Integration · Duck Creek Technologies
Claims automation that runs alongside Duck Creek Claims
Your core keeps the claim, the reserves and the audit trail. Hesper does the reading, the checking and the chasing, then files what it found onto the claim with every statement sourced.
Last updated
In short
Hesper AI is a claims automation layer that sits on top of Duck Creek Claims rather than a replacement for it. Duck Creek remains the system of record. Hesper reads the claim, its endorsements, its documents and its history over Duck Creek's own integration interfaces, works coverage, investigation, settlement and recovery on it, then files cited findings, an evidence package and editable drafts back onto the claim and opens a task on the diary. Your adjusters make every decision.
Does Hesper replace Duck Creek Claims?
No. Duck Creek Claims stays the system of record. It keeps the claim, the coverage types, the reserves, the payments, the authorization thresholds and the audit trail, and it stays where your adjusters work. Hesper writes no policy, moves no reserve of its own and releases no money. What it adds is a layer of work and evidence bolted to the side of the core, which is a smaller thing to buy and a smaller thing to unwind.
That distinction is the whole commercial case. A core replacement is a multi-year programme with a business case, a data migration and a regression plan attached. Connecting Hesper is a scoped integration across interfaces Duck Creek already publishes to other downstream consumers, and it can be narrowed to one product line or switched off entirely without unwinding anything. Nothing in your Duck Creek configuration changes shape because Hesper is reading it.
It matters technically too. Duck Creek Claims is built to record and govern a claim: business rules, routing, reserving, payment controls, and who did what under which authority. Hesper is built for the work that generates those records: reading the file end to end, retrieving and testing evidence against it, and drafting what comes out. One system governs the claim. The other works it.
How does Hesper get the claim file?
Over the interfaces Duck Creek already publishes for everyone else. Duck Creek describes its own architecture as open and API-first: 2,600+ APIs spanning policy, billing, claims, digital and external systems, an Integration Hub that standardises how a third party connects, and 100+ pre-built partner integrations already running through it. Batch and file-based feeds usually sit beside all of that. Hesper subscribes to whichever of them your team is willing to open, exactly as any other downstream consumer on the Hub would. No desktop agent, no browser extension, no screen scraping.
- A trigger, most often a diary task falling due or a document landing on the claim, since those are the two events a Duck Creek workflow already fires on. A reserve change, a status change or a nightly queue filtered by product line and severity work equally well.
- The claim itself, including the part nobody reads: the loss facts and parties, but also the coverage types, the endorsements attached to the policy, the reserve and payment history, and the notes left by anyone who touched the file before.
- The documents on the file: ACORD forms, the policy and its endorsements, contractor estimates and mitigation invoices, police and fire reports, photos with their metadata intact, recorded-statement transcripts and the correspondence thread.
- A written scope, which on Duck Creek is usually drawn along a product line because that is already how the configuration is organised: which lines Hesper may read, which claim types it may never read, and which claims stay inside your tenancy under any circumstances.
Pull, push and scheduled batch are all viable, and on Duck Creek the plumbing usually exists already: the Integration Hub is how the other pre-built connectors reach the claim, so Hesper becomes one more subscriber on a pattern your integration team has run before. A first slice rides an existing feed rather than a new one wherever it can.
What does Hesper write back to Duck Creek Claims?
Finished work, filed onto the claim where a Duck Creek user is already looking for it. Duck Creek's own claims page describes its investigation management as "a single hub for every task, document, and stakeholder", and its recovery workflow as diary-driven. Those are the slots Hesper fills: a document attached to the claim, a written summary on the file, and a task landing on the diary date your adjuster already works to. The claim number, the history and the workflow are untouched.
- A findings summary in plain language, sitting on the claim, where each assertion carries the invoice line, the photograph or the licence record it rests on rather than a confidence score.
- An evidence package filed as a claim document: the underlying records as pulled, a loss-to-repair timeline built from document dates and photo metadata, and every contradiction between them set out with both sides named.
- Drafts your team edits and owns: a coverage position, a reservation-of-rights or denial rationale, a demand response, or the subrogation demand that Duck Creek's diary-driven recovery workflow then carries to its own follow-up dates.
