01Claims47 · The claims platform
The AI agent that runs your insurance claims operation, end to end
Claims47 is the AI platform that runs the whole reimbursement claims operation for insurers, TPAs and assistance companies: intake and triage, evidence and integrity checks, fraud screening, a recommended decision with the amount, the payment-ready file, customer communication and insights. One AI agent with four capabilities, each with specialist skills; the rules engine computes every amount and the adjuster decides. You set the dial on how much runs straight through.
Runs in M47 Cloud. Plugs into your systems, or works as the system itself.
02The product
From any document to a payment-ready decision
Detected type
— · the claim type is detected from the customer's own words, with a confidence scoreDecision
Press Run live demo to watch the claim processed check by check.
Product view. Synthetic data.
claims & insurance processes run per year
capabilities in one agent: intake and triage, evidence, communication, insights
of decisions auditable
typical time to a payment-ready decision
03What it achieves
Cost. Time. Trust.
Cost
of the end-to-end claims process automated. Every manual touch removed is cost removed.
Time
from a complete claim to a payment-ready decision. Not weeks of chasing.
Trust
of claims screened before a decision, every decision auditable after. Not a 10% sample.
04Built for real claims operations
Made to run inside a real operation, not a demo.
You set the dial
Up to 80% runs straight-through within thresholds you configure. Exceptions reach the adjuster with the reasoning attached and exactly what the file needs to decide: a better process first, then automation.
Your rules, encoded
Coverage rules, tariff catalogs and per-client business rules configured, versioned and audit-ready. Not hard-coded.
Every decision explained
Every line checked against the report or the photos, with the page and the sentence that supports it. Full audit trail, plus the operation's metrics: approved, touched manually, throughput per day, where manual time goes.
DocFraud47 forensics
Pixel-level document forensics on every claim. In one production program, 52% of rejections were forged documents a reviewer could hardly see.
Runs in M47 Cloud, connects to your stack
The AI core runs in M47 Cloud (ISO 27001, SOC 2, GDPR compliant, HIPAA-ready with BAA) and connects securely to your policies, CRM, mailboxes and systems of record. Nothing to install, nothing to replace.
Setup fee, then pay per claim
A one-time setup fee for configuration, integrations and rules, then a price per claim processed. No licenses, no seats. Start light, prove the ROI, then scale.
05Configuration
Configured for you. Nothing to code.
Every panel is configuration your team sets. No code.
- Inbound emaile.g. Microsoft 365, Google Workspace, or any mailboxConnected
- Your core, over APIclaims pushed from your claims system, whichever it isConnected
- Web formstood up by M47Live
- Policies & customersread from your database, read-onlyRead-only
- CRM & system of recorde.g. Salesforce, Dynamics, or yoursConnected
- Provider network & tariffsyour catalogs, synced nightlySynced
- Payout systemthe payment-ready file, delivered downstreamConnected
Names are examples. Claims47 plugs into the systems you already run, whatever they are.
06The Claim Journey · Who acts at every step
AI understands. Rules compute. The adjuster decides.
01The claim arrives
Intake & triage · Capture reads the email, the web form or the API payload and every attachment, identifies the claimant and the policy, and opens the claim file.
02Claim type detected, documents checked
Triage detects the claim type from the customer's own words, with a confidence score, and classifies each document: invoice, receipt, report, photo.
Completeness: the documents required for this claim type, checked before any money logic runs.
03Documents read, lines matched, evidence cited
Evidence · Extraction pulls lines, amounts, dates and providers out of PDFs, photos and scans. Matching pairs each line with your benefit catalog. Evidence assessment cites the page and the sentence that supports it. DocFraud47 screens every document at pixel level. Any language and currency: translated with the terminology preserved, amounts converted at the service-date rate.
04Missing information requested
Communication · Request writes and sends the email to the customer automatically, in their language, and reads the reply when it arrives.
Decides what is missing: a required document, a field that could not be read, a mismatch to clarify.
Can ask for anything else from the claim file, in one click.
05Eligibility and integrity checked
DocFraud47 findings and provider billing patterns feed the integrity check.
Policy in force on the service date, waiting period, limits and sub-limits, totals that reconcile, known provider, no duplicate. Any flag routes the file to Review.
06Amount computed, decision recommended
Per line, the lower of what is substantiated and what remains: tariff times reimbursement rate, minus what the evidence does not support. Straight-through within your thresholds, or Review with the reason attached.
07The adjuster decides
Recalculates the amount on every change. Every change stays in the audit trail.
Reviews the recommendation with the evidence attached, changes a line if the file justifies it, accepts or rejects. Mark as Accepted.
08Outcome explained, file payment-ready
Communication · Decision explains the outcome line by line, with the reasons, to the customer or the provider. Insights adds the claim to the operation's metrics and patterns.
The payment-ready file, with the decision, the evidence and the reasons, goes to your systems.
Always on: Status keeps the customer informed at every step, so nobody calls to ask. Every check applies your rules, your tariffs and your data.
One invoice line, three actors
Follow one line of a vet invoice from the document to the amount
Vet invoice
Clínica Veterinaria Levante · INV-2026-0905-063 · 2026-09-05
Vet report · p.2
“A basic biochemistry profile covering six parameters was run.”
