Allstate and IBM are showing that quantum computing has the potential to help build better portfolios of risk and value, and solve hard problems in the insurance industry.

The joint work, published to arXiv in May 2026, addresses the “knapsack problem” from computer science, which has important implications for insurance companies. The knapsack involves finding the best possible combination of items to fill a container without exceeding its weight limit. In the classic knapsack problem, you’re given a group of items with different weights and values, and try to fill the bag with the most valuable possible collection of items without going over the weight limit. Many formulations of the knapsack problem have no practical solution in classical computing, particularly when the number of items gets very large.

The knapsack problem has direct application to insurance portfolios: Insurance companies aim to build balanced portfolios that meet customers’ needs while managing risk responsibly.

The realities of home insurance make balancing risk more complex. Risks are often connected in ways that are hard to predict. If a tornado impacts a community, a home insurer may end up paying to replace your house, your neighbor’s house, and every building in the area.

“On the homeowner side, it really requires us to be thinking from a portfolio perspective and not just from an individual risk perspective,” said Eric Huls, Allstate’s Chief Analytics Officer and Chief Data Officer.

Understanding highly correlated portfolios means computational complexity. When business problems are computationally complex, that’s a signal that quantum-enabled methods—which offer a new paradigm for solving complex problems—might offer value.

How insurers deal with complexity today

Car accidents are largely independent, meaning one driver’s crash has little bearing on the likelihood of other drivers experiencing an accident. Weather is the opposite. A wildfire usually doesn’t just damage one house. It can damage a whole region’s housing stock at once. Hurricanes and hailstorms behave similarly, meaning that the risks across policies in a homeowners insurance portfolio tend to be highly correlated. Those events are among the biggest factors affecting homeowner policies across a portfolio, according to to Jean Utke, Data Scientist and Technical Director in Allstate’s Insurance Product organization.

That correlation changes how an insurer has to think. Allstate can’t judge each property on its own and add up the risks. It has to look at the whole portfolio and ask how bad things could get when many losses strike at the same time, and whether that worst case stays within the company’s risk tolerance. That makes finding the optimal packing of policies into the “knapsack” of an insurance portfolio a very complex version of the knapsack problem.

Today, Allstate answers that question through classical simulations. The company might run 100,000 scenarios of possible futures to understand the range of outcomes, said Utke. The trouble is the rare, expensive events.

“The challenge is that when you’re looking at 1% of tail events out of 100,000 simulations, you’re down to 1,000 events. And there’s a fair amount of uncertainty within those estimates, particularly when you’re looking at many different types of peril across very large geographies,” Utke said.

A simulated wildfire in southern California might not tell you much about hurricanes along the East Coast or hailstorms St. Louis, Missouri.

Quantum computing could offer ways to approach the decision more directly, Utke said, rather than through simulation and approximation.

A quantum approach

Allstate’s version of the knapsack problem is especially hard. The losses, or weights, aren’t fixed numbers but uncertain ranges, since no one knows in advance what a given disaster will cost. And an important function of big, stable insurers is to accept some downside chance of exceeding the cap on total loss (a knapsack’s maximum weight capacity) within a set risk tolerance.

That combination, uncertain weights and a budget that can bend, puts the problem in territory that is classically challenging.

“Current classical approaches either rely on simulating multiple scenarios based on the uncertainties (accurate but computationally expensive) or by considering the worst-case range (safe but overly cautious),” said Vaibhaw Kumar, IBM quantum researcher and co-author of the paper.

In this paper, the joint team explored a novel quantum approach, applied to home insurance portfolios.

Like much of the most important quantum computing research today, their workflow combines quantum and classical computing, with each paradigm solving the parts of the problem to which it is best suited.

The quantum circuit, running on IBM Quantum Heron, generates a batch of candidate answers nudged toward combinations that are both valuable and within budget.

Then, a classical step cleans up the list. It repairs answers that break the budget and learns which houses tend to appear in good solutions, feeding that knowledge back to guide the next round of computation. This drives a virtuous cycle that tightens the results with each pass.

The team also found a way to improve the quantum results. Training the circuit directly on a large problem runs into a known wall where the learning signal fades out as the problem grows. Instead, they train on a small version and transfer what they learned to the larger one.

There’s reason to expect this workflow to get more powerful and efficient as hardware improves.

“If the noise on the hardware keeps getting lesser and lesser,” Kumar said, “the workload on the classical side will get smaller and smaller.”

Comparing quantum and classical methods

To evaluate the results, the team needed a yardstick. They used an exact solver that, given enough time, finds the provably best answer. They then compared the quantum-classical method to four common approximation methods that trade accuracy for speed: parallel tempering, tabu search, simulated annealing, and a genetic algorithm.

Each method, quantum and classical alike, was given 30 minutes per problem.

On problems up to 75 items, everything tied. The quantum-classical method and the approximation methods both matched the provably best answer. Even as the problems scaled, the quantum-classical method remained competitive with the strongest classical heuristic, matching it closely and marginally surpassing it under tight constraints—an encouraging sign for the harder, more constrained instances that matter most in practice.

This workflow isn’t yet ready for scales where exact solvers are too slow to be useful. But these early results show how quantum could yield business value for Allstate as hardware improves and problem sizes scale.

Why Allstate is investing in quantum computing

The point, Utke said, is to build a bridge from both ends: establish what quantum hardware can do now, identify what business problems it might solve as the machines improve, then see where the two meet.

Huls said that even with uncertainty in the timeline to business value, it’s better to be early than late. Quantum is developing quickly, and a company that has already built skills and found the right problems won’t be caught flat-footed.

Both Huls and Utke point to the same advice for other business leaders: start with a real problem. The technology is easier to learn, and progress is easier to measure, when it’s anchored to something the business actually cares about.

That principle shaped how Allstate began exploring the technology. Allstate is headquartered in Northbrook, in the Chicagoland region. IBM participates in the Chicago Quantum Exchange, which Utke said helped orient Allstate toward quantum computing.

The partnership began with IBM’s Quantum Accelerator Program for organizations new to the field, and grew into joint research over the past two years. Utke said he’s proud of the transparent nature of the joint research: public results, with code and data, that anyone can check and build on.

There’s still plenty more to learn about how quantum workflows will play into insurer decision making and portfolio construction, Huls said. But this work has offered valuable insights and learning opportunities that make Allstate prepared for that fast-approaching future.