuine contribution to oversight rather than a claim about it.

Agentic workflows are named as a direction of travel. What is missing is the enforcement layer: nothing states a confidence threshold, an escalation route, a human review band, or what happens at an automated credit decline, and nothing describes what constrains an agent once it is running inside a lending workflow.

The most substantive model transparency disclosure encountered in this sweep. The AI subsidiary publishes quantitative, comparative evaluation of its own interpretability method against named academic baselines using recognised evaluation metrics from the explainability literature: better faithfulness than SmoothGrad, Integrated Gradients and GradCam, better delta most relevant first and least relevant first scores than GradCam for transformer models, and better model parameter randomisation test values than SHAP with a stated 40 percent improvement over LIME.

That is measurement against named alternatives rather than an assertion that outputs are explainable, and it is the difference between this grade and the circular definitions catalogued elsewhere in this index. It is backed by product rather than marketing: the explainability engine, model governance tooling and continuous monitoring are sold to banks and insurers precisely so an institution can validate models under supervisory expectations, which is the standing definition of an A on this axis.

Held at A despite the absence of drift figures for individual deployed models, versioning policy and independent third party validation, because the axis rewards enabling the institution to validate rather than being validated by someone else.

Named customers with qualitative results, one quantified result at an unnamed institution, and an exceptional density of independent recognition. Axis Bank is named for a know your customer deployment that streamlined document verification and reduced manual errors, Tata AIG for insurance onboarding and Redmil for identity verification, all on the vendor's own case study page.

A cheque clearance deployment processing more than twenty cheques every second is attributed only to one of the largest banks in India, which is the anonymised reference shape with a number attached. Independent evaluation is the strongest part: the Euromoney world's best lending solution award for 2026, Chartis category leadership across five corporate lending quadrants, a Chartis RiskTech100 placement, and inclusion as a representative vendor in Gartner's market guide for commercial loan origination. Held at B rather than A on the same basis as Opensee: no figure is attached to any named institution, and awards are evaluators rather than parties with money at stake.

Nothing states whether customer data trains the vendor's models. The question is sharper here than for most suppliers because the AI subsidiary runs API services processing identity documents, cheques and credit files at very high volume across many competing banks and insurers, so a pooled corpus would be both valuable and invisible to any individual buyer.

No exclusion of client data from any training corpus, no separation commitment between institutions and no statement of controls around the generative components are published. Against the reference set of Mortgage Capital Trading, Needl and AlphaSense the silence is a choice, and it sits oddly beside an unusually strong published position on model interpretability.

No privacy posture is published. Nothing addresses retention of the identity documents and financial records the document processing and verification APIs necessarily ingest, purge on termination, data subject rights under India's Digital Personal Data Protection Act, or the handling of personal data across the several jurisdictions the group operates in.

The identity verification and know your customer product lines make this a live question rather than a formality, because they process documents belonging to individuals who are the bank's customers rather than the vendor's.

No security certification is enumerated in the material reviewed. No SOC report, no ISO 27001, no penetration testing statement and no trust portal was located for either the banking software or the AI platform. The absence is notable for a supplier processing identity documents, cheques and credit files for major banks and insurers across several jurisdictions.

Queued check: the group is publicly listed and files detailed annual disclosures, and enterprise attestations for a vendor of this size are commonly held behind a request process rather than published on product pages.

A software supplier with no financial services licence, authorisation or supervisory relationship of its own. The group is publicly listed and therefore subject to securities regulation and exchange disclosure obligations, which is corporate regulation rather than financial services standing and does not read across under the standing bar.

The regulatory language in the product material describes what the platform helps a bank achieve, including regulatory auditability and compliance with lending rules, which is a product claim. No registration, enrolment or supervised programme participation was found.

No fairness position, disparate impact testing, protected characteristic handling or named governance framework is published, and there is no reference to the EU AI Act despite a European treasury acquisition and a product line supporting creditworthiness assessment.

The gap is conspicuous on this particular vendor rather than routine, because it publishes benchmarked interpretability metrics and sells model governance tooling, so it plainly has both the technical capability and the vocabulary to address fairness and has not. Explainability and fairness are different properties, and demonstrating the first says nothing about the second: a model whose reasoning is fully traceable can still distribute credit unequally.

No recourse position is published. Nothing states who carries the consequence when an identity verification wrongly rejects an applicant, a cheque is misread at volume, or a credit assessment misprices a borrower, and nothing addresses whether an affected individual is told that automated processing contributed or has any route to contest it.

The exposure is concrete: a cheque clearance system processing more than twenty items a second and identity verification APIs deployed across multiple institutions both make decisions about individuals at a scale where a small error rate reaches a large number of people, and the individual affected has no relationship with the vendor at all.

No model, family, version or provider is named for any capability. Independent coverage frames the AI acquisition as giving the group a specialised in house capability rather than leaving the lending platform dependent on third party models, which is a statement of posture rather than a disclosure, and it does not extend to naming what the generative components in the AI platform actually run on.

Under the standing rule that building permits silence, an in house model stack removes the commercial pressure to disclose, and this vendor has taken that option. The result is an unusual asymmetry worth noting: it publishes more than almost anyone about how its models can be interrogated and nothing at all about what they are.

Integration is named and specific rather than asserted. The trade limits module integrates with a major independent core banking vendor's trade innovation product, which is a documented connection into a competitor's platform rather than a generic API claim. The architecture is described as API first across the credit lifecycle, and the AI platform ships explicit data connectors with a graph component built to link fragmented proprietary bank data across systems.

The group also owns the layers below and beside the software, including payments gateways and, in its transit business, certified hardware, so the integration surface reaches from core banking through to device level. Successive acquisitions in loan management, treasury and customer engagement have been folded into the same suite, which is itself evidence of integration capability rather than a portfolio of separate products.

No deployment or residency position was found. Nothing states whether the lending suite and the AI platform are delivered on premise, in a managed cloud or as software as a service, which regions are available, or where customer data is processed and stored.

The omission matters more than usual for this vendor because its buyer base spans India, South East Asia, the Middle East and Europe, and several of those jurisdictions impose data localisation requirements on banks that would be a first order procurement question.

No pricing information of any kind was found: no licence basis, no per user or per loan metering, no module pricing, no implementation estimate and no indicative contract size, across either the lending suite or the AI platform. Every route into the products is an enquiry form. The group is publicly listed and therefore discloses revenue and segment performance to its shareholders, which tells an investor a great deal and a prospective buyer nothing about what the software costs.

Broad across institution type, product and region, with independent corroboration in depth. The lending platform is used by leading banks across Asia and the Middle East and the subsidiary is a Chartis category leader in five separate corporate lending quadrants covering limits management, collateral management, loan management, loan origination and alternative finance, which is unusually wide analyst coverage for one vendor.

Buyer types span banks, insurers, fintechs and enterprises, and the credit suite addresses corporate, small business and retail lending. Named institutions include Axis Bank and Tata AIG in vendor material, with State Bank of India, UOB, OCBC and Bank of Ayudhya identified in analyst coverage. Geographic reach runs across India, South East Asia, the Middle East and, through the treasury acquisition, Europe.