How do you choose your first AI use cases?
Start with the work, not the tool. Collect 10 to 20 ideas where AI could help in daily work, rate each one by value, effort and risk, and pick three. Each of the three gets an accountable person and a metric. Everything else waits until the first three are running.
Why choosing is the most important step
What I hear in conversations with managing directors: AI initiatives rarely fail because of the technology. They fail because of the order. A tool gets bought first, and only then does someone look for what it is needed for. Or twenty small trials run at the same time and none of them gets finished. A good selection turns this around: first the work that is worth it, then the right tool.
The Shortlist Rule
No AI project without a shortlist: collect 10 to 20 ideas, sort them by effort, value and risk, pick three.
Step 1: Collect ideas
Do not ask “What can AI do?”. Ask “Where do we lose time every week?”. Good questions for your team:
- Which task repeats every week, almost always the same way?
- Where do customers wait longest for an answer?
- Which knowledge sits in a few heads, so everyone else has to ask?
- Which work keeps getting left behind because nobody likes doing it?
Bring in the people who actually do the work, not only the leadership team.
Step 2: Sort by value, effort and risk
Rate every idea as high, medium or low:
- Value: Does it noticeably save time, bring revenue or better quality for customers?
- Effort: Are the data and systems already there, or does something need to be built first? How much does the process change for the team?
- Risk: Does it involve customer data or personal data? What happens if the AI makes a mistake? Does it make decisions about people?
A tip: also check where AI already sits inside the software you buy. Many vendors have built it in long ago, without anyone in the company having decided on it.
Step 3: Pick three
Take the three ideas with the best balance of high value, reasonable effort and low risk. Not the most exciting idea, not the one that impresses most in the presentation.
Step 4: Set ownership and a metric
Each of the three gets one person who is accountable for the result, not a department. And a metric with today's value, so that after a few weeks you can see whether it works.
A simple template
| Idea | Value | Effort | Risk | Owner | Metric |
|---|---|---|---|---|---|
| Answer product questions in inside sales with an AI assistant | high | medium | low | Head of inside sales | Time to answer |
| Transfer meeting notes into the CRM | medium | low | medium | Head of sales | Share of complete entries |
| … |
The most common selection mistakes
- Starting with the tool: “We have Copilot now, what do we do with it?” Better: first the work, then the tool.
- Too many at once: twenty trials produce twenty half results.
- Nobody accountable: without a named person, every idea stays an experiment.
- No metric: what you do not measure beforehand, you cannot prove afterwards.
Frequently asked questions
How many ideas should we collect?
10 to 20. Fewer, and you choose from too little. More, and sorting turns into busywork.
Who should take part in the selection?
The person who decides, the people who do the work every day, and someone who knows the data and systems. Leadership makes the decision, not IT.
What do we do with the ideas that are not chosen?
They stay on the list. Once the first three are running, the list comes back to the table, together with what you have learned in the meantime.
What if all the good ideas involve customer data?
Then start with an idea that stays internal, for example finding knowledge in the company faster. For ideas involving customer data, clarify with your data protection officer beforehand which systems are suitable. This is guidance, not legal advice.
More answers
- Where does AI deliver measurable results in sales first, in mid-sized companies?
- Why do AI pilots get stuck, and how does AI get into daily business?
- Where does AI pay off in marketing for mid-sized companies, and where not?
About the authorSelina Gaertner, AI for Leadership, Sales and Marketing. TÜV Rheinland certified AI Strategy Consultant, more than ten years in sales and marketing in medical technology, co-author of the Amazon bestseller “Fokus Mensch”.
Your next step: the AI Value Sprint
In 90 minutes you and up to eight people from your team build your shortlist and your first three AI levers.