AI use case

AI copilot for insurance pricing and actuarial analysis

AI that speeds up the work of pricing and actuarial teams, from automated, transparent risk and demand model building to natural language analysis of rate filings, experience data and reserving diagnostics, while actuaries select the models, sign off the rates and own the professional judgment.

By Len Debets · Last verified 26 September 2026 · 5 public deployments

5x
Reported productivity gain
Generali France, organization claim.
Up to 2 days
Cycle time
Europ Assistance (vendor claim).
USD 360,000 to USD 1.4 million
Indicative value per year
An insurer with a pricing and actuarial team of 20 people. Worked example, see how it is calculated.

What problem does it solve?

Pricing and actuarial teams work to fixed review cycles with a high volume of work. Building or updating a risk model in traditional tools can take weeks of data preparation, variable selection and code, often split across separate tools such as SAS, Python or R, where every change to the data means manual code updates that are hard to version and audit. Competitor filings, experience studies and reserving reviews compete for the same people.

The constraint is not only speed. Pricing models must be explainable to supervisors, tested for unfair discrimination and consistent with fair value rules, so machine learning that improves accuracy but cannot be explained is hard to defend. The opportunity is AI that removes the manual work while keeping models transparent and actuaries in control.

How does it work?

  1. Prepare the data. The platform imports policy, claims and quote data, handles missing values and builds candidate features, including approved external data.
  2. Build transparent models faster. Automated search explores thousands of variable combinations and interactions and proposes interpretable models (for example generalized linear or additive models), while actuaries choose and adjust the final model.
  3. Set rates. Actuaries compare pricing scenarios and their effect on volume, loss ratio and fairness tests, then approve the rates.
  4. Ask questions in plain language. A generative assistant answers questions over rate filings, experience data and reserving outputs, and drafts code or documentation for review.
  5. Deploy and monitor. Approved rates go to the rating engine through an audited deployment, and model performance is monitored against actual experience.
Audience
Employee facing
Autonomy
Copilot
Adoption
Early adopters
Channels
Internal tools

What is it worth?

Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.

Value benchmarks for AI copilot for insurance pricing and actuarial analysis
KPIMedianReported rangeData pointsClaimed by
Productivity gainToo few to pool
5x
11 organization
Cycle timeNot pooled
Not pooled: up to 2 days
0plus 1 up to1 vendor

Value drivers: Speed and cycle time, Employee productivity, Risk and loss reduction, Compliance quality.

Indicative value

An insurer with a pricing and actuarial team of 20 people

USD 360,000 to USD 1.4 million

Actuarial capacity released per year

How this is calculated

Formula: actuaries * modellingShare * timeSaved * costPerPerson. The low scenario uses every low input, the high scenario every high input.

InputLowHighBasis
Pricing and actuarial staff actuaries, people2020The reference insurer.
Share of their time spent on model building and data preparation modellingShare, fraction of working time0.30.5Editorial assumption. Replace with your own time allocation.
Share of that time the AI removes timeSaved, fraction of modelling time0.50.75Conservative against the evidence on this page (Generali France's actuarial studies manager reports modelling five times faster; Akur8 reports that Europ Assistance cut work that took weeks to one or two days).
Fully loaded annual cost per person costPerPerson, USD per year120,000180,000Editorial assumption. Replace with your own fully loaded cost.

What it leaves out: Capacity released, not cash saved. It leaves out the usually larger value of better risk selection and faster rate changes on the loss ratio, software licences, and the validation and governance effort that pricing models need.

Who already uses it?

5 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.

Accelerant Holdings

Global · Insurance · 2025

ProductionGrade B

Accelerant runs a risk exchange that connects specialty MGAs (its Members) with risk capital. Its 2025 annual report says incoming data, from Member bordereaux to third party sources, is validated, transformed and governed using AI agents, that internally developed AI tools and models assist Members' underwriting, and that its risk evaluation tools help Members identify, classify, validate, research and price underwriting opportunities. Members also get AI supported claims insights, actuarial analysis and portfolio management to manage rate adequacy. Engineers, data scientists, product managers and designers made up 34% of its workforce at the end of 2025. No outcome figures are disclosed.

No outcome disclosed.

