BBevaya

Roots Automation says the models behind Bevaya were trained on more than 300 million insurance documents, that its agents run in more than 115 production deployments, and that customers realized more than $100 million in value in 2025. In May 2026 the company folded all of that into a new brand, Bevaya, an AI agent platform for carriers, brokers and TPAs. It is best for insurance operations teams buried in documents: submissions, loss runs, claims files, certificates and premium audits.

Bevaya is an enterprise purchase with pricing shared by quote, its results come from the vendor’s own customer examples, and the new name means much of the public record still says Roots. Published pricing and named references would help buyers. We rate it 7.6/10 on our scale, a score shaped by pricing transparency and the absence of public ratings so far rather than by the product’s scope. We have not used Bevaya ourselves; this review is based on the vendor’s product descriptions, funding announcements and trade press.

Roots Automation’s Published Track Record

Bevaya is new as a brand, but Roots Automation has published scale and customer figures over several years.

The growth from 35 customers in 2024 to more than 115 production deployments in 2026 is the strongest signal here, although deployments and customers are not the same unit, since one carrier can run several. A strategic investment from Liberty Mutual’s venture arm also tells buyers that at least one large carrier looked closely at the technology before backing it.

The case results are specific and unnamed. A regional insurer, a commercial carrier, a commercial auto insurer and a TPA are described without names. Treat the percentages as a guide to where the agents have worked well and ask for references in the matching use case.

The rebrand itself is a data point. Moving from Roots to Bevaya came with what the company calls new architecture and a redesigned interface, and with a shift in positioning from digital coworkers that automate single tasks to a platform for designing and governing many agents. CEO and co-founder Chaz Perera framed it simply: insurance people deserve better tools, and Bevaya is the company’s answer. For buyers, that means a product in the middle of a transition, which can bring new capabilities quickly and also some rough edges.

Designing and Governing Agents Across Insurance Ops

Bevaya is one environment where insurance teams design, deploy and govern AI agents across underwriting, claims and policy servicing. It sits beside the company’s existing systems of record rather than replacing them.

Prebuilt Agents

Bevaya’s vendor listing names five P&C areas the platform supports: claims management, document management, policy management, quote management and underwriting management. Inside those areas, the prebuilt agents cover these jobs:

• Submission intake: reads incoming submissions and pulls out the data underwriters need.
• Loss run processing: extracts and organizes loss history from carrier loss runs.
• Policy comparisons: compares policy documents to show differences.
• Claim indexing: classifies and indexes claim documents.
• FNOL and FROI setup: sets up first notice of loss and first report of injury records.
• Certificate of insurance creation: produces certificates from policy data.
• Premium audit processing: handles the documents and data behind premium audits.

What the Agents Do With a Document

1. Read and classify: identify what the document is and route it.
2. Extract: pull the fields that matter for the workflow.
3. Analyze and summarize: check the data and condense it for a human or the next system.
4. Recommend: suggest the next action, which humans review or which executes under the rules the team set.

Custom Agents and InsurGPT

Beyond the prebuilt set, teams can create and manage agents by writing natural-language prompts. All agents run on InsurGPT, the insurance-tuned models Roots introduced in 2023. Roots also joined Guidewire’s Insurtech Vanguards program, which fits the Guidewire and ClaimCenter integrations Bevaya lists.

Human Review of Flagged Items and Agent Ownership

Human-in-the-loop controls and explainability features are part of the design, so people still review flagged items and decide where agents may act alone. Someone has to own agent configuration, exception queues and the rules that govern which recommendations are executed automatically. The vendor lists training through documentation, live online sessions, webinars, in-person sessions and videos, which suggests a guided rollout rather than a sign-up-and-go product.

Enterprise Contracts Shaped by Scope

Bevaya is sold through enterprise contracts. These are the parts that shape a quote, based on the vendor’s descriptions of the platform.

How Much Keying Work Must Disappear to Pay Off

With no public price list, the question is how much work the agents must remove to pay for themselves. These scenarios apply Bevaya’s own reported percentages to volumes we chose for illustration.

• A TPA processing 1,000 claims a month: the reported 85% capacity increase would let the same team handle 1,850 claims a month, or absorb growth without new hires.
• A commercial carrier with a 30-minute intake task: a 97% cut in handling time brings it to under a minute (0.9 minutes). On 2,000 such tasks a month, that is about 970 staff hours freed.
• A commercial auto insurer with 200 premium errors a year: a 90% reduction leaves 20. The value is the rework and leakage those 180 avoided errors would have caused.

Other costs to plan for: integration work with policy and claims systems, staff time to define rules and review exceptions, and a period of tuning before accuracy reaches production levels on your own documents.

Before a sales call, gather three numbers for each candidate workflow: monthly document volume, average handling time per document and the current error or rework rate. Those let you hold any quote against the vendor’s reported results and decide which agent to start with. Ask, too, whether custom agents built with prompts are priced the same way as prebuilt ones, since that affects how cheaply you can extend the platform later.

