NTT DATA has launched AI for Insurance, an agentic artificial intelligence solution designed to transform complex insurance sector processes such as underwriting, claims management, and customer service into repeatable and governed services. The offering leverages AIVista, the company’s AI platform, aiming to address one of the sector’s main challenges: moving from isolated pilots to integrated, auditable AI operations with human oversight.
The key points of NTT DATA AI for Insurance in 20 seconds
- The solution combines AI agents, industry-specific models, orchestration, and governance controls.
- It’s designed for underwriting, claims, customer service, and other core insurance processes.
- Can integrate with existing systems without requiring core replacement.
- NTT DATA also wants to avoid reliance on a single foundational model.
- The 2026 report shows strong AI adoption in back and mid-office, but still limited actual industrialization.
The launch comes at a time when insurers are increasingly using AI but continue to face difficulties in deploying it consistently in production.
The NTT DATA 2026 Global AI Report for Insurance indicates that 86.7% of leading organizations apply AI in back and mid-office processes, while 66.7% also extend it to growth-oriented use cases in the front office. However, the gap between experimenting with models and establishing them as stable operational components remains substantial.
This is precisely the space AI for Insurance aims to occupy.
From chatbots to agents involved in processes
NTT DATA’s proposal goes far beyond adding a conversational assistant to an insurer’s website.
The architecture is designed to introduce specialized agents within complete processes. These agents can gather information, analyze documents, query internal systems, apply rules, suggest decisions, and coordinate with humans and enterprise applications.
For example, in underwriting, an agent could compile information from multiple systems, complete risk data, and present a recommendation to the underwriter.
In claims, it could help classify files, gather documentation, identify cases needing further review, and provide context to the handler.
And in customer service, it can combine policy, claim, and CRM system information for more context-aware responses.
NTT DATA describes this approach as a transition from simply assisted workflows to operations partially delivered through AI, maintaining human intervention where necessary.
This emphasis on human involvement is crucial because insurance decisions are especially sensitive to quality. Improper automation can impact premiums, payouts, risk acceptance, or policyholder rights.
Therefore, the solution includes human supervision and controls within the operational layer itself.
Governing the agent, not just the model
Governance features are central to the platform.
NTT DATA incorporates specific guardrails to control agent decisions, ensure traceability, and generate audit evidence for later review.
This matters because in an agent-based architecture, the question isn’t only what response a model produces.
You also need to know:
- which data sources were consulted;
- which rules were applied;
- what tools were used;
- what action was proposed;
- which system was modified;
- who approved the decision;
- and what evidence is recorded.
This complexity increases in regulated industries, where organizations must explain not only outcomes but also the processes used to achieve them.
According to NTT DATA’s study, 67.4% of insurers it considers AI leaders use centralized governance models, compared to just 23.5% among less advanced organizations.
This difference suggests that the ability to deploy AI at scale largely depends on having common controls over models, data, and decisions.
Without replacing the insurer’s core system
Another key aspect is integration.
Large insurers often operate on systems developed and expanded over decades. Replacing the entire core to implement a new AI layer would be costly, slow, and risky.
NTT DATA presents AI for Insurance as a layer that can integrate with these existing systems.
The company describes its approach as cloud-agnostic and asserts that it can connect with common industry platforms and with the existing systems, data, and processes already in use.
Practically, this architecture could look like:
Core insurance system
│
CRM ──┼──────── External Data
│
▼
NTT DATA AI for Insurance
│
┌────┼───────────┐
│ │ │
▼ ▼ ▼
Agent Agent Agent
Risk Claim Customer
│ │ │
└────┼───────────┘
▼
Persona / System / WorkflowThe goal is to introduce a layer of intelligence and orchestration without necessarily transforming the entire transactional system.
Also aims to avoid AI model lock-in
Technological dependency does not end at the core system.
A company building all its processes around a single large model might face similar issues as any proprietary platform down the line: costs, evolving capabilities, or commercial conditions could change.
NTT DATA states that its solution incorporates model routing to select different models based on the task.
This is significant because not all tasks require the same LLM.
A simple classification task could use a small, inexpensive model, while more complex cases might need a larger, more reasoning-capable model.
Conceptually, such architecture might look like:
Request
│
▼
Model Router
│
├── Small model → classification
├── Specialized model → documents
├── Advanced model → reasoning
└── Local model → sensitive dataBeyond reducing dependency, this approach can also impact inference costs directly.
Sector-specific data and knowledge
A particularly important part of the solution is what NTT DATA calls the insurance data genome and its tailored knowledge base focused on the sector.
