The conversation around AI in insurance underwriting often focuses on speed. Faster quote turnaround times, greater automation and increased straight-through processing have become common benchmarks for success. But according to Peeyush Rai, Founder and CEO of Weav.ai, focusing solely on these velocity metrics risks missing the bigger picture.
“If you look at underwriting, especially in commercial lines and specialty lines, it’s very different from personal lines," Rai told IB. “While personal lines underwriting is often driven by a small set of variables, commercial risks require underwriters to evaluate a much broader and more nuanced set of factors. Each risk is very different, and you need a broader lens to assess the risk. It's not just a number thrown by an AI model.”
That distinction shapes Weav.ai's approach to underwriting automation. While the company has helped customers dramatically reduce quote turnaround times, Rai quickly pointed out that speed without underwriting discipline creates its own problems.
"You could be turning around lots of quotes, but if they are not profitable it's not a good thing. You're doing bad things faster."
For Weav.ai, the objective really is twofold - improving both speed and accuracy. And that first opportunity comes before an underwriter even begins assessing a risk - across commercial insurance, a significant amount of time is still spent gathering, validating and cleaning submission data. According to Rai, carriers often face a "huge upfront overhead" simply ensuring information arrives completely and accurately. The process typically involves repeated exchanges between brokers, agents and underwriting teams, while many insurers continue to outsource portions of this work to third-party providers.
This is where AI is already delivering measurable gains. By automating much of the pre-underwriting process, Weav.ai has been able to eliminate the majority of manual effort involved in reviewing incoming submissions. Rai told IB that the company has achieved "an 85% improvement in terms of automation", with AI handling most data extraction, validation and document processing tasks.
Importantly, the company has not removed human oversight. Instead, it applies a confidence-based model in which underwriters are only involved when the system identifies uncertainty.
“The 15% that goes to human-in-the-loop are the cases where AI says, 'I'm not really confident about this particular data point or information or document that has come in’,” added Rai.
And the impact on operational efficiency has been substantial. For some customers, underwriting preparation work that previously took three or four days can now be completed in as little as 15 to 20 minutes. However, Rai believes that one of the less-discussed benefits of AI-powered underwriting is its ability to reduce market noise. In commercial insurance, carriers frequently receive large volumes of submissions that ultimately prove uncompetitive or fall outside appetite. And, as a result, underwriters often spend valuable time reviewing business they were never likely to write.
"Some brokers and agents spray carriers with lots of submissions for the same business," Rai told IB. The result is that highly attractive opportunities can become buried beneath dozens of less relevant applications.
To address this, Weav.ai enables insurers to issue indicative pricing much earlier in the process. This allows brokers and agents to quickly determine whether a risk is likely to fit a carrier's appetite and pricing expectations. And, by filtering unsuitable opportunities earlier, carriers can focus their attention on higher-quality submissions while reducing wasted underwriting effort.
Once a submission reaches the underwriting stage, Weav.ai’s focus shifts from efficiency to decision support. Rather than presenting underwriters with a flat list of variables, the platform attempts to model risk in a way that reflects how individual insurers actually evaluate exposures.
"Two property carriers will not look at the risk the same way," Rai explained. Even within the same line of business, carriers may have different underwriting philosophies, product structures and risk appetites. Traditional underwriting systems often struggle to capture this complexity, relying instead on relatively rigid rule-based approaches.
Weav.ai takes a different route. The platform researches business operations, validates classifications, analyses operational characteristics and builds a more complete representation of each insured. And rather than relying on a generic model, the system actually attempts to mirror the carrier's own view of risk.
"We have AI scorecards that assess the risk even before the underwriter sees it," added Rai. This means that carriers can establish thresholds that determine whether submissions should be automatically approved, declined or referred for further review. Risks that meet predefined standards can move directly through the underwriting workflow without human intervention, while more complex or ambiguous cases continue to receive underwriter attention.
Building trust in these automated decisions remains critical. Rai compares the adoption process to self-driving vehicles.
"Initially, we are cautious of what the car is doing, but then you get used to its behaviour. From there, you start trusting it for certain things. And that's how AI works in general - that's the philosophy we also deploy to our customers. Metrics-wise, pre-underwriting we reduce 85% of the manual effort - we've also reduced the initial turnaround of quotes from days to 15 minutes. While we are still early in our journey, we are seeing significant traction - and customers are seeing significant traction too."
Underlying all of this is data. But according to Rai, the challenge is not simply accessing more information, it’s about making sense of it – something which is often easier said than done.
"We partner with several data providers," he told IB, “but external datasets alone rarely provide a complete picture.”
Instead, Weav.ai combines internal carrier data, third-party information and independently gathered intelligence into what it calls knowledge graphs. These knowledge graphs allow the platform to create a structured representation of businesses, locations, operations and exposures. The system can also supplement underwriting information by researching publicly available sources, including company websites, operational updates and other indicators that may affect risk quality.
"The key is to combine them into one view of the risk," added Rai. “Pricing and rating engines are basically input and output. If you give them high-quality input, you'll get a high-quality output."
Although underwriting automation continues to attract headlines, Weav.ai sees the longer-term opportunity elsewhere. The company is building systems that connect underwriting decisions with policy performance over time, incorporating audits, endorsements, renewals and claims activity into future decision-making.
"Risk is never static," warned Rai. “We track the dynamic nature of the risk over its lifecycle. And then, as renewals happen, we have a pretty good idea of that risk at a point level as well as a portfolio level. We [can see] how things are trending and we can, in real time, inform not just the underwriters, but the product managers and the executive teams on what influence this is having at a business level.”
This article was created in partnership with Weav.ai