Bevaya Compared With Sixfold, Clive by Five Sigma and Agentech

These agents all sell to insurers, but each focuses on a different part of the work.

Bevaya is the broadest of the four, which suits operations leaders who want one platform for intake work across departments. Sixfold goes deeper on underwriting judgment, and Clive goes deeper on handling the claim itself. Agentech overlaps most on claims intake, but it sells task automation for adjusters rather than a platform for building agents. See more on our AI agents for insurance page and in AI agents for enterprise.

Carriers With Submission Backlogs and Busy TPAs

• Carriers with submission backlogs: submission intake and loss run agents target the slowest manual steps in commercial underwriting.
• TPAs and claims teams: claim indexing and FNOL and FROI setup agents speed up the start of every claim.
• Brokers issuing many certificates: the certificate agent turns a repetitive service task into an automated one.
• Guidewire shops: listed Guidewire and ClaimCenter integrations reduce integration risk.
• Teams that want to build their own agents: natural-language agent creation lets operations staff extend the platform.
• Workers’ compensation and commercial auto carriers: two of the published case results, a 97% handling-time cut and a 90% drop in premium calculation errors, came from these lines.

Agencies Wanting a Phone Receptionist Need a Different Tool

• Independent agencies wanting a receptionist: Bevaya works on documents and back-office workflows, not inbound calls. A voice agent such as Sonant AI fits that need.
• Buyers who need published prices: every deployment is quoted.
• Teams wanting an AI that decides claims: Bevaya recommends actions under human-in-the-loop controls; Clive by Five Sigma goes further toward straight-through claim handling.
• Very small operations: the volumes behind the reported results suggest the platform pays off at carrier, broker and TPA scale.

A Brand Change and Unnamed Case Results

• Name change: contracts, case studies and review profiles may still use Roots or Roots Automation, so check that you are comparing the current platform.
• Anonymous results: the 99%, 97%, 90% and 85% figures come from unnamed customers. Ask for references in your line of business.
• Deployments versus customers: more than 115 production deployments does not mean 115 customers; ask how many accounts are live.
• Governance setup: human-in-the-loop rules and exception queues need owners, or work piles up in review.
• Scope: the platform covers documents and back-office tasks; policyholder conversations need a separate tool.
• Platform transition: customers on the older Roots platform should ask how and when their agents move to the new Bevaya architecture, and whether pricing or terms change with it.

Roots Automation Relaunches as Bevaya in May 2026

• May 29, 2026: Roots Automation launched Bevaya as its new flagship AI agent platform for insurance, with new architecture and a redesigned interface.
• 2026: Coverager reported the rebrand, noting that Bevaya replaces the Roots platform and that the company counts more than 115 production AI deployments.
• September 23, 2024: Roots raised a $22.2 million Series B led by Harbert Growth Partners, with Liberty Mutual Strategic Ventures among the follow-on investors.

Real Production Scale Behind a 5.2 Score

For carriers, brokers and TPAs whose bottleneck is reading and keying insurance documents, Bevaya is a strong candidate with real production scale behind it. It scores 7.6/10 on our scale: credible on autonomy and insurance focus, with room to grow through published pricing, named results and public user ratings.

The sensible test is one agent on one workflow, such as loss runs or FNOL setup, measured against your current handling time and error rate. If you see results anywhere near the reported 97% cut in handling time, adding more agents is an easy decision.

Compared with narrower tools, Bevaya’s advantage is breadth: one vendor and one governance model for intake work across underwriting, claims and servicing. That breadth is worth most to operations leaders who would otherwise buy three point solutions. If your problem sits in one place, such as underwriting risk analysis or claim handling, a specialist like Sixfold or Clive by Five Sigma may get you further in that one area.

Is Bevaya the Same Company as Roots

Is Bevaya the Same as Roots Automation?

Yes. Roots Automation launched Bevaya in May 2026 as its flagship platform and primary brand, replacing the earlier Roots platform. The InsurGPT models carry over, and older case studies, funding news and review profiles still use the Roots name.

How Much Does Bevaya Cost?

Bevaya is sold through enterprise contracts with carriers, brokers and TPAs. Quotes depend on the agents you use, custom agents you build and document volumes. Measure your current handling times first so you can judge the quote against the work it removes.

What Is InsurGPT?

InsurGPT is the set of insurance-tuned AI models behind Bevaya’s agents. The company says it was trained on more than 300 million proprietary insurance documents, which is meant to help it read loss runs, submissions and claim files that general-purpose models handle less reliably.

Does Bevaya Integrate With Guidewire?

Bevaya lists Guidewire and Guidewire ClaimCenter integrations and offers API access for other systems of record.

Can Bevaya Agents Act Without Human Review?

The agents read, extract, summarize and recommend actions, and human-in-the-loop and explainability controls decide what is executed automatically. Teams set those rules for each workflow, so a cautious team can keep every recommendation under review at first and allow automatic execution only after accuracy is proven on its own documents.