How to Select and Fund AI Use Cases

A live poll at the Industrial AI Summit 2026 showed 37% of attendees targeting plant floor maintenance and quality vision inspection as their first or next AI project, while 21% had not yet selected a use case. Benson Chan of Strategy of Things presented a seven-step framework for narrowing a list of 50 or more candidate AI use cases down to the 10 that deserve funding. The framework separates selection from funding: two decisions that require different criteria and different conversations inside the organization.

What Three Criteria Filter a Valid AI Use Case?

Before building a candidate list, every potential AI use case should pass through three decision gates. Failing any one of them eliminates the use case from consideration, regardless of how appealing the technology looks.

Gate 1: Business fit. Does this use case align with the business strategy? What return can the organization expect? The fastest way to test business fit: ask whether the business owner will fund the project if the pilot succeeds. If no one can answer that question, or if no owner exists who would write the check, the use case has no viable path from pilot to production. That single question eliminates use cases that lack organizational commitment before any money gets spent.

Gate 2: Feasibility. Can the organization actually execute? This means having the data in usable form, the infrastructure to support the solution, and the technical skills to build and maintain it. A use case with strong business value but no supporting data or no qualified staff is not feasible, and pursuing it wastes resources.

Gate 3: Scaling readiness. Most organizations defer this question until after the pilot, which is too late. If the pilot succeeds, what does it cost to scale? Does the infrastructure support production-level data volumes? Is the workforce willing to adopt the new system? Are there compliance or governance requirements that change at scale? Asking these questions before the pilot prevents the common outcome of discovering after a successful demo that production deployment costs multiples of the pilot budget.

Use cases that pass all three gates form the candidate list, while those that fail any gate get discarded entirely. Including marginal use cases on the list wastes evaluation effort.

How Does the Seven-Step Prioritization Framework Work?

Once the candidate list exists, the challenge shifts from selection to funding. A manufacturer with 50 valid use cases and budget for 10 needs a structured process for deciding which 10 to fund.

Steps one through four transform a scattered collection of ideas into a structured, comparable set. Step five catches hidden dependencies, because a computer vision use case may require a data collection infrastructure upgrade that must come first. Steps six and seven force the trade-off decisions and portfolio balancing that keep the final selection grounded.

The framework treats the use case list and the funding decision as separate problems. Many organizations conflate them, which leads to either pursuing too many use cases without adequate resources or selecting use cases based on enthusiasm rather than fit.

Why Should the Use Case Portfolio Mix Quick Wins with Strategic Bets?

An AI portfolio that contains only quick wins produces incremental improvements that competitors can easily replicate. A portfolio that contains only strategic or innovative bets is expensive, risky, and may not deliver any return in the near term.

The portfolio should work like a diversified investment. Some use cases are low-risk and fast to implement, building organizational confidence and demonstrating measurable value early. Others are strategic: harder to execute, longer in timeline, but necessary for competitive positioning because in a competitive industry, skipping strategic investments means a competitor will make them first.

The right mix depends on the industry, the business strategy, and the organization’s risk tolerance. A company in a rapidly evolving market needs at least one strategic or competitive use case in the portfolio to avoid being displaced. A company focused on operational efficiency may weight the portfolio more heavily toward quick wins that accumulate savings.

What makes the framework practical is that it forces each use case into a defined bucket, then compares them within that context. A quick win and an innovative project are not competing directly. They serve different portfolio functions and should be evaluated by different criteria.

What Did 37% of Manufacturers Choose as Their First AI Target?

The audience poll at the summit captured where manufacturers are concentrating their first or next AI investment:

The concentration in plant floor maintenance and quality reflects where operational data tends to be most available and where downtime and scrap costs create clear ROI pathways. These use cases typically pass all three decision gates more easily than cross-functional projects: the data exists in sensor feeds and inspection records, the business owner is the plant manager, and the scaling path is replicable across production lines.

The 21% who had not selected a use case represent the exact audience a prioritization framework is designed to serve. Without structure, those organizations risk either paralysis from too many options or impulsive selection driven by vendor marketing rather than business fit.

One consistent point across all panelists: start with the operational problem. The challenge most manufacturers face is throughput, quality, downtime, or cost, and the technology choices follow from that problem definition. AI solutions are abundant; the scarce resource is organizational focus on the right problem.

AI vendors are shifting their own pricing models toward value-based contracts, where the price reflects the business outcome rather than a flat SaaS subscription. That shift makes the ROI conversation more concrete for both sides but also raises the stakes of selecting the wrong use case. A value-based contract tied to a poorly chosen metric delivers neither value for the manufacturer nor revenue for the vendor.

Before starting any AI initiative, the panel’s collective advice: assess readiness across three dimensions. Infrastructure determines what is technically possible, operations determine what workflows exist to support AI, and organization and people determine whether the workforce will adopt the solution or resist it. That assessment, combined with the three decision gates and the seven-step prioritization framework, gives manufacturers a structured path from a list of possibilities to a funded portfolio of AI projects with clear ownership.