How to Score 10 AI Use Cases in One Afternoon

Every organization I have worked with has the same problem. Not a shortage of AI ideas. A surplus. The VP of Operations wants document processing. The Head of Compliance wants risk scoring. The CTO wants a customer-facing chatbot. The CEO read an article about generative AI and now wants 'something with that.' You have 10 ideas and budget for one. The question is not 'which one sounds exciting?' The question is 'which one has the conditions for success?'

Why Gut Instinct Picks the Wrong Project

In my experience, the AI project that gets funded is usually the one with the loudest sponsor, not the best odds. The SVP who pounds the table in the steering committee meeting wins. The director with the quiet, well-researched proposal loses. Then the loud project fails because nobody asked whether the data was ready, whether the team had bandwidth, or whether the technology was proven.

I have seen this pattern play out at three different banks. The politically powerful project gets approved. It stalls at month four because the data is scattered across seven legacy systems. Meanwhile, the boring document processing project that scored highest on every objective measure sits on the watch list, waiting for a champion who can yell louder.

A scoring rubric fixes this. It does not eliminate politics entirely. Nothing does. But it forces every project through the same five dimensions. The math becomes the argument. And in my experience, even the loudest SVP has trouble arguing with a project that scored 14 out of 25 when the alternative scored 22.

The 5 Dimensions That Predict AI Project Success

After running this exercise across dozens of AI initiatives, I have found that five dimensions reliably predict whether a project will succeed or stall. They are not equally obvious. Everyone thinks about technical feasibility. Almost nobody thinks about organizational readiness. Here are all five.

Dimension 1: Data Readiness. This is the dimension that kills more AI projects than any other. A 5 means your data is structured, in a single system, regularly updated, and well-documented. A 1 means the data does not exist yet. Most enterprise projects land at a 2 or 3 — data exists but it is scattered, inconsistent, or locked in legacy systems that require months of extraction work.

The honest assessment is brutal. I have sat in rooms where the project sponsor insists the data is 'mostly clean' and then the data engineer quietly says it will take four months to build a reliable pipeline. Trust the data engineer. They are the only person in the room without a political reason to be optimistic.

Dimension 2: Impact Magnitude. How much is this worth in dollars? A 5 means $1M+ annual impact or a strategic competitive advantage. A 1 means under $50K annual impact. Be specific. 'This will save time' is not impact magnitude. 'This will reduce false positive alerts by 60%, saving 2,400 analyst hours annually at a loaded cost of $85 per hour, which is $204K per year' is impact magnitude.

Dimension 3: Organizational Readiness. This is the dimension nobody wants to score honestly. A 5 means you have an executive sponsor, an eager team, clear ownership, and available bandwidth. A 1 means active resistance, competing priorities, or organizational dysfunction. Most teams rate themselves a 4 when they are really a 2. Ask this question: 'If I pulled two people off their current work to staff this project, whose projects would suffer and would those leaders support it?' If you cannot answer that cleanly, you are not a 4.

Dimension 4: Technical Feasibility. A 5 is an off-the-shelf product with standard integration and a proven playbook. A 1 is cutting-edge technology with no proven solutions. Most enterprises should be targeting projects that score 4 or 5 here. The time for research-grade AI experiments is not when you are trying to prove ROI to the board.

Dimension 5: Strategic Alignment. This one is counterintuitive. A 5 means internal-only, no customer impact, easily reversed, and no regulatory exposure. A 1 means direct customer impact, high regulatory risk, and reputational exposure. Higher-risk projects are not bad projects — they just need more governance, more budget, and more time. For your first AI win, you want a project that scores 4 or 5 here. Low risk. High visibility. Easy to course-correct if something goes wrong.

Running the Exercise in 90 Minutes

Here is how I run the scoring session. Book 90 minutes with your AI steering committee or leadership team. You need the people who understand the data, the business impact, the team capacity, and the technology options. That is usually 4-6 people.

Minutes 1-10: Frame the exercise. List your top 5 AI project candidates on a whiteboard or shared doc. No more than 5. If you have 10, pre-filter down to 5 using a quick impact-vs-feasibility gut check. The scoring rubric works best with 3-5 candidates.

Minutes 10-70: Score each project. Take each dimension one at a time across all projects. Do not score one project completely before moving to the next — that creates anchoring bias. Instead, score all 5 projects on Data Readiness, then all 5 on Impact Magnitude, and so on. Each dimension should take about 12 minutes of discussion.

Minutes 70-80: Total and rank. Add up the scores. The ranking usually generates one or two surprises. The project everyone assumed would win often does not. The boring, well-positioned project that nobody was championing floats to the top. That is the rubric doing its job.

Minutes 80-90: Make the call. Projects scoring 20-25 are your 'start now' candidates. 16-19 should be scoped for next quarter. Below 16 goes on the watch list. If two projects are within 2 points of each other, use the Data Readiness score as the tiebreaker. Data readiness is the single best predictor of first-year success.

What Happens After You Score

The rubric gives you a ranking. It does not give you a project plan. For the winning project, you need four more things: a cost estimate across five layers, an ROI projection, a 90-day implementation timeline, and a one-page brief for executive approval.

These are not separate initiatives. They are a single afternoon of work if you have the right templates. Score in the morning. Budget in early afternoon. Draft the brief before end of day. Walk into Monday's meeting with a funded project, not a vague pitch.

The difference between a director who says 'we should invest in AI' and a director who says 'I have scored 5 candidates, picked the one with the best conditions for success, built a 3-scenario budget, and here is the one-page brief for your approval' is the difference between being seen as strategic and being seen as someone who gets things funded.

Actionable Takeaway