Editorial illustration for NTT DATA AIVista Tackles Complexity in Insurance Forms
NTT DATA AIVista Simplifies Complex Insurance Forms
NTT DATA AIVista Tackles Complexity in Insurance Forms
Bratin Saha has a specific example in mind when he talks about where AI breaks down inside big companies: multinational insurance claims forms. Handwritten fields, dense checkboxes, regulatory quirks that shift by jurisdiction. NTT DATA AIVista's CEO raised the problem at VB Transform 2026 in a conversation with VentureBeat CEO and editor-in-chief Matt Marshall, framing it as the central obstacle standing between enterprises and the AI spending they've already committed to.
The issue isn't model quality in the abstract. It's what happens when a frontier model meets a claims form that doesn't look like anything in its training data. Saha pointed to models including Fable 5, Opus 4.8, and GPT-5.5, capable systems by most measures, that still stumble on the kind of paperwork insurers process every day.
Most enterprise AI projects don't fail at the model layer, he argued. They fail during implementation, when integration gaps, missing domain knowledge, thin governance, and fuzzy ownership pile up. Closing that gap, Saha said, requires more than a better model.
It requires building an entire system around one.
"When you're deploying AI in the enterprise, you're not deploying a technology," he said. "You are taking a workflow that exists and taking it from point A to point B." The value is created by the workflow that gets moved, not the model that helps move it.
Why this matters
Saha's point about handwriting and checkboxes sounds small until you remember that's where most enterprise AI budgets actually go to die. Frontier models keep getting bigger benchmarks and bigger price tags, but insurance claims don't care about MMLU scores, they care about whether a scanned form with a coffee stain gets parsed correctly. NTT DATA is betting that the money is in the boring layer: context handling, guardrails, security, the stuff that turns a demo into something a compliance officer will sign off on.
For founders building on top of GPT-5.5 or Opus 4.8, that's a useful reality check. The model isn't the product anymore, the last mile is. For enterprise buyers, it's a reminder to ask vendors exactly how they handle the messy 20% of documents that don't fit the clean training distribution, because that's usually where the ROI conversation actually gets decided.
Watch whether NTT DATA publishes any real accuracy numbers on those regulated workflows, rather than just talking about complexity in the abstract.
Related Reading
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
From Archives Karpathy Spends USD 10 on Claude Opus to Render Tolkien in 3D
From Archives Fender CEO Says Your Bandmates Are "Analog AI
Common Questions Answered
What specific enterprise challenge does NTT DATA AIVista address with its AI solution?
NTT DATA AIVista tackles the complexity of processing multinational insurance claims forms, which contain handwritten fields, dense checkboxes, and jurisdiction-specific regulatory requirements. According to CEO Bratin Saha, these forms represent a central obstacle preventing enterprises from realizing the value of their AI investments, as they require sophisticated handling of unstructured and variable data formats.
How does Bratin Saha differentiate between deploying AI technology versus creating enterprise value?
Saha emphasizes that enterprise AI deployment is not about the technology itself, but rather about moving existing workflows from point A to point B. The actual value is created by the workflow transformation, not by the underlying AI model that facilitates the movement, shifting focus from model capabilities to practical business outcomes.
Why do insurance claims processing and benchmark scores like MMLU represent different priorities in enterprise AI?
Insurance claims don't care about frontier model benchmarks or MMLU scores; instead, they require accurate parsing of real-world documents like scanned forms with coffee stains and handwritten annotations. This highlights that enterprise AI budgets often fail because they prioritize impressive model metrics over the boring but critical infrastructure layer needed for reliable production deployment.
What are the key components NTT DATA focuses on to turn AI demos into production-ready solutions?
NTT DATA bets on the importance of context handling, guardrails, and security as the critical layers that transform AI demonstrations into functional enterprise systems. These foundational elements are essential for handling the complexity and variability of real-world business processes like insurance form processing, rather than relying solely on frontier model capabilities.
Further Reading
- NTT DATA AI for Insurance - NTT DATA
- NTT DATA Named a Leader in Insurance GenAI and Agentic AI Services - NTT DATA
- NTT DATA AIVista: Operationalize Enterprise AI at Scale - NTT DATA AIVista
- NTT DATA AIVista launches in Silicon Valley - Consulting.us
- NTT DATA AIVista and Snowflake: Identity alone won't secure enterprise AI agents - VentureBeat