- A task carrying a recommended next action and a diary date, so the follow-up is governed by the same queue mechanics as the rest of the claim instead of an adjuster's memory.
What Hesper withholds matters as much as what it writes. It holds no payment authority, moves no reserve, denies nothing and closes nothing. Duck Creek's authorization thresholds stay the only gate on money leaving the building, and every Hesper output lands as a draft somebody approves, edits or rejects, recorded against that person exactly as hand-worked output is.
What does this change for the adjuster?
A Duck Creek adjuster opens the claim to a file that is already assembled. Duck Creek's own claims page describes its investigation management as a single hub for every task, document and stakeholder, and the hub is genuinely good at holding that material in one place. Holding it is not working it. Somebody still has to read the mitigation invoice against the recorded statement, chase the report that never arrived, and run the contractor's remittance address to ground.
Hesper does that reading before the adjuster gets there, and cites it. What lands in the hub is a finished workup rather than a folder of inputs, so the adjuster's first act on the claim is a judgement rather than an hour of assembly. Straightforward files clear with the checks on the record. The ones with a problem in them arrive with the problem named, the sublimit or the date conflict already pinned to the document it came from.
Decision authority does not move. Every call that costs money stays where it is: coverage, reserve, settlement authority, referral and denial, all under the thresholds your Duck Creek configuration already enforces. Claims triage covers how a file gets scored and routed before any of that; claim file review covers what the workup itself contains.
What does a Duck Creek integration actually involve?
The schedule is set by your environment rather than by Hesper, so we will not put a go-live date on a web page. What drives it is your side: whether you are on OnDemand or self-managed and who therefore owns the upgrade path, how long your security and data review queue is, and how many product lines you want in the first slice. The shape of the work is the same everywhere.
- Scope: one product line, one trigger, and a written list of the claim types that are never in scope, signed off before anything connects.
- Access: a non-production Duck Creek environment, reached over an interface your integration team already runs, with the exact field and document set written down before anything connects.
- Calibration: Hesper works a batch of your own closed Duck Creek claims and your team marks the output file by file against what the adjuster actually concluded, before anything touches a live claim.
- Write-back: summaries, evidence documents and diary tasks start landing on live claims, in shadow mode first if your claims leadership wants to watch before it counts.
- Widen: more triggers, more product lines, more of the lifecycle, on the cadence your release calendar allows rather than ours.
A go-live date quoted before anyone has read your release calendar is a number somebody made up. Show us the deployment shape and we will scope against it in the room. Pricing works the same way: it is set against the book and the product lines in scope, which is why there is no number on this page.
Where does the claim data go, and who can see it?
Everything moves over encrypted connections and sits encrypted at rest. Hesper holds a SOC 2 Type I attestation and is GDPR-ready. Your claim data is never used to train models, and both the source data and Hesper's output are held for the subscription term unless you instruct otherwise. Which Duck Creek fields and which document types are in scope is written down before the first connection opens, not negotiated after it.
Every agent step carries its sources, its reasoning and a timestamp. Duck Creek's audit trail remains the authoritative record of who did what to the claim; Hesper's log sits underneath it and explains how the entries Hesper added were arrived at. Those are the two records a market conduct examiner or a reinsurer actually asks to see together.
What does Hesper do, and what does your team decide?
| Hesper's agents | Your team |
|---|---|
| Reads the claim, its coverages, its history and every document attached to the file | Sets which product lines, claim types and severity bands Hesper may touch |
| Works coverage, investigation, valuation and recovery on the same file at once | Keeps the authority matrix and payment thresholds configured in Duck Creek |
| Checks contractors, repair facilities and remittance addresses against external sources | Approves which external sources may be used |
| Attaches a cited findings summary, an evidence package and editable drafts to the claim | Owns and signs every document that leaves the file |
| Opens a task on the adjuster's diary naming the recommended next action | Holds the coverage, reserve, settlement, referral and denial calls |
| Screens every file for subrogation and salvage and drafts the demand | Makes the pursue or waive call |
| Logs every step with its sources and a timestamp | Keeps Duck Creek Claims 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
- Commercial property, water damage, single location
- Trigger
- A new document on the file: a contractor's mitigation invoice
- Read from Duck Creek
- Loss facts, coverages and endorsements, parties, reserves and payments, and the documents on the claim
- Agents run
- Coverage and endorsement analysis, document forensics, entity and vendor verification, timeline reconstruction, recovery screening
- Written back
- A findings summary, an evidence document, a subrogation draft, and a dated task on the adjuster's diary
Evidence cited
- The mitigation invoice carries line items dated before the reported date of loss, and the invoice metadata shows the file was regenerated after the date printed on it.