Kinsale Capital Group

United States · Insurance · 2025

ProductionGrade B

Kinsale, a US excess and surplus lines insurer that sources about 95% of its premium through wholesale brokers, told investors in January 2026 that AI driven routing improves the accuracy of submission routing and underwriter productivity, alongside an average submission clearance time of 9 minutes. Its 2025 annual report says it gave every employee an enterprise AI tool licence in 2025, that use is most prevalent in its IT, actuarial and analytical teams with selective use in underwriting, and that it also uses internally developed agents. Kinsale does not attribute the clearance time to AI, so no metric is recorded.

No outcome disclosed.

Europ Assistance

France · Insurance · 2026

ProductionGrade C

Europ Assistance adopted Akur8's cloud pricing platform, whose automated modelling cut the time spent running and updating pricing models. The vendor reports that work that took weeks now takes one or two days, that teams can reuse fitted models on new datasets, and that built in documentation lets stakeholders review and challenge the whole pricing process.

  • Cycle time: up to 2 days, pricing model execution and updates
    "As a result, what once took weeks in the pricing process can now be completed in just one or two days."
    Claimed by: vendor

Generali France

France · Insurance · 2026

ProductionGrade C

Generali France's actuarial studies team uses Akur8, a pricing platform that automates the repetitive parts of building risk models while keeping the process transparent and auditable for actuaries. Its actuarial studies manager says modelling is five times faster and that the shared interface improved communication inside the team.

  • Productivity gain: 5x, speed of pricing model building
    "Modeling speed is 5x faster, while keeping a thoroughly transparent and auditable process."
    Claimed by: organization

MAIF

France · Insurance · 2026

ScaledGrade C

MAIF, a French mutual insurer, moved its pricing workflow from separate SAS and Python tools into Akur8, where data preparation, model building and geographic modelling happen in one place. The vendor says its machine learning explores thousands of variable combinations in parallel to find the most predictive features while actuaries keep control of the final selection. MAIF's pricing teams now manage several hundred databases of up to 30 million rows and several thousand models and versions on the platform; no time saving is quantified.

No outcome disclosed.

How do you implement it?

A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.

Data you need

  • Clean, joined policy, exposure, claims and quote data at the level pricing needs
  • Approved external data sources with documented use conditions
  • Model governance standards and fairness testing methods
  • Rate filings and experience studies in a searchable form

Systems to integrate

  • Data warehouse or lakehouse holding policy and claims data
  • Rating engine for deployment
  • Model risk management inventory
  • Version control and documentation repository

Complexity: High

Automated modelling platforms already run in production at insurers such as MAIF and Generali France; the effort is in data quality, integration with rating engines, model governance and fairness testing. Generative assistants for actuarial analysis are newer and need controls on data access and on code they generate.

  1. 1

    Pick one product with a rate review due

    Run the new approach in parallel with the existing process on a real rate review so the results can be compared on accuracy, time and explainability.

  2. 2

    Keep models explainable by design

    Prefer methods that produce interpretable models (GLM or GAM style) or add robust explanations, because regulators and fair value reviews will ask why each factor is there.

  3. 3

    Build fairness testing into the workflow

    Test rating factors and outcomes for proxies of protected characteristics before approval, as rules such as Colorado's SB21-169 require for covered lines.

  4. 4

    Govern generated code and analysis

    Treat code or documentation drafted by a generative assistant like a junior's work: reviewed, tested and versioned before it touches production models.

  5. 5

    Monitor after deployment

    Compare predicted and actual experience monthly and set triggers for review.

Guardrails

  • Actuaries select and sign off every model and rate; nothing deploys without approval
  • Fairness and proxy testing before any new factor is approved
  • Full lineage from data to deployed rate, with version control
  • Generative assistants have read only access to data and cannot deploy
  • Price optimization constrained by fair value and renewal pricing rules where they apply

KPIs to instrument

  • Elapsed days from data extract to approved model, per review
  • Models built or refreshed per actuary per quarter
  • Predictive lift of new models versus the current rates on holdout data
  • Fairness test results per model version
  • Actual versus expected loss ratio after deployment

Human in the loop

Actuaries own model selection, rate approval and professional sign off. Model risk management validates pricing models independently, and compliance reviews fairness test results and rate filings before they go to regulators.