This concept addresses another common challenge in enterprise AI: a general model can understand language but may not know the internal structure of an insurance company.
It needs context about elements such as:
policy
├── coverages
├── exclusions
├── insured
├── beneficiaries
├── premiums
└── conditions
claim
├── associated policy
├── damages
├── documentation
├── fraud
└── indemnityThe usefulness of an enterprise AI depends on these concepts being correctly linked to real systems and data.
NTT DATA attempts to address this through ontologies and insurance-specific models that enable agents to better interpret policies, claims, documentation, and regulations.
Deployment can be three times faster, according to NTT DATA
The company claims its preconfigured agents can enable deployment up to three times faster.
This is an estimate from the provider itself and should be taken as such, since actual timing depends on each insurer’s systems, data quality, integration needs, and regulatory requirements.
But the strategy is clear.
Instead of building from scratch:
Model
+ prompts
+ RAG
+ agents
+ security
+ integration
+ auditing
+ workflowNTT DATA provides components already oriented to specific industry use cases.
Each insurer can then customize agents to their products and processes.
AI aims to directly intervene in underwriting and claims
The main shift compared to previous generations of enterprise AI is the proximity of these tools to operational decision-making.
For years, the most visible use cases involved chatbots, document analysis, or employee assistants.
Agentic AI begins to approach processes such as:
Underwriting. Gathering information and assisting in risk assessment before accepting and pricing a policy.
Claims. Classifying claims, identifying complex cases, gathering data, and assisting the handler.
Fraud detection. Connecting dispersed signals and prioritizing files for investigation.
Customer service. Resolving requests using context from multiple systems.
Operations. Automating tasks that previously involved manual data movement between applications.
NTT DATA notes that leading sector players are already focusing AI precisely on these areas. Its report identifies underwriting, claims processing, fraud detection, pricing, and distribution as the main values being generated.
Regulation ensures human oversight remains essential
Total automation remains particularly challenging in insurance.
Decisions can directly impact individuals and companies, involving regulatory obligations, data protection, explainability, and internal risk policies.
NTT DATA proposes an architecture of human oversight, where certain decisions may ultimately require review or approval by humans.
This approach can be depicted as:
AI Agent
│
▼
Evaluation
│
├── Low risk → automation
│
├── Exception → human review
│
└── High risk → blocking/escalationThe goal is not necessarily to eliminate underwriters, appraisers, or claims handlers, but to reduce manual work surrounding their decisions.
A shift towards Service-as-Software model
NTT DATA also uses an interesting term to describe this offering: Service-as-Software.
The semantic shift is significant.
Traditional enterprise software provides tools that a person uses to perform a task.
In the Service-as-Software approach, part of the work itself becomes a function delivered by the software.
For example, instead of just providing an application to review a claim:
employee → software → decisionthe architecture evolves to:
data
↓
agents + rules + systems
↓
proposed action / decision
↓
human supervisionThis is one of the most significant trends behind enterprise agentic AI: moving from selling software that helps execute processes to offering platforms that directly execute parts of those processes.
NTT DATA’s initial focus on a highly specific sector
The company enters this market with a significant presence in insurance.
NTT DATA states it has more than 12,000 insurance specialists, 16 delivery centers, and works with 10 of the top 25 global insurers by non-bank assets.
This sector expertise can be crucial, as one of the challenges of enterprise AI is no longer just accessing a good model.
Foundational models are increasingly becoming an interchangeable layer.
The competitive advantage begins shifting towards:
own data
+ processes
+ integration
+ governance
+ sector knowledgeAnd insurance is one of the sectors where integrating these elements is particularly complex.
NTT DATA’s strategy reflects a broader industry shift. The next phase of enterprise AI adoption will likely be measured not by how many employees have access to a chatbot, but by how many complete processes can incorporate agents without losing traceability, regulatory compliance, and human control.
Frequently Asked Questions
What is NTT DATA AI for Insurance?
It’s an agentic AI platform tailored for insurers that combines specialized agents, industry-specific models and data, process orchestration, and governance controls.
What processes can it automate?
NTT DATA mainly targets underwriting, claims management, customer service, and other operational insurance processes.
Does it require replacing the insurer’s core system?
No. The solution is designed to integrate with existing systems. NTT DATA affirms it can be used without depending on a single core platform or foundational model.
Does the AI make decisions without human oversight?
The architecture includes guardrails, auditability, and human supervision. The level of automation will depend on the process, insurer policies, and regulatory obligations.
