Most companies announce AI as a breakthrough. Go Digit General Insurance did something less common in its 15th Transparency Report (H1 FY27, data as at 30 September 2026): it first published where service got slower, then said which AI tools it is deploying to fix each problem.
The problem AI is meant to solve
Digit has always aimed to give same-day vehicle repair approvals, while the report notes that approvals typically take 1-2 days across the industry. In H1 FY27, however, its average vehicle repair approval time rose to 17 hours 39 minutes from 14 hours 44 minutes in H1 FY26.
The Gujarat floods were part of the reason, but not all of it. Without the flood claims, the average was 16 hours 27 minutes. The report states plainly that “the remaining increase is ours to address.”
Digit also notes that the metric covers the full stretch from the First Notice of Loss to the issuance of the repair order. That includes claim intimation, assessment of admissibility and communication of work approval to the garage.
AI tool #1: Speech-to-text during vehicle inspection
For vehicle repair approvals, the report says Digit has “introduced AI-enabled speech-to-text during vehicle inspection.”
The idea is to capture the assessor’s findings by voice instead of manual entry. For straightforward claims, Digit says this “would help us complete assessments faster and issue work orders at the garage, helping reduce the overall wait time.”
This matters because the report links the slowdown to physical surveys. A higher share of claims required them, and during the floods, inaccessible vehicles and garage congestion added survey and travel time. Speeding up what happens during the inspection attacks the delay at its source. The report also sets the goal of ensuring “similar turnarounds even when claim volumes rise cyclically due to nat-cat events.”
AI tool #2: Human-assisted OCR for policy changes
The second tool targets retail health policy changes. The average turnaround for non-claim changes rose from 29 minutes to 41 minutes. Digit’s explanation is specific: “a small number of requests took significantly longer, pushing up the average.”
Its response is “a human-assisted OCR flow combining AI-led document reading with human validation.” Rather than aiming for full automation, the design lets AI read the documents and sends them to people for validation. For a regulated insurer handling personal data, that is a sensible balance, and the stated goal is to “bring down outlier processing times and the overall average.”
The technology around the AI
Not every efficiency gain in the report is AI. During the Gujarat floods, Digit settled 2,109 motor claims, or 72.5% of those received, and paid ₹28.5 crore by the end of H1 FY27. The report credits a mix of tools and approaches:
Automated RC verification through Vahan/Parivahan
Home surveys for 45% of customers
Towing assistance
No mandatory licence checks for parked vehicles
65+ claims team members across field and back-office processing
These are better described as digital automation and process design than as AI, and they show the larger system the new tools will plug into.
The case for AI at Digit
The business logic is visible in the financials. Policies sold rose 22% to 8.8 million, and customers served rose 24% to 94 million. Insurance revenue grew 5% to ₹5,386 crore. When volumes grow faster than revenue, the cost and time to serve each customer has to fall.
The foundation is already strong. Cashless health approvals averaged 25.1 minutes against a 60-minute regulatory limit, and 95% of grievances were resolved on first contact. On peak days, Digit has issued 3.5 lakh policies and settled 11,750 claims in a single day. AI is being applied where this happy path breaks down.
The test ahead
Digit says its next Transparency Report will follow in May 2027. Because it publishes comparable year-on-year figures, the repair-approval and policy-change numbers will show whether the AI tools worked. If repair approvals move back toward same-day and health policy-change times fall, the claim that AI helps where customers wait will rest on published data.
Source: Digit General Insurance Transparency Report (15th edition), Form NL-37 Claims Data as at 30 September 2026.