Common failure modes

More accurate, less explainable
A complex model wins on lift but cannot be explained in a filing. Set explainability requirements before modelling starts.
Proxy discrimination
External data or interactions act as proxies for protected characteristics. Test outcomes by group and document the reasons for every factor.
Speed without governance
Faster models mean more changes than validation can keep up with. Scale validation capacity with modelling capacity.
Unreviewed generated code
Code drafted by an assistant contains a subtle error that flows into rates. Require review and tests for every change.

What are the risks and rules?

EU AI Act

Depends on design

Pricing and risk assessment of natural persons for life and health insurance is high risk under Annex III point 5(c). Pricing for property and casualty products, and actuarial analysis that does not price individuals, are not listed, although supervisors still expect sound model governance.

Guidance

Controls to put in place

  • Model inventory entries with owners, validation status and approval history
  • Fairness and proxy testing records per model version
  • Rate change approval workflow with actuarial sign off
  • Access controls separating analysis tools from production deployment
  • Post deployment monitoring with documented triggers

When it went wrong elsewhere

Frequently asked questions

How much faster does AI make insurance pricing?
Figures published by the vendor Akur8 report large gains: Generali France's actuarial studies manager says modelling is five times faster, and Akur8's case study says Europ Assistance now completes pricing work that took weeks in one or two days. These are not independently audited and do not measure effects on loss ratios.
Is AI pricing high risk under the EU AI Act?
For life and health insurance of individuals, yes, under Annex III point 5(c). For motor, home and commercial lines the Act does not list pricing as high risk, but fairness, explainability and national pricing rules such as the FCA's renewal pricing rules still apply.
Do actuaries still decide?
Yes. The tools on this page automate data preparation, variable search and analysis, while actuaries select models and sign off rates. Kinsale reports that AI tool use is most prevalent in its IT, actuarial and analytical teams.

How to cite this page

Blits.ai AI Use Case Library, "AI copilot for insurance pricing and actuarial analysis", last verified 26 September 2026, https://www.blits.ai/ai-use-cases/insurance-pricing-and-actuarial-copilot. Licensed under CC BY 4.0. Method: how we verify use cases.

Changelog
  • 27 September 2026: First published

Related use cases

Insurance

AI copilot for underwriting risk assessment

A copilot that assembles everything relevant to a risk (the submission, loss history, internal guidelines, third party data and public information), highlights exposures and gaps against the insurer's underwriting guidelines and drafts the underwriting narrative or referral note, while the underwriter makes and signs every decision.

Deployments
8 public, best grade B
Reported cycle time reduction
50%
Generali Global Corporate & Commercial, vendor claim
BankingInsurance

AI copilot for model risk validation and monitoring

A copilot for independent model validation and review, whether run by a bank's validation function, an external tester or a supervisor, that checks model documentation against the model risk standard, generates and scores challenger tests (for generative AI, often with an LLM as a judge calibrated against human experts), watches production models for drift and drafts and consistency checks the validation report. An accountable validator owns every conclusion.

Deployments
3 public, best grade B
Autonomy
Copilot
Insurance

AI for insurance renewal processing and customer retention

AI that prepares and runs the renewal cycle: it digitizes renewal submissions and changes in risk for underwriters, flags policies at risk of lapsing or leaving, prepares the renewal conversation and answers customers' renewal questions, while renewal prices stay governed by the insurer's pricing rules and fair value obligations.

Deployments
3 public, best grade B
Autonomy
Copilot
Cross industryBanking

Governed text to SQL analytics assistant

An assistant that turns a business user's plain language question into a query against governed data, runs it under that user's own data permissions and returns the table or chart together with the SQL and the tables used, so routine ad hoc questions no longer queue for the data team.

Deployments
3 public, best grade B
Autonomy
Assist
Insurance

Conversational AI for insurance quote and buy

A customer facing AI agent that sells insurance directly in a conversation: it asks the rating questions in plain language, explains cover options, returns a price from the insurer's rating engine, handles objections and takes payment to bind the policy, with a licensed human available for advice and anything outside its limits.

Deployments
1 public, best grade B
Autonomy
Supervised agent