Anand Mahalingam, Vice President – Data Science, Digit Insurance, on how bounded autonomy, real-time decisioning and agentic AI are reshaping claims, underwriting and customer experience.
Insurance is entering a new phase of AI adoption. The focus is shifting from using AI to accelerate individual processes to building intelligent systems that can interpret data, make decisions and act in real time. From claims and underwriting to fraud detection and customer engagement, agentic AI is beginning to reshape how insurers operate and how customers experience insurance.
For insurers, however, greater autonomy comes with an equally important responsibility: ensuring that AI-driven decisions remain transparent, secure and accountable. The challenge is no longer simply how much can be automated, but where autonomy creates value, and where human judgment must remain in the loop.
In this conversation with ET Edge CIO&Leader, Anand Mahalingam, Vice President – Data Science, Digit Insurance, discusses the organisation’s approach to bounded autonomy, real-time claims and underwriting, AI governance, and the emerging use cases that could redefine the insurance experience.
ET Edge: How is Digit approaching the shift from AI-assisted workflows to autonomous decision-making systems?
Anand Mahalingam: At Digit, our approach to autonomy is evolutionary rather than abrupt. We are moving from AI-assisted workflows towards bounded autonomous systems, where AI operates within clearly defined decision frameworks, while humans retain oversight over high-impact decisions.
AI systems are already making autonomous decisions in areas such as document routing, processing pathways and risk identification. This enables standard cases to be handled instantly, while complex or ambiguous scenarios are escalated for human review.
In claims processing, for instance, multiple steps—including document validation, risk screening and duplication checks—can be executed automatically and in parallel. This significantly reduces manual intervention while improving consistency and turnaround times.
We are also extending agentic principles beyond core operations. AI-powered conversational and calling agents are increasingly managing customer interactions across lifecycle touchpoints, enabling personalised engagement at scale.
Looking ahead, we see AI playing a larger role in straight-through claim approvals, autonomous fraud triage and intelligent escalation. However, in insurance, autonomy must always be paired with accountability, transparency and human oversight. The objective should be to automate the service standard while preserving human judgment where it matters most.
ET Edge: What role is AI playing in enabling scalable, real-time claims and underwriting processes?
Anand Mahalingam: AI is playing a pivotal role in how we scale operations at Digit, particularly across claims and underwriting. We have been moving from sequential processing towards real-time, event-driven decisioning.
In claims, AI enables document classification, information extraction, validation and fraud checks to run simultaneously rather than sequentially. This allows us to handle large volumes more efficiently while significantly reducing turnaround times.
High-confidence cases can be processed automatically, while complex or ambiguous cases are routed for human intervention. This creates a balance between speed and accuracy.
On the underwriting side, AI-driven models enable us to assess risk at a much more granular level and in near real time. Processes that traditionally took hours can increasingly be completed in minutes as structured and unstructured signals are evaluated simultaneously.
This capability becomes particularly important as insurance is increasingly embedded into real-time customer journeys through APIs and partner platforms. The ability to make fast, contextual decisions is becoming a key differentiator.
ET Edge: As AI systems become more agentic, what are the key governance, risk and security considerations?
Anand Mahalingam: As AI becomes more agentic, governance cannot be an afterthought; it has to be embedded into the architecture from the outset. At Digit, our focus is on ensuring that every AI-driven decision is transparent, traceable and auditable.
Each decision should have a clear trail of inputs, outputs and confidence levels. This allows us to review outcomes internally while also supporting regulatory and compliance requirements.
Equally important is maintaining a strong human-in-the-loop framework. Automation can handle standard and high-confidence scenarios, but critical decisions—particularly around complex claim approvals—require human judgment where necessary.
From a risk and security perspective, we operate within defined data-governance boundaries. Sensitive customer information needs to be anonymised, securely processed and used responsibly. As we move towards more agent-driven architectures, we are also placing greater emphasis on system-level controls that define roles and permissions across AI agents, monitor feedback loops and ensure that autonomy remains bounded.
Ultimately, increased autonomy should lead to greater transparency and trust—not opacity. The more agentic these systems become, the more important it is to make their behaviour explainable and accountable.
CIO&L: What are the emerging use cases where AI could drive the next wave of innovation in insurance?
Anand Mahalingam: The next wave of insurance innovation will be driven by AI’s ability to enable real-time, embedded and event-driven insurance experiences.
Take travel insurance, for example. AI can enable faster and more frictionless payouts when customers face disruptions such as flight delays or cancellations. As data ecosystems mature, we could see more use cases where payouts happen almost instantly, without conventional claims-processing journeys.
Embedded insurance represents another significant shift. Insurance can become part of the customer journey rather than a standalone purchase. Across travel, mobility and financial services platforms, AI can enable real-time underwriting, contextual pricing and instant policy issuance, allowing insurance to become seamlessly integrated into everyday transactions.
Conversational AI is also evolving from a support tool into a decision interface. Instead of simply answering customer queries, AI agents could guide customers, explain policies and assist with decision-making.
Autonomous claims settlement is another area with significant potential. AI systems could eventually validate, assess and approve standard claims with minimal human intervention. Combined with advances in real-time underwriting and continuous risk monitoring, this could move insurance towards a model that is more proactive, personalised and responsive.
The larger opportunity is to move from insurance that responds to events to insurance that can anticipate and act on them.