Introduction
When AI arrives in an SME, ideas come quickly while the budget stays limited. Prioritizing AI use cases then becomes the step that decides what comes next. This article details the second step of our guide AI Strategy for SMEs: Key Steps and Priorities: how to rank your ideas with an impact × feasibility matrix.
The goal is to focus your resources on the projects that will pay off the most for a reasonable effort, then turn them into tools that actually run. That is what we build, business function by business function, on the diagonall AI systems page.
What is an impact × feasibility matrix?
The impact × feasibility matrix is a simple tool for assessing and ranking AI projects. Each project is placed according to its potential impact and how easy it is to implement. This gives four zones:
- High impact, high feasibility: the projects to launch first.
- High impact, low feasibility: the structural projects, to be prepared.
- Low impact, high feasibility: the small wins, to do if time allows.
- Low impact, low feasibility: to be set aside.
For an SME, this makes it possible to allocate an often tight budget where it matters.
How to assess potential impact
To assess the impact of an AI use case, look at several factors:
- Revenue: can AI improve existing sales or create new ones?
- Costs: which expenses can automation reduce?
- Efficiency: how many hours per week does the task cost today?
- Risk: which costly mistakes can the project prevent?
The most reliable approach is to start from your own measurements: time spent, volume handled per month, cost of an error. A figure from your own business is worth more than a market average.
Analyzing the feasibility of an AI project
Feasibility depends on several elements:
- Data: it must be accessible, relevant and of good quality.
- Skills: does your team know how to do it, or do you need training or support?
- Infrastructure: can your current tools connect to an AI solution?
- Acceptance: are the teams concerned ready to change the way they work?
An audit is often useful to answer these questions. See our article on assessing AI maturity.
The steps to prioritize your use cases
To prioritize effectively:
- List the opportunities: list the processes that could benefit from AI, together with the teams who deal with them.
- Score each idea: a score from 1 to 5 for impact and for feasibility.
- Place the ideas on the matrix: one sticky note per use case is enough.
- Choose: focus on two or three projects with high impact and high feasibility.
For what comes next, explore our AI roadmap for SMEs.
Example: a distribution SME
Let's take a fictional distribution SME that lists eight ideas. Three stand out in the top right of the matrix:
- Preparing replies to product availability requests: frequent requests, data already in the ERP, a team asking for it.
- Automatically extracting data from purchase orders received as PDFs: a time-consuming task with clear rules.
- Summarizing exchanges with major clients before a meeting: a clear time saving for sales reps.
Detailed stock forecasting, on the other hand, has high impact but low feasibility as long as the sales history remains incomplete. It goes into the "to be prepared" column, with a data cleaning workstream upstream.
Best practices and pitfalls to avoid
To succeed:
- Start small: test the idea with a POC before a broad rollout. See how to make an AI POC succeed.
- Track indicators: define before starting what will tell you whether the project worked.
- Don't overestimate impact: an optimistic estimate skews the whole matrix.
For the most common mistakes, read the 5 mistakes to avoid in an AI project.
Adopting an agile approach
The agile approach, well known in software, applies very well to prioritization. By breaking the project down into small steps, an SME tests its hypotheses quickly, adjusts and focuses its resources on what works. This reduces risk and brings technical and business teams closer together. The matrix is not fixed: a project judged difficult can become feasible after an initial data workstream.
To go further on implementation, read Scoping an AI POC: data, risks, schedule.
Integrating AI into company culture
Integrating AI is not just a technical matter. It requires a culture that understands and values what AI can bring: training employees on the tools, encouraging experimentation and collaboration between departments, and creating a climate of trust where people can test and make mistakes. Involving the teams in scoring the matrix is, incidentally, a good way to bring them on board.
Practical tips
- Involve all stakeholders from the start to align objectives.
- Make the matrix visible: a whiteboard or a shared document is enough.
- Reassess priorities every quarter.
- Train the teams to make adoption easier.
- Start with pilots before committing to a broad rollout.
Frequently asked questions
How do you determine the budget of an AI project?
You need to add up development, subscriptions, training, maintenance and internal time, then compare them with the expected gains. Our article on calculating the ROI of AI details the method.
What is the typical duration of an AI project for an SME?
A pilot generally takes from a few weeks to three months. A full rollout can stretch over one to two years depending on the number of processes involved.
What skills are needed?
A mix of business knowledge, development and project management. Many SMEs rely on an external partner for the technical part and keep business knowledge in-house.
How many use cases should you launch at the same time?
Two or three at most for an SME. Beyond that, the teams' attention gets scattered and no project goes the distance.
Conclusion
Prioritizing your use cases with the impact × feasibility matrix is the simplest way for an SME to choose its first projects without spreading itself too thin. By combining an honest assessment of impact and feasibility with short pilots, you reduce risk and get visible results.



