Pick your first AI project with math, not opinions

Welcome to the first issue of Systems Thinker. One practical way to make AI actually work inside a business. No hype, no tool-of-the-week.

Here’s how most companies pick their first AI project: someone senior sees a demo, the room gets excited, and three months later there’s a half-built chatbot nobody asked for. The problem was never the technology. The problem is that the project was chosen by enthusiasm instead of evidence.

The fix is boring, which is why it works: score your candidates like investments.

The AI Use Case Scorecard, in four steps

1. List 5–10 candidate workflows, real ones. “Draft the weekly client report”, not “marketing”.
2. Score each 1–5 against ten criteria: impact, frequency, process clarity, data availability, feasibility, risk, speed to validate, strategic value, review requirement, integration complexity.
3. Weight what compounds: business impact and workflow frequency count double.
4. Rank the totals and commit to the top one. One initiative, one quarter. The rest go on a written roadmap.

Two rules keep the scoring honest. First, use anchors: know what a 1 looks like and what a 5 looks like for every row before you score anything. Second, gut-feel 3s are usually a 2. When you’re unsure, round down.

The maximum weighted score is 60. Anything above 45 deserves a serious look. Anything below 35 is a quarter of wasted work you just avoided.

The full scorecard is free on the site: the AI Use Case Scorecard, with all ten criteria and their 1-vs-5 anchors, ready to copy into a doc or spreadsheet.

If you run your list through it this week, hit reply and tell me what came out on top. I read every reply.