When a customer reported a loss, opening the claim meant an agent copying the same details into as many as seven separate systems by hand. The old document scanner read barely 40% of a claim correctly, so people re-keyed the rest, the customer waited on the line, and the work that should have been reassurance became data entry.
A new claim arrived as documents and a conversation, and the agent's job was to get it into the systems: the claims platform, the policy system, the customer record, the payments system and more. The same facts were entered by hand, five to seven times over.
The tool meant to help barely did. The existing document scanner read only about 40% of a claim accurately, so an agent still checked and corrected most of it, and any mistake in the re-keying followed the claim downstream.
And it landed at the worst moment. Someone opening a claim has usually just had an accident, a break-in or a loss, and instead of being helped they waited while an agent worked through screens, so the experience at the hardest point was slow and impersonal.
Put an AI assistant on the agent's screen that reads the claim accurately, enters it everywhere at once, and confirms cover in real time, with the agent in charge.
A document-understanding model reads the claim, its forms, photos and supporting documents, and extracts the details accurately, lifting reading accuracy from about 40% under the old scanner to around 99%, so people stop correcting what the tool got wrong.
The AI works as an assistant alongside the agent, aware of the systems already open on the screen, so it fits into how the team already works instead of forcing a new tool or a rebuild of the core platforms.
Instead of an agent copying the same details into five to seven systems, the assistant enters the claim into every relevant system at once, from the claims platform to the customer record, so the data is captured a single time and stays consistent everywhere.
As the claim is opened, the assistant checks it against the policy in real time, confirming the cover is valid and flagging anything that needs a closer look, so the agent has the answer in the moment rather than after the customer has hung up.
The assistant prepares and fills; the agent confirms. Nothing is committed without the agent, and anything read with low confidence or falling outside policy is surfaced for a person to decide, so speed never comes at the cost of a wrong call on someone's claim.
With the typing done in seconds, the agent's attention goes back to the person on the line. The intake that used to be minutes of screen work becomes a short, human conversation at exactly the moment it matters most.
The systems fill themselves, and the person on the line gets a person back.
Opening a claim went from minutes of re-keying across systems to seconds. The assistant reads the claim at around 99% accuracy, against roughly 40% under the old scanner, and enters it everywhere at once, so claims start clean and move faster from the first step. The team's solutions architect described the jump from 40% to near-perfect reading as the change that made the whole thing trustworthy.
Because the data entry no longer eats the call, agents spend the opening of a claim reassuring the customer rather than working through screens, turning the hardest moment in the relationship into the one where the service feels most personal.
Claims intake is slow and impersonal when a person has to re-key the same facts into every system by hand. The value here is an AI that reads the claim accurately, fills every system at once and checks the policy in real time, with the agent confirming each step, so opening a claim takes seconds and the agent is free to be present for the customer. It is the same document-understanding and decision-support discipline behind our prior authorization work, aimed at first notice of loss rather than payer approvals.
A provider's prior-auth requests took days and many bounced back denied on a technicality, writing off revenue. AI builds each request to the payer's own rules, flags what's missing before sending, and drafts appeals for a manager to approve, cutting approvals to hours and recovering lost revenue.
Read the case → Client intake · AccountingA firm's real bottleneck wasn't the accounting, it was chasing clients for documents. AI sends tailored requests, auto-chases, sorts every incoming file and assembles a ready-to-work pack, with a dashboard flagging the stuck ones, so work starts far sooner and deadlines stop slipping.
Read the case →
Reef TRH