Human-authored finding
SupportedThe listed vehicle was scheduled for business use on the loss date.
Entered by claims staff · source · Policy declarations · p. 14
For auto claims organizations
AI agents built for
your insurance
company’soperations
Sentanis starts with represented demand response: catch the clock, deconstruct the file, and prepare source-linked work for review. Commercial-auto investigation carries the same record into fleet- and trucking-specific questions, sources, and staff review. Adjusters and counsel still decide, sign, file, and send.
Choose where you work
One claim record. Two accountable teams.
Sentanis begins with auto claim operations across non-standard personal and commercial lines. Claims and legal see the same evidence, while each retains its own duties and decision authority.
Claims operations
I lead claims operations
Catch a time-limited demand, open a commercial loss in the right investigation workspace, and put the adjuster’s next decision in one sourced view.
See the demand workflow →Carrier legal and defense
I handle claim litigation
Start with the same cited chronology, open questions, and recorded claim position—without rebuilding the file before legal work begins.
See the defense workflow →- First workflow
- Represented auto demand response
- Operating contexts
- Non-standard personal + commercial auto
- Evaluation measures—not claimed results
- Deadline recall · issue precision · time to review-ready
Experience behind Sentanis
First wedge · Demand response
Start with the demand. Expand across the claim.
See how Sentanis identifies the deadline, deconstructs the package, and prepares source-linked work for review. Then inspect the adjacent claim workflows. Every file shown is fictional.
Loading capability demo
Product expansion · Distinct auto solutions
One reviewed record. Distinct paths for each operation.
Route auto work to the right team, save human-authored findings across the claim, and turn unanswered questions into owned evidence requests.
Implemented in synthetic evaluation · production and customer acceptance pending
S
Northstar Casualty · Claim operations
Fictional product view · synthetic records
Human review active
Claim routing
Light commercial auto
CA-40817Two-vehicle intersection loss
Source remains separate
Imported type · Commercial automobile. The imported value does not become the reviewed routing decision.
Claim-wide reviewed work
Findings and evidence requests
Persists across investigation lenses
Evidence request
Human ownedDriver qualification file and trip purpose record
Approved for follow-up · not sent
The product
Four areas, one claim record.
Some modules are on our current product roadmap.
First notice of loss
InsuranceDocument intelligence
Every page is read, classified, and extracted on arrival. Each fact links back to the page it came from.
Standing claim brief
Every claim carries a short reading: what it is, where it stands, and what it needs next.
Coverage checks
Checks each claim against the policy for excluded drivers, lapses, and misrepresentation.
Claims intelligence
InsuranceCommercial-auto investigation
Fleet and trucking lenses over the same claim record, with driver, vehicle, and trip questions kept sourced.
Fraud signals and SIU triage
Explainable indicators for staged crashes, provider networks, and billing anomalies. SIU decides.
Ask the book a question
Describe a cohort in plain English and get back the matching claims.
Executive posture
Exposure, unanswered demands, and files over limits, counted from the claims themselves.
Valuation and settle-vs-defend
What the claim is worth, what defense will cost, and the exposure — side by side.
Demand response
Insurance + legalTLD Sentinel
Catches time-limited demands, explicit or disguised, and runs the countdown.
Demand & medical deconstruction
Turns a demand package into a cited medical and billing chronology, with buildup and duplicates flagged.
Citation QA
An independent audit checks every legal citation in a draft response. Counsel still signs.
Review gates
Decisions open on the evidence, not a recommendation. A person with authority records the call.
Litigation defense
LegalLitigation drafting
Answers, discovery, and routine motions drafted from the claim file. Counsel edits, signs, and files.
Deposition second chair
Outlines built from the full file, with contradictions against prior statements flagged.
Legal spend review
Panel-firm invoices checked against billing guidelines, line by line.
Safeguards
The AI never decides. That’s the design, not a disclaimer.
Claim CL-24-08817 · Time-limited demand
9 days leftSystem recommendation
Tender the $30,000 policy limit
- Specials exceed the limit. Billed medical $41,280 against a $30,000 BI limitBills p. 4
- Liability is clear. Rear-end collision; insured cited at scenePolice report p. 2
- The clock is running. Demand expires in 9 days; a lapse risks excess exposureDemand p. 1
Adjuster decision · required
Tender limits
$30,000
Counter-offer
Set amount
Request records
Extend review
Reject
Counsel review
Recorded to the adjuster by name, append-only. The system cannot choose for them.
The system never
- Denies a claim
- Rejects a demand
- Contacts a claimant
- Files with a court
- Releases a document
- Signs anything
The gates are wiring, not policy
Every settlement or tender decision goes to a licensed adjuster. Everything filed with a court or sent to opposing counsel is reviewed, edited, and signed by licensed counsel — that's Sean's side of the house. In the running system these are enforced below the interface: named permissions, gate-specific authority, and row-level security in the database decide who may record a decision, and the browser hiding a button changes nothing.
Source documents are evidence, never instructions
A demand package is an adversarial document written by someone who would very much like your system to do what it says. Imported text and attachments are treated as untrusted evidence throughout — the model reads them, and never takes orders from them. That distinction is architectural, and it is the one most AI claim tools get wrong.
An independent audit with no power to release
Courts have sanctioned filings that included fabricated AI citations. Sentanis runs a separate audit agent over every drafted response: it inventories each matter of law and each explicit authority, resolves them against claim-free legal sources, and binds numbered findings to that exact revision. Edit the letter and the proof is void. A passing audit is evidence for the person signing — it never releases anything itself.
