IOanyT Innovations

Insurance

Which submissions should a person see, and which can a rule decide?

For MGAs, TPAs and specialty carriers. Appetite, delegated authority and coverage run on written rules, and a wrong answer costs money or a regulator’s attention. Those are the decisions worth getting right first.

Book a Reliability Review Get a fixed quote

What to compile, what to keep on a model, what to leave alone

Typical calls, not a verdict on your workflow. The Workflow Diagnostic decides yours, on your own cases.

  • Submission triage and appetite checks

    Compile into code

    Appetite and referral rules are already written in your underwriting guides. The same submission should get the same answer every time, with the rule that decided it on record.

  • Delegated-authority and bordereaux validation

    Compile into code

    Checking each line against the binding authority’s limits is rule work, and every exception needs a trace a capacity provider can follow.

  • FNOL and claims intake with coverage checks

    Keep with an AI model

    Reading a free-text loss description is language work. A model extracts the facts inside a fixed schema; the coverage check itself runs as code.

  • Subrogation flagging

    Keep with an AI model

    Spotting recovery potential in adjuster notes needs language understanding. A model raises the flag; a person decides whether to pursue it.

  • Complex or disputed claim settlements

    Don’t automate

    Low volume and high judgement. Keep them with your adjusters, and give them the full decision trace instead.

What your regulator or auditor will ask

  • The NAIC adopted its Model Bulletin on the Use of Artificial Intelligence Systems by Insurers on 4 December 2023. As of 8 October 2026, 29 jurisdictions (27 states plus DC and Puerto Rico) had adopted it. It expects insurers to keep a written AI systems program covering governance, risk management and internal controls, including oversight of third-party AI systems and data.

    Source: NAIC adoption map, status as of 8 October 2026

  • Under the EU AI Act (Regulation (EU) 2024/1689, Annex III point 5(c)), AI used for risk assessment and pricing in life and health insurance is high-risk. Under the 2026 Digital Omnibus agreement, Annex III obligations apply from 2 December 2027.

    Source: EUR-Lex; Regulation (EU) 2026/1744

Last reviewed: 11 October 2026 by the IOanyT advisory team. This is general information, not legal advice.

Closest work we can show

Chosen by the kind of problem, not the industry. Each case says which sector it comes from.

A build case · Insurance, Zambia

Motor Insurance Digital Transformation

We built the platform that took an offline motor insurance market online, for customers, agents, administrators and the regulator. It shows how we work inside an insurance operation; it was not a decision-automation engagement.

Our own product · Contact centres

The Heartbeat

AI reviews 100% of calls instead of a 1% sample, and human analysts verify every flag: the same split between what a model reads and what a person decides.

Old system holding the rules?

Policy administration and claims systems often hold rules nobody has written down since the system was built. Before you replace one, write down what it actually does.

Legacy System Specification Assessment

How we’d start

  1. 1. AI Reliability Review. Free, 60–90 minutes, a 2-page findings note you keep.
  2. 2. Workflow Diagnostic. Two to three weeks to decide what to compile, keep on a model or leave alone.

Fixed price, agreed at scoping. Stop after any step. The findings are yours.

Tell us which decision is stuck.

We'll tell you in writing whether we can help.

Book a Reliability Review

Free · 60–90 minutes · a 2-page findings note you keep