- Extracted
- Biochemistry panel, 15 parameters · 1 × 70.00 EUR · 2026-09-05
- Matched
- PA0114 · Biochemistry panel, 15 parameters · 100%
- Evidence
- Partial6 of 15 parameters documented
Result
18.24 EUR
Annual maximum 2,000.00 · used 0.00 · no cap applied
Same inputs, same output. Versioned rule set.
A corrected report arrives: 15 parameters were run.
Recommended for this line
The AI never sets the amount. The rules never read the report. The adjuster changes both with one click, and it stays in the audit trail.
No payout is a model's guess. Every amount comes from explicit, versioned rules you control, over facts a person can inspect: reproducible on appeal, defensible before a regulator, and in line with automated decision-making rules such as GDPR Art. 22.
07The Claims47 agent
One agent, four capabilities
Intake & triage
Every submission, from any channel, language, currency or format, becomes one clean, structured, correctly-routed case.
See the skills (2)
Reads whatever arrives (emails, PDFs, photos, forms), extracts and structures all the data, verifies documents are complete, writes a smart summary.
Identifies the policy and the product, detects the claim type from the customer's own words with a confidence score and explains why, routes the case to the right queue, detects duplicates.
Evidence
Turns the documents into facts the rules can compute: every line extracted, matched to the benefit catalog, the provider and the beneficiary, and checked against the report or the photos, with the page and the sentence cited.
includes DocFraud47See the skills (4)
Line items, amounts, dates, provider, beneficiary and vehicle or property lifted from PDFs, photos and scans into a typed structure. Medical coding (ICD-10-CM/CPT) where the line of business needs it. Any language and currency: automatic translation with the medical terminology preserved, amounts converted at the service-date rate.
Each line matched to your benefit catalog and its limits, the provider to your network, the beneficiary to the policy, with a confidence score on every match.
Reads the whole file and returns a verdict per line: supported, partial, unsupported or needs information, with the reason and the citation. The rules engine then computes eligibility, integrity and the recommended amount.
The fraud specialist. Pixel-level document forensics plus pattern analysis across claims, providers and members. Catches what humans can't see.
Communication
Your customer always knows where their claim is. Nobody calls to ask.
See the skills (3)
Proactive updates at every state change (received, in review, decided, paid), any channel, any language.
Asks the customer for exactly what's missing, tracks the response, automatically resumes the claim when it arrives.
Communicates approvals and denials with concrete, clause-level reasons in clear language.
Insights
Every claim reviewed becomes business intelligence: what you pay, to whom, for what, and where money leaks. Built on reviewing up to 100% of decided claims.
See the skills (2)
Cross-checks invoices against medical acts: prices out of range for the same act, treatments that don't match the diagnosis, patterns per provider.
The bulk view across claims: emerging fraud patterns, provider benchmarks, cost drivers, and recommendations your operation can act on.
08Specialized for different claims environments
One engine. Everything starts with the policy: product, coverages, benefits and rules.
Claims47 for
Pet Reimbursement
Semantic coding of veterinary acts, priced against your own tariff catalog. Live in weeks, not months.
Claims47 for
Home Insurance
Water, glass, electrical, fire or theft: the coverage detected from the description, the photos read as evidence, every line matched to the policy.
Claims47 for
Auto Insurance
Roadside assistance and repair reimbursement: invoice and proof of payment read, every service matched to the partner tariff and the policy.
Claims47 for
Travel Insurance
Reimbursement claims for medical bills, baggage, delays and cancellations. Any document, any language, any currency, processed like it came from next door.
Claims47 for
Health Insurance
Consultations, hospital stays, pharmacy or dental: every bill coded and priced against your benchmarks or tariffs. US bills included, where 837 auto-adjudication ends.
09Proof
From deployments in production
claims per adjuster and day, health program, first weeks in production (40-50 → 100)
typical decision on a complete claim, any line of business
documents auto-structured at intake, any language or currency
claims reviewed by AI in one travel-medical deployment
ICD-10 / CPT coding accuracy, health program in production
"Claims47 helped us transform a slow, manual process into something faster and more consistent. The AI system assists our teams in reviewing medical forms, reducing human workload by up to 80%, while ensuring that final decisions remain in expert hands and fully comply with regulatory standards."
Four production references: a travel-medical TPA, an international health plan, a pet-insurance carrier and a health mutual. References available under NDA.
10The collaboration model
We grow when you grow
Start light
A one-time setup fee covers configuration, integrations with your systems and your business rules: live in weeks, in M47 Cloud.
Prove the ROI
First claims run on your live flow. Cost and time per claim are measurable from day one.
Scale together
Then a price per claim processed: our revenue follows your operation. When it grows, we grow.
Pricing, in one line:
a one-time setup fee for configuration, integrations and business rules, then a price per claim processed. No licenses, no seats.
11Beyond claims
For insurers, assistance companies and TPAs, the same architecture runs the whole operation: assistance, coordination, engagement, network and finance.
AI for Insurance OperationsAssistance
Coordination
Engagement
Network
Finance
12Book a demo
Watch a claim like yours become a payment-ready decision
A 20-minute demo on synthetic data. Then we show how your policies, tariffs and rules plug in.