An Avalon Health & partners venture

Stop the claim before the money leaves.

Hutano AI is the real-time fraud and authorization firewall for medical aids. It scores every electronic and paper claim before you pay, explains every flag, and gives your investigators a console to act, hosted in-country.

49 automated tests passing 6 of 8 fraud families at 100% recall ~1% clean-claim false positives
12
Detection rules, one 0–100 risk score
2
Channels: Health263 & scanned paper
100%
Paper ↔ electronic equivalence
$0.50
Per principal member / month
The problem

Money leaves before anyone checks.

Every medical aid pays claims it should never have paid. Fraud, waste and abuse are detectable at submission, yet most aids only find them after payment, if at all.

US$160M
Estimated annual loss to fraud, error & leakage across Zimbabwe's health-insurance industry (CIMAS CEO estimate).
US$60M
Estimated lost by PSMAS 2018–2022 through a defective claims system & undeclared purchases.
15–20%
Of claims lost to FWA in South African private healthcare (Health Funders Association).
477
Fraudulent claims from a single pharmacy for medication never dispensed (CIMAS).

Sources: CIMAS Health Group / allAfrica; ZimLive, Zimbabwe Independent, Newsday; Health Funders Association, Daily Maverick, Council for Medical Schemes. Zimbabwe figures are executive estimates, not audited totals; the SA 15–20% is a labelled regional proxy. We do not estimate a figure for your book, we baseline it with you, from your data.

The platform

A firewall for your claims pipeline.

Hutano AI sits between the switch (and scanned paper) and your adjudication system. Advisory and fail-open by design, a slow or failed score never blocks a legitimate claim.

Forensics engine

A hybrid of deterministic rules, peer benchmarking and ML anomaly detection, blended into one explainable score.

  • 12 detection rules, per line item
  • Upcoding, unbundling, ghosting, card sharing
  • Network views for collusion rings

Data protection

A fraud firewall handles health data. We treat that as a first-order constraint, not an afterthought.

  • Deployable fully in-country (Zimbabwe)
  • Tokenized identifiers, encrypted datastore
  • Licensed controller, DPO, full audit log

Investigator console

Not a dashboard of charts, a tool your Special Investigations Unit works in every day.

  • Risk-ranked queues & explainable flags
  • Provider peer benchmarks
  • Multi-scheme, role-based, whitelabel-ready
How a pilot works

Three months, free, in shadow mode.

We ingest a feed of your claims and score everything. It takes no action, nothing is held, declined or delayed. Your adjudication runs exactly as it does today.

Connect

The easiest feed

A batch export, a tee from your receiver, or a live copy, plus a paper sample.

Tune

Weekly review

We hand your team the top flags; they tell us signal vs noise, and we tune.

Report

Value report

What we caught, what it would have saved, the false-positive rate on your book.

Go live

One category

Advisory & fail-open, expanding only as accuracy holds. You decide.

Built on real infrastructure

Not a greenfield bet.

Hutano AI is productized from infrastructure Avalon Health already runs in production in this exact claims pipeline, a Health263 integration, a real-time switch adapter, and the Zimbabwe ZRVS tariff schedule (7,930 codes). We reuse proven parts, not a vendor starting cold.

A flag is investigative, never accusatory. No provider is sanctioned on an algorithm alone. Flagging logic and false-positive rates are auditable, with provider appeals, a lesson taken directly from South Africa's Section 59 findings on disproportionate flagging.

One working session. Then the clock starts.

Tell us your process and we fit into it. One session with your claims and IT leads is enough to agree the feed and start the three-month shadow pilot, at no cost, no operational change.