Stories of real impact

Allstate: exploring insurance portfolios with quantum optimization

A collaboration with IBM turns a real underwriting challenge into a hybrid quantum-classical experiment, building evidence and skills for the next generation of optimization.

Independent editorialSource publication: June 18, 2026

Sector: Insurance
Organizations involved: Allstate · IBM
Maturity of the work: Hybrid research experiment

Independent editorial coverage of public industry research. StrategiQ did not participate in the work described and does not claim its results.

The business challenge

For a home insurer, risk is not simply the sum of individual policies. A wildfire or hurricane can damage many properties at once. Selecting a portfolio therefore means balancing value against uncertain, correlated losses and the company's tolerance for extreme events.

Allstate and IBM framed this as a chance-constrained knapsack problem: choose valuable items whose uncertain weights respect a budget within an agreed risk tolerance. It is a concrete business question, rather than an algorithm looking for a use case.

The approach

The experiment combined circuits on IBM Quantum Heron with classical processing. Quantum circuits sampled candidate portfolios; a classical recovery step repaired infeasible candidates and fed information about promising selections into the next round.

To address training difficulties as problems grow, the researchers also transferred circuit parameters learned on smaller instances to larger ones. The published paper reports experiments using up to 150 qubits, with classical methods serving as explicit benchmarks.

What was demonstrated

IBM reports that, with 30 minutes allocated to each method per problem, both the hybrid workflow and classical heuristics reached the exact optimum on instances up to 75 items. At larger sizes, the hybrid approach remained competitive with the strongest classical heuristic, with a marginal improvement under tight constraints.

The achievement is a tested workflow tied to underwriting, not a documented reduction in Allstate's operating losses. The research offers a baseline that can be revisited as processors improve and experiments reach more demanding scales.

The lesson for business

The practical lesson is to start with a measurable decision problem and a strong classical baseline. Business teams can define acceptable risk, data scientists can formalize constraints, and quantum specialists can test where a hybrid method contributes.

For organizations preparing today, capability building and honest benchmarking are valuable outcomes in their own right. They make future investment decisions more informed without requiring a promise of near-term quantum advantage.

Sources and further reading

This article is an original synthesis, not a reproduction of the sources. Findings are attributed to the organizations that reported them.