An airline trains its pilots in a simulator so they can eventually transport passengers. The simulator helps the pilots develop skills and prepare for conditions they will encounter in flight. Its value comes from the operating capability it helps create.
Quantum computing pilots should serve a similar purpose. They help organizations understand the technology, explore applications, and develop skills. But a promising quantum pilot should lead to a different conversation: Which application are we preparing to put into use, and what will it take to get there?
Another experiment may answer another technical question. Eventually, however, the organization needs to select a real problem and commit to developing a solution. That work can begin before the hardware needed for production is available.
WHAT CHANGES AFTER THE PILOT
A pilot usually asks a bounded question. Can we express this problem in a form a quantum computer can process? Does a particular algorithm look promising? What computing resources might a useful version require?
Simplifications are often appropriate. A small dataset or reduced model can help investigate one question without taking on every complication of the eventual application.
A real project takes responsibility for those complications. It has a business owner, a defined outcome, and people assigned to deliver it. The team must establish how its proposed solution compares with the best available classical computing approach and fits into an existing workflow. Consider a hypothetical materials manufacturer whose pilot explores whether a quantum algorithm can calculate a property of a small molecule. Encouraging results give the team confidence to continue.
The next step is to identify a calculation that would affect an actual materials decision. Which property matters? How accurately must it be calculated? Would a better answer help researchers select candidates for expensive laboratory testing?
Those questions connect the computation to a business outcome. They also expose requirements the pilot may never have addressed, including validation procedures, operating costs, and the time available to produce an answer.
A LESSON FROM AI PILOTS
Artificial intelligence offers a useful warning. Many organizations proved that a model could perform a task yet struggled to move beyond the pilot because they had not assigned an accountable business owner, redesigned the surrounding workflow, prepared production data, or agreed on measures of value.
Quantum teams can avoid the same pattern. The goal is not simply to demonstrate that an algorithm works. It is to determine who will use the result, what decision it will improve, how it will be validated, and what conditions must be met to deploy it.
WHY PREPARATION BELONGS ON TODAY’S SCHEDULE
Hardware roadmaps provide a reason to begin now. My company is working with Amazon Web Services on the first fault-tolerant quantum computer (FTQC) on the Amazon cloud. Fault tolerance means correcting errors during a computation so the machine can perform longer, more demanding calculations reliably.
Roadmap dates can change. Nor does a fault-tolerant machine make every proposed application practical. Each project has requirements for computing capacity, accuracy, and cost.
Still, these milestones give enterprises something concrete to plan against. A team can estimate what its application needs, compare that with expected capabilities, and identify work it can undertake before the hardware arrives.
Preparing data, selecting algorithms, and establishing validation methods take time. So does bringing business and computational specialists together. Waiting for useful hardware before beginning adds an organizational delay to the technical timeline. The computer may be ready while the company is still deciding what to ask it.
DON’T MAKE THE TRANSITION ALONE
An enterprise needs to understand its problem, the value of a useful answer, and the conditions under which that answer can be trusted. A quantum vendor brings complementary knowledge of its particular computer—which approaches suit it, what resources an application will consume, and where its limitations affect the plan.
Two approaches to the same calculation may place very different demands on a machine. A vendor can help evaluate those choices, distinguish what can be tested today from what requires future capabilities, and revise estimates as the application develops.
An effective partner will explain the limits as clearly as the opportunities. That includes identifying cases where a classical method remains preferable or an application requires capabilities beyond the current roadmap.
For our hypothetical materials manufacturer, the joint work might begin with the required scientific accuracy. The teams would select a computational approach, estimate hardware needs, and agree on validation. The manufacturer would determine whether the output could improve a laboratory decision. The vendor would help establish a credible route to producing it.
The result should be a shared development plan, with responsibilities and evidence required at each milestone.
BUILD TOWARD AN OPERATING CAPABILITY
Moving toward production involves progressively more demanding tests. A simplified problem becomes a representative one. Results are compared with credible alternatives. The team evaluates repeatability and total workflow cost, including the classical computing and human work surrounding the quantum calculation.
Intended users should participate before deployment. An impressive answer has limited value if researchers cannot interpret it or it arrives too late to influence their next experiment.
Production also needs a definition appropriate to the application. For a materials team, it might mean a validated calculation used periodically to select experiments. The capability must become a dependable part of the work.
Funding can grow as evidence improves. Some projects will need a different algorithm, a narrower scope, or a later hardware milestone. A shared plan should accommodate those adjustments while keeping the intended outcome clear.
The hardware roadmap is only part of the schedule. Enterprises need a roadmap for their applications, workflows, and people, developed alongside partners who understand the machines they intend to use.
A promising pilot provides a place to start. The next step is to select a real problem, agree on what success requires, and begin building toward production together.
Yuval Boger is the chief commercial officer of QuEra.