- The contractor's licence number resolves to a different trading entity, and the same remittance address appears on two other open claims in the book.
- The policy's water-damage endorsement caps mitigation at a sublimit the submitted invoice exceeds, which the adjuster had no reason to spot inside a 140-page policy.
- A third party may have caused the loss, so the file is flagged for subrogation with a drafted demand attached.
Recommended next action: Route to the investigation queue with the evidence package attached, hold the invoice pending a second quote, and open the subrogation diary. No payment moves, no denial issues and no reserve shifts on Hesper's say-so; the adjuster and the supervisor do all three, in Duck Creek Claims.
Frequently asked questions
Is Hesper a Duck Creek partner or listed on the Duck Creek Content Exchange?
No. Hesper is not a Duck Creek Technologies partner, is not listed on the Content Exchange, and holds no Duck Creek certification. Saying so plainly beats implying an endorsement we have not earned. Nor does it need one. Everything Hesper uses is already open to the other connectors running through your Integration Hub.
Duck Creek launched its own agentic AI platform. Why would we add Hesper?
Because they solve different parts of the claim. Duck Creek announced its Agentic AI Platform on 28 April 2026, with an AI Gateway offering a marketplace, a registry and standardised protocols including MCP and A2A so Duck Creek, partner and customer-built agents can be integrated, and the named claims application at launch was Agentic FNOL at intake. Hesper is the kind of third-party agent such a gateway is built to connect, though we are not listed in theirs and claim no place in it, and Hesper works the claim after intake: coverage, investigation, settlement and recovery, with cited evidence behind each one. See Duck Creek's own announcement for what their platform covers.
We are on Duck Creek OnDemand. Does that change anything?
Both work. What separates them is not what Hesper needs but who controls the release path: on OnDemand the upgrade cadence is Duck Creek's, on a self-managed install it is yours, and that decides how quickly a mapping change can be pushed. The questions we ask either way are which interfaces your team will open, who owns the release calendar, and whether there is a non-production instance we can calibrate against first.
Will this force changes to our Duck Creek configuration?
It should not need to. Duck Creek ships a large body of pre-configured content, including coverage types, tasks and business rules, and Hesper writes back as documents, summaries and tasks, which your configuration already uses. Carriers who want Hesper's output reportable on its own sometimes add a queue or a flag for it, but that is a reporting preference and nothing depends on it.
Can Hesper pay, deny or close a claim on its own?
No, and it is built so it cannot. Hesper holds no payment authority, cannot move a reserve, cannot deny and cannot close. What it produces is evidence, findings and drafts, plus a task on the claim naming the action it recommends next. The person holding the authority makes the call, in Duck Creek Claims, against the thresholds you already configured.
Does Hesper replace the fraud detection tool we already run beside Duck Creek?
No, and the two usually sit better together than apart. Hesper detects fraud itself and runs standalone, but where FRISS, Shift or Verisk is already scoring the book, Hesper picks the claim up on the far side of the score. That is the gap worth pricing: a detection tool ends at a flag, and a flag is only worth what the investigation behind it is worth. With manual investigation running 14 or more days a case against 200 or more open files per investigator, roughly 75% of flagged claims are never fully investigated. See fraud investigation.
Other claims systems Hesper works beside
Book a demo on a Duck Creek claim
Half an hour on your own product lines: which Duck Creek fields and documents Hesper reads, what lands back on the claim, and where your authority matrix takes over.
Book a demo