AI that makes actuarial work smarter, faster and more explainable
AI at Addactis, Intelligence beyond automation
- Machine Learning
- GenAI / LLM
- Agentic
- Computer vision
Addactis embeds governed AI into actuarial workflows to automate repetitive tasks, explain complex results, challenge assumptions and support better-informed decisions. Proven actuarial engines ensure calculation accuracy. AI brings additional intelligence, actuaries remain in control.
Deterministic where accuracy matters, intelligent where judgement matters
What stays determinisitc
Actuarial and regulatory calculations run on controlled, reproducible and auditable engines. The same inputs always give the same results.
Controlled
Reproducible
Auditable
Where AI comes in
Wherever judgement is needed: exploring options, explaining drivers, challenging assumptions and comparing alternatives. AI informs the decision. It never takes the calculation’s place.
Explore
Explain
Challenge
Compare
What AI does in you actuarial workflow?
Every AI capability we bring to market answers three questions. What work does it take on? What better signal does it bring? And how can you check what it did?
Assist
Access actuarial knowledge and interact more naturally with complex tools and workflows.
Explain
Understand results, identify drivers and generate structured explanations for review.
Recommend
Challenge assumptions, compare alternatives and support actuarial choices.
Challenge
Brings a systematic second look. Alternatives you might not have had time to run are tested and backtested on your history.
What does AI do in your actuarial workflow?
Every AI capability we bring to market answers three questions.
What work does it take on? What better signal does it bring? And how can you check what it did?
Assist
Access actuarial knowledge and interact more naturally with complex tools and workflows.
Explain
Understand results, identify drivers and generate structured explanations for review.
Recommend
Challenge assumptions, compare alternatives and support actuarial choices.
Challenge
Brings a systematic second look. Alternatives you might not have had time to run are tested and backtested on your history.
Explainable by design, accountable by default
Understand what drives the results. Trace the actions and assumptions behind each recommendation. Review, challenge and validate AI-supported decisions with human oversight.
Four layers for AI without a black box
Speed comes from automationn and trust comes from transparency. Our model keeps both visible.
- Proven engines: Controlled, reproducible calculations. The same inputs always give the same results.
- Automation: Preparation, controls, reconciliation, documentation and repetitive runs, done faster and in one chain, with fewer hand-offs between files.
- Governed AI that explains and challenges: Your rules, your context, your decision history and specialist actuarial tools, used to explain, challenge and recommend.
- Explicit human control: Editable assumptions, reviewable recommendations, traced actions and clear human accountability.
AI in practice, workflow by workflow
Reserving: the Automated Reserving Agent
Non-life reserving still relies on actuaries testing combinations of methods and parameters by hand, comparing results and justifying every choice to auditors and regulators.
The Automated Reserving Agent, built into Reserving by addactis®, explores those combinations automatically, backtests each scenario on your data and recommends a combination for each origin period.
- 3 components: a Configurator that sets the parameters, an Engine that runs the calculations and an Advisor that selects.
- A knowledge base combining Addactis expertise, your own data and the memory of past closings.
- Results grouped by strategy: keep the previous method, adjust it slightly or change it. Actuaries rarely change method between closings without solid justification, and the Agent is built around that reality.

Regulatory reporting: Explainability for IFRS 17 and Solvency II
Explaining why indicators move often means rebuilding the calculation by hand before every committee. Explainability breaks down indicator variations, contextualises results and documents the analysis, so your committees understand where movements come from.
- Automatically identify the main drivers of variation.
- Generate documentation ready for audit and committees.
- Rely on explainable AI, under human validation.

AI-powered smart entry of benefit schedules
addactis® PO Predict automatically reads, extracts and structures health benefits from your tender specifications, whatever their format (native or scanned PDF, Excel, PowerPoint), with no re-keying.
Each extracted field comes with a confidence score and areas of ambiguity are flagged. The underwriter then confirms or corrects the data before it is saved, so the decision remains their responsibility. Underwriters start with the analysis, not the data entry.
- Multi-format: native or scanned PDF, Excel, PowerPoint, no prior formatting required.
- Structured extraction of benefits, reimbursement bases, caps and exclusions.
- A confidence score for each extracted field, with ambiguities flagged.
- Systematic human validation before saving.

Property risk: addactis® Property Risk Intelligence
Home insurance risk is often estimated at regional or local level, which hides the differences between neighbouring buildings. Next-door buildings face the same hazard but never the same risk.
addactis® Property Risk Intelligence uses machine learning and computer vision to assess each building. It calculates nine risk scores, graded 1 to 10, so underwriters have a consistent technical basis for acceptance, pricing and prevention.
- Machine learning and predictive models: algorithms trained on real-world data score each peril (storm, hail, flood, theft, fire…), continuously refined and updated.
- Computer vision: images of roofs, building condition and surroundings (trees, swimming pools…) are analysed to describe each building from the outside.
- Smart data: open and third-party data are qualified, corrected and enriched by business experts and AI, to describe buildings with 200+ attributes.
- Underwriters decide: policy-based alerts help justify acceptance, loadings, exclusions or prevention.

What changes for your teams
Actuarial processes that are faster to review, safer to validate and simpler to explain.
Speed
Shorten preparation, review and decision cycles, and spend more time on analysis.
Explainability
Understand and document results, drivers and assumptions.
Capacity
Explore more scenarios without multiplying complexity.
Risk Control
Reduce manual rework and strengthen traceability.
Knowledge
Capture methods, rules & past decisions in repeatable workflows.
Decision Quality
Challenge assumptions & decide on stronger evidence.
Everything you need to know about Actuarial AI
What can AI actually do in actuarial workflows? Addactis embeds AI into actuarial workflows to help automate repetitive tasks, explain results, explore alternatives and support recommendations. The capabilities available depend on the solution and the workflow.
AI-supported recommendations are designed to remain reviewable by actuarial professionals. In the Automated Reserving Agent, actuaries can examine proposed methods and parameters, challenge alternatives and validate the final selection.
The Agent explores combinations of reserving methods and parameters, backtests each scenario on your data and recommends the best combination for each origin period. It relies on three components (Configurator, Engine, Advisor) and on a knowledge base combining Addactis expertise, your data and the memory of past closings.
Explainability is the trust layer of our AI. It shows how a result was reached: the key steps, the main factors behind a variation and a clear validation path for the actuary. In IFRS 17 and Solvency II reporting, it breaks down indicator variations and documents the analysis, without rebuilding the calculation by hand.
Addactis aims to make AI-supported workflows reviewable and traceable. In the Automated Reserving Agent, methods, parameters and decisions can be documented and reviewed. The exact level of traceability and audit evidence depends on the solution and its available capabilities.
Data protection and confidentiality are essential requirements for AI in insurance. The applicable data handling, model training and security arrangements are detailed in our contracts.