If you want to leverage AI potential , your company usually doesn't lack the technology. Many SMEs lack a clearly defined process bottleneck, a manageable AI use case, usable data quality, and a simple stop-or-go decision before investing money and energy in larger solutions.
I've seen this for years working with small and medium-sized enterprises: technology is rarely the primary bottleneck. Far more often, it's a lack of clarity in decision-making. Many teams are testing AI tools, but the potential of AI remains untapped in daily operations: emails pile up, knowledge is confined to individual minds, proposals take too long, data is scattered, and nobody knows what the most sensible first step is.
An AI pilot project is a small, measurable practical test that, before a larger investment, checks whether AI creates benefits in concrete everyday work.
That's precisely why AI needs less tool euphoria and more craftsmanship: strategic prioritization, human control, and a 30-day plan that turns an idea into a robust test.
Why companies are not using their AI potential
Many companies are now experimenting with generative AI. McKinsey & Company reported in 2024 that 65% of the organizations surveyed used generative AI regularly. At the same time, measurable business value is not automatically created by using tools, but primarily when companies rethink workflows and improve concrete processes.
This aligns with my experience of over 20 years in strategy, web development, branding, and digitalization: AI potential rarely fails due to a lack of access to ChatGPT, Claude, Microsoft Copilot, or other tools. Rather, AI potential is more often thwarted by vaguely defined goals.
Typical causes include:
- No clear process bottleneck: The company wants to use AI, but has not chosen a specific process that actually costs time, money or concentration.
- Initial projects that are too large: Instead of learning with a mini-prototype, the discussion immediately turns to complete systems, integrations, and automation.
- Poor data quality: Customer data, product information, offer logic, or internal documents are incomplete, outdated, or contradictory.
- Lack of responsibility: Nobody decides who checks results, which errors are acceptable, and where human control remains essential.
- No stop-or-go logic: A test continues even though it is unclear which metric will determine whether it continues, is adjusted, or is terminated.
MIT Sloan Management Review and BCG emphasized in 2019 that successful AI companies consider strategy, organization, and technology together. This is precisely where the opportunity lies for SMEs: You don't need a large AI department. You need a clear sequence.
Imagination means translating everyday problems into AI use cases.
By imagination, I don't mean creative daydreaming. Entrepreneurial imagination means: You recognize a recurring problem in everyday life and translate that problem into a testable AI use case.
An example from SME practice: A small service company loses time every week because email inquiries are worded differently, important information is missing, and offers have to be gathered manually.
The first AI use case wouldn't be: We build an AI agent for everything. The first AI use case would be: structuring incoming requests, highlighting missing information, and preparing an internal proposal template for human review.
This is how AI potential can be used in a practical way. AI does not replace the team's responsibility. AI relieves the team of repetitive sorting tasks, allowing people to make better decisions.
From lack of imagination to AI pilot in 30 days
A 30-day plan helps you transform the potential of AI from an abstract concept into a verifiable workflow. I like to use the following five steps as a sparring framework when an SME doesn't want to test yet another tool, but rather make a sound decision.
1. Choose a real bottleneck
Don't start with the question: Which AI is the best? Start with the question: Which process noticeably costs us time, quality, or concentration every week?
Good starting points include:
- Recurring customer inquiries that are always answered similarly,
- Manual quote preparation with many copy-and-paste steps,
- internal knowledge search in documents, emails or folders,
- Sorting leads, support cases, or applications,
- Summary of meetings, feedback, or project notes.
If you're still unsure which processes are suitable first, our article " Which processes should SMEs automate first?" will help you.
2. Outline three possible AI use cases.
A single bottleneck can have multiple AI use cases. That's precisely why a quick comparison is worthwhile before you get started.
Example customer inquiries:
- AI Use Case 1: Automatically group and sort inquiries by topic.
- AI Use Case 2: Identify missing information and prepare follow-up questions.
- AI Use Case 3: Draft responses according to tone, offer, and responsibility.
All three ideas can be worthwhile. But not all three belong in the initial test. For an AI pilot project, what matters is not what is theoretically possible. For an AI pilot project, what matters is what can be tested quickly.
3. Conduct a benefit-risk assessment
The cost-benefit analysis is the moment when strategy takes precedence over technology. You're not just assessing whether AI can do something; you're assessing whether its use is justifiable in the specific context.
Check at least these questions:
- Use: Does the AI use case regularly save time or improve the quality of a decision?
- Data: Is the required information available, up-to-date, and legally usable?
- Risk: Can incorrect results cause harm, for example through incorrect pricing, incorrect advice, or inappropriate communication?
- Control: Who checks the results before customers or employees are affected?
- Measurement: How can you tell after 30 days whether the test was worthwhile?
Data protection and regulation come into play as soon as personal data is processed or automated decisions can significantly affect people. In the EU, GDPR rules on profiling and automated individual decision-making are particularly relevant.
Additionally, the EU AI Act creates a risk-based regulatory framework for AI systems. For practical guidance, I recommend our overview of GDPR and AI for SMEs.
4. Build a mini-prototype
A mini-prototype is not finished software. A mini-prototype is a controlled test using real or realistically anonymized examples. The goal is not perfection. The goal is learning.
A simple mini-prototype might look like this:
- You collect 20 typical requests from the last few months.
- You anonymize personal data where necessary.
- You define a clear task for the AI: summarizing, classifying, preparing follow-up questions, or creating drafts.
