Perfios

Their 'Perfios Analyse' suite extracts structured data from ITR forms, bank statements, and payslips with domain-trained precision — which was magnificent until GPT-4V started reading financial documents natively and Anthropic's computer use started filling out loan applications end-to-end. The API calls still work; the reason to pay for specialized parsing is evaporating.

Risk analysts, document processors, and fraud investigators are the core workforce here — roles that multimodal AI handles with terrifying competence. The engineers who built the OCR pipelines will pivot to prompt engineering; the domain experts who trained the models just became expensive consultants on datasets Claude reads for free.

Here's the real moat: enterprise integrations so deep that changing vendors means re-engineering every connected system, plus 15 years of Indian financial document edge cases baked into the models. When your fraud detection API is hardcoded into a bank's loan approval workflow, switching costs are genuinely structural — not the frictional kind that AI dissolves overnight.

They rebranded to 'perfios.ai' and launched 'Perfios Lens' as their 'AI-driven intelligence suite,' which sounds suspiciously like wrapping their existing document parsing in conversational interfaces. Points for self-awareness, minus points for thinking the solution to AI disruption is adding AI as a feature to the thing AI is disrupting.

Financial compliance, fraud detection, and risk assessment survive and compound — regulation gets more complex as AI adoption creates new attack vectors and audit requirements. The need grows; the question is whether specialized fintech infrastructure survives or gets absorbed into foundation model capabilities that every bank licenses directly from Anthropic.