Why most AI starts go wrong
The first AI project sets the tone for everything after it. Pick something flashy that never pays back and the board loses patience. Pick something dull, frequent and measurable, and you have proof within a quarter.
The AI Opportunity Finder gives a leadership team a fast, fair way to compare ideas. You score the work your people already do, not the tools you have seen in a demo.
The five criteria
Score every process from 1 to 5 on each. A higher score always means a better place to start, so the maximum is 25.
| Criterion | What it asks | 5 means |
|---|---|---|
| Hours | How much team time goes on this each week? | Over 20 hours a week across the team |
| Repetitive | How similar is the task each time? | Same steps, same inputs, nearly every time |
| Data ready | Is the information digital and easy to reach? | In one or two systems, reasonably tidy |
| Low risk | How bad is it if the output is wrong? (Reverse scored) | A mistake is cheap, easy to spot and easy to fix |
| Impact | Does doing it better help customers or revenue? | Customers feel it fast or revenue moves |
The template lists 20 common processes across sales, marketing, customer service, finance, HR and operations, such as writing proposals, replying to common customer emails and taking meeting notes, with space for your own.
Reading your totals: 20 to 25 is a strong candidate. 14 to 19 is worth a look once the first wins are in. Under 14, park it for now. On a tie, pick the one with the lower risk. Early wins should be safe wins.
How to run it in 45 minutes
- Get the right people in the room. Two or three people who know the day-to-day work, not just the leadership view.
- Score quickly. Go with your gut. Do not debate a single score for more than a minute.
- Add up and sort. The highest totals rise to the top.
- Pick three and name owners. One named person each, and a first step you can finish within two weeks.
Red flags: don't start here
One red flag means think twice. Two means pick another process.
- It decides about people, such as hiring, pay or performance. Start with drafting help, never the decision itself.
- It uses sensitive personal data and you have not checked how UK GDPR, or your local data rules in the GCC, apply.
- No one owns it today. AI will not fix that.
- The process is already broken. Automating a messy process gives you a faster mess.
- A wrong answer goes straight to a customer with no human check in between.
- It happens a few times a year. Low volume rarely pays back.
Once you have your top three, plan the first one with the 90-Day AI Roadmap and check the numbers with the AI ROI Worksheet.
Questions
How long does the AI Opportunity Finder take?
About 45 minutes with two or three people who know the work. Score quickly and save the debate for the top three.
Do we need to know which AI tools to use first?
No. You score the work, not the tools. Choose tools once you know which process you are improving.
Why is risk reverse scored?
So risky processes sink down the list. Your first project should be a safe win that builds confidence.
Is it suitable for GCC businesses?
Yes. The processes and criteria are the same. Check your local data protection rules before personal data goes into any tool.