Integration · Guidewire
Claims automation that runs alongside Guidewire ClaimCenter
ClaimCenter stays the system of record. Hesper reads the claim file, does the legwork, and writes evidence-backed findings and drafts back onto the claim.
Last updated
In short
Hesper AI runs beside Guidewire ClaimCenter, not instead of it. ClaimCenter stays the system of record. Hesper reads the claim, its exposures and its documents, runs its agents across coverage, investigation, settlement and recovery, then writes cited findings and drafts back onto the file as notes, documents and activities. Adjusters keep every decision.
Does Hesper replace Guidewire ClaimCenter?
No. ClaimCenter stays the system of record. It keeps the claim, the exposures, the reserves, the payments and the audit trail, and it stays where your adjusters work. Hesper does not write policy, set reserves on its own, or move money. It is an automation and evidence layer that runs beside the core system.
That distinction matters commercially. Replacing a core claims system is a multi-year programme. Adding claims automation beside one is a scoped integration against interfaces ClaimCenter already exposes, and it can be narrowed or switched off without unwinding anything. Nothing in ClaimCenter changes shape because Hesper is reading it.
It matters technically too. ClaimCenter is built to record and govern a claim: who did what, when, and under what authority. Hesper is built to do the work that produces those records, which is reading the file, pulling and checking evidence, and drafting the output. The two are complementary rather than overlapping.
How does Hesper get the claim file?
Through the integration surfaces ClaimCenter already has. Guidewire exposes claim data over APIs and can publish events when something changes on a claim, and most carriers run batch or file-based feeds alongside those. Hesper consumes whichever of those your team prefers to open, the same way any other downstream system would. There is no agent installed on an adjuster's desktop and no scraping of the user interface.
- A trigger: a new claim, a new exposure, a document landing on the file, a status change, or a scheduled queue of claims selected by line and severity.
- The claim record itself: loss facts, exposures, parties, coverages, reserves, and the claim's own history.
- The documents attached to it: ACORD forms, police and fire reports, medical records, estimates, invoices, photos, recorded-statement transcripts and correspondence.
- Your rules for scope: which lines of business, which claim types, which severity bands, and which claims never leave your environment at all.
Pull, push and scheduled batch all work. Carriers who want a narrow first slice usually start with one line of business and one trigger, then widen once the output has been checked against files their own adjusters already closed.
What does Hesper write back to ClaimCenter?
Work product, on the claim, in the places a ClaimCenter user already looks. In Guidewire terms that means notes, documents attached to the claim, and activities assigned to a person with a due date. The claim keeps its identity and its history; Hesper adds to it.
- A findings note on the claim, written in plain language, with every statement tied to the document or source it came from.
- An evidence package attached as a document: the source material, a reconstructed timeline, and the inconsistencies the agents found, each one cited.
- Drafts your team edits and owns, such as a coverage position, a reservation-of-rights or denial rationale, a demand response, or a subrogation demand.
- An activity for the adjuster or supervisor naming the recommended next action, so the work lands in a queue rather than in somebody's inbox.
What Hesper does not write is just as important: no payments, no reserve changes it decided on its own, no denials, no closed claims. Nothing is auto-paid and nothing is auto-denied. Every output arrives as something a person approves, edits or rejects, and that approval is recorded in ClaimCenter exactly as it is for work done by hand.
What does this change for the adjuster?
The file is already assembled when they open it. Today an adjuster working in ClaimCenter spends much of a claim building context: reading the documents, chasing the missing ones, querying databases, reconciling a recorded statement against a repair estimate. The core system records all of that faithfully. It does not do any of it.
With Hesper in place that legwork arrives finished and cited. The adjuster reads an evidence-backed summary instead of assembling one, then makes the call. Clean claims move through with the checks documented; claims with something wrong in them arrive with the problem already identified and evidenced. Speed stops coming at the cost of accuracy, because the same engine does both.
Decision authority does not move. Coverage positions, reserves, settlement authority, referral and denial stay with your people, under your existing authority matrix, logged in ClaimCenter. See claims triage and claim file review for what the agents produce at each point.
How long does a ClaimCenter integration take?
That depends on your environment rather than on Hesper, so we will not put a go-live date on a web page. The schedule is set by three things: how your ClaimCenter instance is deployed and how change is released into it, how long your security and data reviews take, and how wide a first slice you want. The sequence itself is the same everywhere.
- Scope: one line of business, one trigger, and a written definition of which claims are in scope and which never are.
- Access: a connection to a non-production ClaimCenter environment, with the field and document set agreed in writing.
- Calibration: Hesper runs against closed claims your team already knows the answer to, and you compare its output with what actually happened.