Written for the jury that may read it
Courts have already compelled insurers to hand over their AI's claim evaluations. So decisions are append-only, evidence is recorded alongside the judgment a reviewer saw, feedback and later evaluation stay attached to the file, and a derived assertion carries the source line that produced it. Cheap to build in from the start, expensive to retrofit under a market-conduct exam.
Retrospective evaluation
Closed files in. Claim-level evidence out.
Start with a bounded cohort of represented auto bodily-injury claims. Sentanis replays only the information available at the time, produces no live action, and is scored against labels and reviewer rules your team approved first. It does not change a claim, contact a claimant, or send a response.
01
What you provide
A representative set of closed, represented auto bodily-injury files — routine files as well as known misses — plus their attachments, correspondence, event timestamps, and outcome labels your team approves.
02
What runs
The same demand-intake, critical-date, document-deconstruction, evidence-linking, and response-preparation pipeline shown above. It works from point-in-time records so later information cannot leak into an earlier recommendation.
03
What we measure
Deadline recall and precision, date accuracy, false-alert rate, lead time where the record supports it, issue completeness, evidence traceability, and blinded reviewer scores. Avoided loss and dollar savings remain modeled, not measured.
04
What you receive
A claim-level scorecard, every miss and false alert, the reviewer adjudication record, a data-flow summary, and a documented recommendation to stop, retest, or proceed to a bounded live shadow.
Carrier-owned gate
Stop, retest, or proceed. Your team owns the gate.
Before scoring begins, your claims, legal or compliance, and security owners agree the cohort, outcome rubric, data route, retention, and acceptance thresholds. No live workflow starts because a date arrived. Critical misses are reviewed, and the retrospective evaluation advances only when your team accepts the evidence.
A closed-file retrospective evaluation measures the bounded workflow above. It does not, by itself, prove portfolio-wide economic impact. Any avoided-loss or savings estimate is shown separately with its assumptions.
Longer-term expansion
The vision is broad. Permission expands one measured workflow at a time.
Sentanis is building toward a shared operating layer for insurance and legal work. Represented demand response remains the first evaluation wedge; non-standard personal and commercial auto are current product contexts. Each further cohort, jurisdiction, and workflow still earns permission through evidence and governance.
Scope
One defined workflow
Progression
Evidence, not elapsed time
Authority
People remain accountable
Closed files, agreed scoring
Retrospective evaluation
Start with a representative set of closed claims. Sentanis replays intake, demand identification, file synthesis, and response work against outcomes your team already knows. The evaluation set, baseline, and scoring rules are agreed before results are reviewed.
Evidence gate
Advance only when the carrier accepts the evaluation design and the results meet the thresholds set for the selected workflow.
Decision authority
Claims sponsor with legal, compliance, and model-risk review.
Deliverable
A scored evaluation pack: source-linked outputs, misses and false positives, baseline comparison, and a written recommendation on whether to proceed.
The sequence is fixed; the pace is not. A phase can continue, narrow, or stop when the evidence does not support expansion. Production decisions remain with the carrier and its licensed professionals.
From claim work to legal work
Claims catches the exposure. Counsel inherits the evidence.
The claim does not become a different matter when legal work begins. Sentanis keeps the chronology, sources, open issues, and named decision-makers intact across both teams—without a handoff packet to rebuild.
Fictional matter · claim 26WASYN000241
Elena Rivera v. Northline insured
Blue · evidence lineageOchre · human authority
Insurance operations
The adjuster sees the next claim decision.
Needs attention
Reported impact and treatment pattern diverge. Confirm the settlement position before the demand clock closes.
Decision packet
- 01 Standing claim brief, current to July 18
- 02 Liability and treatment inconsistencies
- 03 Demand deadline and open evidence gaps
Named authority
Jamie Diaz
Claims adjuster · records the tender decision
Shared source record
Evidence once. Chronology once.
JUN 11 · 08:42
Loss reported
Rear impact at a controlled intersection; no injury reported at intake.
Source · Recorded FNOL call · rev 1
JUN 12 · 14:32
Treatment begins
Passenger begins care after the reported loss.
Source · Northside clinic invoice · p. 1
JUL 18 · 09:05
Demand received
$145,000 time-limited demand; response due August 17.
Source · Rivera demand package · p. 2
Every summary, deadline, and draft points back to the source record and revision that supports it.
Legal operations
Counsel sees the same facts, ready for legal work.
Draft posture
Respond to the time-limited demand from the verified chronology; preserve the adjuster’s recorded position and surface the unresolved treatment gap.
Counsel starts with
- 01 Deconstructed allegations and damages
- 02 Cited facts bound to source revisions
- 03 Draft response held behind counsel review
Named authority
Maya Chen
Defense counsel · reviews, edits, and signs the response
First workflow · demand response
A focused way to start
Identify the TLD clock, deconstruct the demand and medical record, and prepare source-linked work for adjuster and counsel review.
Commercial auto · available for evaluation
The same record, specialized questions
Open commercial-auto, light-fleet, or heavy-trucking investigation profiles without copying the claim. Candidate records, exact pages, open questions, and the collection plan stay source-linked.
Broader institution · direction
A shared operating layer for insurance and law
Extend the same evidence, authority, and audit model across insurance lines and the legal work around them. That is the direction; broader permission is earned one measured workflow at a time.
Human review begins here
Start with one claim file. Prove it on a closed cohort.
The demo shows the workflow on a fictional file. A retrospective evaluation then scores one bounded use case on your closed claims, with the test and thresholds agreed in advance. Nothing reaches a live workflow unless your team approves the evidence.