Step 01
Free diagnostic
We take a real chain — the opening of a claim, for instance — and read it end to end: volumes, documents, dead time, points of human decision, and the sensitive data at stake.
Sector — Insurance
In insurance, everything rides on turnaround time
AI use cases in insurance in Morocco: claims handling, underwriting, distribution network, duty to advise. ACAPS framework, Law 09-08 and the CNDP.
In brief
In insurance, generative AI first goes to work where documents pile up: claim declarations, supporting papers, underwriting files, questions from the distribution network. The framework, meanwhile, is set by the ACAPS, which expects sound governance and internal control, and by Law 09-08 for policyholders' data. Hunter BI designs systems where AI prepares the file and a human keeps hold of the act of insurance.
A policyholder does not judge their insurer on the premium. They judge it on the day of the claim — on the turnaround, on the number of times they were asked for the same document again, on the clarity of the answer. Yet between the declaration and the settlement, most of the time is not spent on expert assessment: it is spent reading documents, chasing missing pieces, re-keying data.
This is exactly where generative AI belongs. It reads an accident report, a quote, a certificate, an expert's report; it checks what is missing; it prepares the file for the handler. It does not set a cover, it does not grant a payout, it does not turn a policyholder away. Hunter BI works with Moroccan insurers, mutual funds and intermediaries to build that boundary into the tool itself — and to make it demonstrable to internal control.
Updated 14 July 2026
Sector stakes in Morocco
Insurance in Morocco: what the framework demands before the tool
The ACAPS, the authority that supervises insurance and social welfare, expects organisations to have a firm grip on their governance arrangements, their internal control and their operational risks — which includes the tools that underpin underwriting, claims handling and the relationship with policyholders. Introducing an AI system into these chains means introducing a new operational risk: it must be identified, documented, tested and monitored like any other. No text forbids it; every governance text asks you to know what the tool does and who answers for it.
Policyholder protection adds a demand specific to the trade. The duty to advise, the clarity of pre-contractual information and the handling of complaints are obligations for which the burden of proof falls on the organisation: an answer generated automatically and left untraced is an answer you cannot defend. On top of this sit Law 09-08 and the role of the CNDP, with particular sharpness whenever a product touches welfare or health cover: such data falls into the most protected category, and its passage through an AI tool must be explicitly framed — or avoided. We work through both of these before opening the slightest access.
Governance and internal control
The ACAPS expects a framework for governance and operational-risk control: an AI tool used in underwriting or claims falls within that scope and must be documented there.
Duty to advise and proof
The burden of proof falls on the organisation. Any AI-assisted answer or recommendation must be traceable, sourced and validated by an authorised human.
Health and welfare data
Health data falls into the most protected category under Law 09-08. Its use in an AI tool is framed explicitly with the CNDP — or kept out of scope entirely.
Use cases
What AI concretely changes — insurance
Every use case links to the Hunter BI offer that delivers it. We claim no result figures until they are measured at your organisation.
Sovereignty
Welfare and health cover: where sovereignty stops being theoretical
An insurer that distributes welfare, health or retirement cover handles data that the law places in the most protected category. A sick-leave note, a certificate, a hospital discharge report passed into a claim file: these documents have no business passing through a service whose location, retention period and access conditions you do not control. Caution here is not a pose — it is the condition for the processing to remain defensible before the CNDP.
Within these perimeters we deploy open models hosted in your own infrastructure or with a qualified host in Morocco: the medical document is read and structured without ever leaving your network, and logging stays entirely under your control. The rest of the company — drafting assistance, general questions from the network, market watch — can perfectly well rely on enterprise cloud platforms. This deliberate split is the right insurance architecture: the level of protection follows the sensitivity of the data, not the fashion of the moment.
Where to start
Three steps, in this order
In insurance, the first useful use case is almost never the most spectacular one: it is the one that unblocks turnaround time without touching the act of insurance.
Step 02
Pilot on one line
One line of business, one team, one indicator agreed in advance. The pilot produces the before/after measurement and the compliance file that internal control will ask for — both at once.
Step 03
Roll-out to the network
Once the chain is made reliable, the tool extends to the other lines and to the distribution network, with a per-seat subscription and continuous monitoring of answer quality.
Frequently asked
Can AI accept or refuse a claim in the handler's place?
No. Our systems prepare the file — reading the documents, extraction, completeness checks, summary — and stop dead before the act of insurance. Acceptance, refusal, assessment and settlement remain human decisions, traced and attributable. This is a policyholder-protection requirement as much as an internal-control one.
What does the ACAPS say about the use of artificial intelligence?
The regulator does not expect you to abstain, but to stay in control. The requirements for governance, internal control and operational-risk management apply to AI tools as to any other critical component: risk identification, documentation, testing, monitoring. An undocumented AI system is a blind spot in your risk-management framework.
Can medical documents be processed with an AI assistant?
Only within a framework built explicitly for it. Health data falls into the most protected category under Law 09-08: for these flows we recommend a sovereign deployment, where the document never leaves your infrastructure, and prior framing with the CNDP. Many companies choose, at first, to keep such documents out of scope altogether.
Can our network of agents and brokers use the same assistant?
Yes, and it is often the most cost-effective deployment: the network asks, in volume, questions whose answers already exist in your policy terms and internal notes. The assistant replies by citing the exact document, in the user's language, and you finally see what the field really asks for — data that no one collects today.
AI in your field: let's start with a diagnostic.
A free scoping session with a Hunter BI consultant: your data, your regulatory framework, the use cases worth launching first — and the ones better set aside.
- Réponse sous 24 h ouvrées, par un ingénieur
- Diagnostic gratuit, sans engagement
- Membre des réseaux partenaires OpenAI et Anthropic