InsurTech NY: Adlib on Fixing Claims Data

At InsurTech NY, Kunal Bargotra from Adlib explains why one of the biggest challenges in insurance isn’t necessarily the systems but the data that feeds them.

Bargotra points to the claims intake process as a clear example. He says claims often arrive in highly unstructured formats, such as emails with multiple attachments and little context. From there, claims agents are required to manually open files, read through documentation, identify the type of claim, and input the relevant information into internal systems.

According to Bargotra, this creates a workflow that is heavily manual and time-intensive, particularly at the earliest stage of the process.

Adlib’s focus is on automating that initial intake layer.

Bargotra explains that the platform ingests incoming documents and emails, reads and classifies them, and then extracts and structures the relevant data. For example, it can identify whether a claim relates to fire, health, or another category, while also pulling out the key information needed to move the process forward.

The result, he says, is a much cleaner starting point for claims teams.

Instead of beginning with fragmented, unstructured inputs, analysts receive a summarised and organised package of information, allowing them to focus on decision-making rather than data preparation.

Bargotra also highlights that this challenge is not unique to one workflow or organisation.

Across insurance and particularly in regulated environments data is consistently messy, fragmented, and difficult to work with. This creates a recurring need to clean, structure, and enrich data before it can be used effectively in downstream processes.

This is where Bargotra sees a gap in how the industry is approaching AI.

While many organisations are focusing on implementing AI tools, he suggests that the underlying issue often lies earlier in the workflow. If the data feeding those systems is incomplete or unstructured, the outputs will be unreliable regardless of how advanced the AI model is.

He adds that large language models, while powerful, can produce inaccurate results or “hallucinations,” particularly when working with poor-quality data. In response, companies often try to refine prompts or add more layers to their AI workflows, rather than addressing the root cause.

For Bargotra, the solution is clear.

Focus on improving the quality of data at the source before it reaches any AI system.

By enriching and structuring data at the intake stage, insurers can ensure that downstream processes, including AI-driven ones, are more accurate, efficient, and reliable.

In that sense, Bargotra’s message is less about adding more technology and more about getting the foundation right first.