AI in action: What really works for small and medium-sized businesses and what doesn't

1. The reality for small and medium-sized businesses: between skepticism and buzzword bingo
2. What works (and what doesn't): Three observations from practice
3. Use case: Automating quotation processes in technical sales
4. What matters: data, clarity, integration
5. Conclusion: Small steps, clear goals—and no AI drama

1. The reality in small and medium-sized businesses: Between skepticism and buzzword bingo

Many decision-makers have long had AI on their radar, but the technical possibilities rarely lead to a concrete use case. Why is that?

• Because many offers are too generic or too technical
• Because benefits and costs are not made tangible
• Because AI projects are often treated as innovation showcases rather than genuine business solutions.

👉 Tip: When you receive your next AI suggestion, ask directly: What specific problem does this solve—and what is the business case? This separates show from substance.

2. What works (and what doesn't): Three lessons learned from practical experience

❌ What doesn't work:

• "AI implementation projects" without a clear use case
• Isolated pilots in IT, without involvement of the specialist departments
• Tools that promise more than they deliver in everyday use and within existing structures

✅ What works:

• Specific problems arising from operational business (e.g., quotation processes, lead prioritization, after-sales)
• Existing data + clear process + tangible goal
• Involvement of the people who will be working with it, and no black box AI

👉 Best case: A company tests an AI assistant in just one sales team, focusing on prequalifying leads. The model is quickly improved thanks to real feedback. After six weeks, it is scaled up.

3. Use case: Automating quotation processes in technical sales

A mechanical engineer receives daily inquiries about standard products with individual requirements. Sales and engineering prepare quotes manually, often under time pressure, with queries and media breaks.

The process without AI:

• Incoming requests (e.g., PDFs, emails) are automatically read
• The AI recognizes technical parameters and compares them with existing orders.
• Suggestions for configuration, price, and delivery time are generated.
• Sales reviews and finalizes the offer.

The result:

• Offer time drops from 3 days to less than 24 hours
• Standard inquiries are automated, special cases are handled in a more focused manner
• Sales gains time for consulting and closing deals

👉Tip: You don't need a huge tool. Often, a targeted module or low-code solution that complements existing processes—and really uses your data—is enough.

4. What matters: data, focus, integration

AI is only as good as its setup. This means:

• Data access: Historical offers, product data, CRM—without a clean database, there is no benefit.
• Clear focus: Not: "We do something with AI." But rather: "We want to make XY more efficient."
• Process integration: AI must fit into everyday life. If employees avoid using it, it's a waste of budget.

What is often overlooked is that it is not about automating as much as possible, but rather automating the right things. The best projects start with a clear question: Where are we currently wasting too much time on recurring tasks with clear rules?

👉 Best case: A sales team defines the three most common types of inquiries—and starts right there. After the test run, the system is refined and expanded. This way, the solution grows with demand.

5. Conclusion: Small steps, clear goals, and concrete use cases

AI can bring real added value, especially in small and medium-sized businesses, where capacities are limited and skilled workers are in short supply. But only if you...

• Start with real problems, not with visions.
• calculate specifically, not just present
• involve your teams, don't overwhelm them

You don't need an AI strategy. You need a first project that works.

👉Tip: Take an hour with your sales or service team. Ask: Which task is holding us back—even though it's actually standardized?
That's where your AI project begins. Not buzzwords, but concrete use cases.

‍