- Write-back: notes, documents and activities begin landing on live claims, read-only first if your team prefers.
- Widen: more triggers, more lines of business, more of the lifecycle, at whatever pace your claims leadership sets.
Anyone who quotes you a go-live date before seeing your change-release calendar is guessing. Bring your architecture to a demo and we will scope it against what you actually run.
Where does the claim data go, and who can see it?
Claim data moves over encrypted connections and is encrypted at rest. Hesper is SOC 2 Type I attested and GDPR-ready. Claim data and Hesper's output are retained for the subscription term unless you instruct otherwise, and your claim data is never used to train models.
Every agent action is logged with its sources, its reasoning and a timestamp. That is the part that matters when a regulator, a reinsurer or a capacity partner asks how a decision was reached. The audit trail in ClaimCenter stays authoritative; Hesper's log explains the work behind the entries it added to it.
What does Hesper do, and what does your team decide?
| Hesper's agents | Your team |
|---|---|
| Reads the claim, its exposures and every document attached to the file | Decides which lines, claim types and severity bands are in scope |
| Runs coverage, investigation, valuation and recovery checks in parallel | Sets the authority matrix the results are routed under |
| Verifies entities, vendors and addresses against external sources | Approves which external sources may be used |
| Writes a cited findings note, an evidence package and drafts onto the claim | Edits, signs and owns anything that leaves the file |
| Opens an activity naming the recommended next action | Makes the coverage, reserve, settlement, referral and denial decisions |
| Logs every step with its sources and a timestamp | Keeps ClaimCenter as the authoritative record |
What does the output look like?
- Status
- Illustrative walkthrough, not a customer file. No Hesper customer, pilot or prospect is described here.
- Claim
- Auto, first-party collision, single exposure
- Trigger
- New document on the file: a supplemental repair estimate
- Read from ClaimCenter
- Loss facts, exposures, parties, coverages, and the documents on the claim
- Agents run
- Coverage analysis, document forensics, entity verification, timeline reconstruction
- Written back
- One findings note, one evidence package as a claim document, one activity for the adjuster
Evidence cited
- The supplemental estimate's metadata shows the file was regenerated after the date printed on it, with the line-item total changed.
- The repair facility's billing address resolves to a residential unit, and the same address appears on two other open claims in the book.
- The insured's recorded statement places the vehicle at a location the photo timestamps contradict by two days.
Recommended next action: Route to the investigation queue with the evidence package attached, and hold the supplement pending a second estimate. Hesper makes no payment, no denial and no reserve change; the adjuster and supervisor do, in ClaimCenter.
Frequently asked questions
Is Hesper a Guidewire partner or listed on the Guidewire Marketplace?
No. Hesper is not a Guidewire partner, is not listed on the Guidewire Marketplace, and holds no Ready for Guidewire validation. We would rather say that plainly than imply an endorsement we do not have. The integration does not depend on one: it runs against the API, messaging and file-based interfaces your ClaimCenter already exposes to other downstream systems.
Does this work with ClaimCenter on Guidewire Cloud as well as a self-managed install?
Both are workable, and the difference is mostly in how change is released rather than in what Hesper needs. What matters is which interfaces your team is willing to open, who controls the release calendar for that environment, and whether a non-production instance is available for calibration. Bring your deployment shape to a demo and we will scope against it instead of guessing.
Will this force changes to our ClaimCenter configuration?
It should not need to. Hesper reads the claim and writes back as notes, documents and activities, which are native concepts your configuration already uses. Some carriers choose to add a field or a queue so Hesper's output is easy to filter and report on, but that is a preference rather than a requirement.
Can Hesper pay, deny or close a claim on its own?
No. Hesper does not release payments, change reserves on its own, deny claims or close files. It produces evidence, findings and drafts, and it opens an activity naming the recommended next action. A person with the right authority makes every decision, and that decision is recorded in ClaimCenter.
We already run a fraud detection tool next to ClaimCenter. Does Hesper replace it?
It does not have to. Hesper has built-in fraud detection and works standalone, and it also works alongside FRISS, Shift or Verisk where a carrier already runs one. The difference is what happens after a claim is scored: a detection tool hands a flag to a person, while Hesper runs the workup behind the flag and attaches the evidence. See fraud investigation.
Do we have to automate the whole claims lifecycle to start?
No. Hesper can pick a claim up at any stage. Carriers running ClaimCenter usually start with one stage and one line of business, often claim file review or investigation, then widen once the output has been checked against claims their own team closed.
Other claims systems Hesper works beside
Book a demo on a ClaimCenter claim
A 30-minute walkthrough on your own claim types: what Hesper reads off the file, what it writes back, and what stays your team's call.
Book a demo