- You have the result reviewed by a responsible person.
- You document where AI helps, where AI makes mistakes, and where rules are lacking.
If you want to differentiate more precisely when a prototype, a pilot project or already a product makes sense, you will find a clear decision logic in the article " Choosing the right AI prototype, pilot project or product".
5. Make a stop-or-go decision
After 30 days, you don't need a perfect AI strategy for the next five years. You need an honest stop-or-go decision.
The decision can take three forms:
- GB: The test saves time, improves quality, and remains controllable. The AI use case is being expanded.
- Adjust: The benefits are visible, but data, rules, or responsibilities need improvement.
- Stop: The benefit is too small, the risk too high, or the process is not yet mature enough.
A halt is not a failure. A halt is good corporate governance when a test reveals in time that another bottleneck is more important.
Checklist: When is an AI use case suitable for launch?
A good first AI use case for SMEs is small enough for testing and relevant enough for everyday use. This checklist will help you choose:
- Repeatability: This process occurs regularly, ideally weekly or daily.
- Data access: The necessary information is available, understandable, and not just stored in individual minds.
- Measurable benefits: You can observe time savings, quality improvements, faster response times, or less manual work.
- Low risk: An error does not cause serious damage because a human checks and approves.
- Clear human control: Responsible individuals remain involved in the process and make the crucial approvals.
If an AI use case fulfills these five points, it is usually better suited for starting out than a large, unclear automation plan.
Automation and innovation belong together
In many discussions, AI is viewed either as automation or as innovation. For SMEs, this distinction is often too simplistic. Good automation first and foremost creates relief. This relief then enables better communication, more precise decisions, and sometimes even new services.
For example, if a small team spends hours each week sorting through customer feedback, AI can first cluster topics and summarize sentiments. The immediate benefit is less manual work. The strategic benefit comes later: The team identifies recurring problems more frequently, improves offerings, and communicates more clearly.
This is where AI potential becomes practical: not as an end in itself, but as an enhancer of competence. A craft business, a law firm, a hotel, a planning office, or a consulting firm doesn't need AI gimmicks. An SME needs less chaos, better decisions, and processes that serve people.
The 30-day plan at a glance
If you want to start tomorrow, you can easily structure the 30-day plan:
- Days 1 to 5: Identify the process bottleneck, formulate a goal, and determine who is responsible.
- Days 6 to 10: Outline three AI use cases and prioritize them according to benefit, risk, data availability, and effort.
- Days 11 to 20: Test a mini-prototype with real or anonymized examples.
- Days 21 to 25: Review results, document error patterns, and improve rules.
- Days 26 to 30: Make a stop-or-go decision and plan the next sensible step.
This logic is deliberately simple. Small teams don't need months of preliminary research before they're allowed to learn. But small teams do need enough structure so that testing doesn't end in a chaotic mess of tools.
How Berger+Team proceeds
Berger+Team is a freelance collective from Bolzano and doesn't operate on a "tool first" logic. We start with the goal, the bottleneck, and the responsibility. Only then do we decide whether automation, AI strategy consulting, workflow integration, or a small prototype makes sense.
In our work with AI and digitalization, we combine strategic consulting, technical implementation, automation, and process optimization. For many small businesses, this combination is crucial: an AI idea must not only work technically, but also align with the brand, the team, the data, the customers, and the company's responsibilities.
If you are currently facing several options and don't know which AI application makes sense first, structured advice on digitalization and AI can be a better first step than the next tool test.
FAQ: AI potential, pilot project and use cases for SMEs
What is the potential of AI?
AI potential is the concrete possibility of using artificial intelligence to make an existing workflow faster, clearer, or of higher quality. For SMEs, AI potential arises primarily where repetitive tasks, scattered information, or slow decision-making slow down daily operations.
Why do AI projects fail in companies?
AI projects often fail not because of the technology itself, but due to a lack of strategy, poor data quality, unclear responsibilities, and overly high expectations. A small, measurable test reduces this risk because its benefits, limitations, and the need for human oversight become apparent early on.
How do I start an AI pilot project?
Start with a clear process bottleneck, choose a manageable AI use case, and define metrics for the stop-or-go decision in advance. Then test a mini-prototype with real or anonymized examples and have the results consistently reviewed by a human.
Which AI use cases are suitable for SMEs?
For SMEs, AI use cases with high repeatability and low risk are particularly suitable: inquiry sorting, summarizing, offer preparation, internal knowledge retrieval, text drafting, feedback evaluation, or simple process analyses. It is important that the benefits are visible in everyday practice and that no uncontrolled decisions are made.
How long does it take to develop a first AI prototype?
A first mini-prototype can often be created within a few days if the goal, data, and test examples are clear. However, for a robust AI pilot project, I recommend a 30-day plan, because this allows for an examination not only of technical feasibility but also of benefits, risks, and team acceptance.
Does a small company need its own AI department?
No, a small business doesn't need its own AI department to get started. An SME needs clear priorities, data privacy awareness, simple testing, and people who take responsibility for results, approvals, and improvements.
Your next step
If you want to leverage your AI potential, don't start with the grandest vision. Start with a bottleneck that puts a strain on your team every week.
I'd be happy to support you in prioritizing the first meaningful AI applications, setting up a solid AI pilot project, and preparing an honest stop-or-go decision. Strategically, practically, and with the goal that AI serves your company – not the other way around.
