How to choose your first quantum use case (and when to wait)

Your first quantum use case should be a problem you already solve with classical methods, that has a structure quantum algorithms are known to suit, that is small enough to run on today’s machines, and where a better answer would change a real decision. Very few problems pass all four tests. For many companies in 2026 the honest result of the exercise is a shortlist of two problems to watch and a decision to wait on the rest, and that is a good outcome.

This guide is for innovation, data and operations leads who have been asked to “find something to do with quantum”. It gives you a way to score candidates, a test that removes most of them, and the signs that tell you to hold off.

Which problems have a quantum shape?

Quantum algorithms help with three broad kinds of problem, and almost every serious quantum use case falls into one of them. If your candidate does not, stop there.

The first is simulation of nature: molecules, catalysts, batteries, materials, anything where the behavior of electrons decides the answer. This is the use case physicists expected from the start, because a quantum computer is itself a quantum system. It has the strongest theoretical case and the longest wait, since the chemistry that matters for industry needs error-corrected machines.

The second is combinatorial optimization: routing, scheduling, allocation, portfolio construction. It gets the most attention in business, and it is also where claims deserve the most suspicion. Classical solvers such as Gurobi, CPLEX or open-source heuristics are very good, and no quantum method has yet shown a clear, independently confirmed win over them on real industrial instances.

The third is sampling and some machine learning tasks, including Monte Carlo methods in finance. Results here are mixed and very sensitive to how the classical comparison is set up.

McKinsey’s Quantum Technology Monitor 2025 names chemicals, life sciences, finance and mobility as the sectors with the most to gain. That matches the three shapes: chemistry and pharma for simulation, logistics and mobility for optimization, finance for sampling and optimization.

A scoring framework for quantum use cases

Score each candidate from 0 to 2 on the six criteria below. The point is less the total than the conversation it forces: most candidates fail on one row, and it is usually the baseline.

A candidate needs at least a 1 on every row and a 2 on baseline and owner. A high score on shape with a 0 on baseline is the most common trap: it looks like a quantum problem, but you cannot tell whether quantum helped.

Why does the classical baseline decide everything?

Because every published claim of quantum advantage is a comparison, and the comparison is only as good as the classical side. If you do not know how well your current method performs, a quantum pilot can only produce a result nobody can interpret.

Two recent cases show how this plays out. In March 2025 D-Wave published a paper in Science claiming its annealer had beaten classical computers on a useful magnetic-materials simulation. Within days, researchers including a group at EPFL argued that classical tensor-network methods could handle the same problems, as HPCwire reported. In September 2025 HSBC and IBM reported up to a 34% improvement in predicting whether bond trades would fill, using IBM Heron processors on about 1.1 million trade requests. Scott Aaronson criticized the study on his blog the same day, noting that when the noiseless quantum computation was simulated classically, the advantage disappeared, which suggests the gain came from the hardware’s noise rather than from anything a classical computer could not do.

Neither case means the work was worthless. They show that the classical comparison is where results are won or lost. IBM seems to agree: in November 2025 it launched an open quantum advantage tracker with Algorithmiq, the Flatiron Institute and BlueQubit, where claims are tested against the best classical approaches anyone can submit. If you run a pilot, hold it to the same standard. Our review of whether quantum computing is useful yet covers more of these claims.

Should your first project be quantum-inspired?

Often, yes. Quantum-inspired methods are classical algorithms that borrow mathematics from quantum physics, mainly tensor networks, and run on ordinary servers or GPUs. You can put them into production this year, and they give your team practice with the same ideas a real quantum project will need.

The clearest commercial example is Multiverse Computing, the Spanish company that raised €189 million in June 2025 for CompactifAI, which uses tensor networks to compress large language models. That product has nothing to do with quantum hardware. It came from people who think in quantum terms.

A quantum-inspired first project also protects you from the baseline problem. If the quantum-inspired method beats your current solver, you have a result you can use now. If it doesn’t, you have a much stronger baseline for any quantum pilot later.

When is it better to wait?

Waiting is the right call more often than vendors will tell you. You should probably hold off on a quantum pilot if one of these is true:

• You cannot say how well your current method performs, in numbers.
• Your real bottleneck is data quality, integration or people, not the algorithm.
• The only candidates are large optimization problems that need thousands of variables to mean anything.
• No business owner has agreed to act on a positive result.
• The main reason for the project is a press release.

Waiting is not the same as ignoring. Pick two or three public milestones (a verified advantage result on IBM’s tracker, an error-corrected machine you can rent, a published result in your own domain) and agree in advance what you will do when each one happens. The quantum hardware roadmaps article lists the dated ones.

Security is the exception. Post-quantum cryptography is not a use case you can wait on, because it responds to a threat rather than an opportunity and it has regulatory dates attached. That work belongs in a separate plan.

What does a sensible first quantum project look like?

A good first project is small, time-boxed and designed so that a negative result still counts as a success. The goal is a benchmark and a trained team, not a production system.

1. Inventory. List ten to twenty candidate problems with the people who own them. Do not filter yet.
2. Score. Use the table above. Most candidates drop out here, usually on baseline or size.
3. Build the baseline. For the one or two survivors, measure the best classical method you can reasonably run, including a quantum-inspired one.
4. Define success in writing. Agree before anyone writes code what result would justify a next step, and what result would close the topic for a year.
5. Run the pilot. Use cloud access to real hardware and simulators, on a reduced version of the problem.
6. Decide. Continue, park with a dated trigger, or stop. All three are valid outcomes.

Quantum-South, a Uruguayan startup founded in 2019 at Universidad de Montevideo, is a useful regional reference. From May 2022 IAG Cargo trialled its air-cargo optimization work: a problem with a clear optimization shape, a measurable baseline and an owner who cared about the answer. That is the profile you are looking for, whatever the final result.

Where are the best first candidates likely to be?

In Latin America, the problems that score well tend to sit in a few places. Mining and energy companies have materials and chemistry questions (catalysts, batteries, corrosion) that are real simulation problems. Logistics operators have routing and scheduling problems with good data and strong existing solvers. Pharmaceutical and agrochemical firms have molecular questions, though usually through research partners.

Each of these has a long wait before quantum hardware beats classical methods on industrial sizes. The value of starting early is the benchmark, the team and a clear signal for when to invest more. The pages on logistics and transport and healthcare and pharma go into sector detail.

Getting from a list to a decision

If you already have a list of ideas and want to know which one, if any, deserves a pilot, that is what our quantum use case discovery work does: we score your candidates with your team, build the classical baseline, and tell you plainly when the answer is to wait. When a candidate survives, the next step is a hybrid quantum proof of concept with success criteria agreed up front.