AI & Automation

How to Prioritise AI Use Cases: A Scoring Method That Fits on One Page

November 11, 2026 · 5 min read · Etzal Group Team

Once an organisation decides to invest in AI, ideas arrive from every direction. Sales wants lead scoring. Finance wants invoice extraction. Operations wants demand forecasts. Customer service wants an assistant. All of them sound reasonable, and the budget covers two or three. A transparent way to prioritise AI use cases prevents the choice from going to whoever argues longest.

Collect the ideas properly

Ask each department for candidates using the same short form:

  • The task today. What is done, by whom, and how often.
  • The problem. Slow, error-prone, expensive or missed entirely.
  • The idea. What the AI would do.
  • The measure. How improvement would show up.
  • The data involved and where it lives.

A common form makes ideas comparable. It also filters out suggestions that are a technology looking for a purpose.

Score on three dimensions

Rate each idea from 1 to 5 on the following. Agree in advance what each number means.

Value

  • Volume and time. How much work is involved?
  • Financial effect. Cost saved, revenue protected or gained.
  • Quality and speed. Fewer errors, faster response to customers.
  • Strategic fit. Does it support what the company is trying to become?

A 5 means a measurable effect on a figure the management team watches. A 1 means a convenience for a few people.

Feasibility

  • Data. Does it exist, is it accessible, is it good enough?
  • Process clarity. Can the task and its rules be described?
  • Integration. Do the systems involved have a way in?
  • Technology maturity. Is this something current models do reliably?
  • People. Is there an owner with time, and will users accept it?

A 5 means the data is ready, the process is clear and similar projects exist. A 1 means a data project and a process redesign are needed first.

Risk

  • Cost of an error. What happens when the AI is wrong?
  • Reversibility. Can a mistake be undone?
  • Privacy and regulation. Does it touch personal or regulated data?
  • Exposure. Is the output seen by customers or only by staff?

Score risk so that 5 is low risk and 1 is high, so that higher is always better across all three.

Run the scoring session

  1. Bring together process owners, IT or data, and finance or risk.
  2. Present each idea in two minutes.
  3. Score individually first, then discuss the differences. Disagreement usually reveals a missing fact.
  4. Record the reason for each score in a sentence.

Two hours is enough for twenty ideas. The written reasons are as useful as the numbers.

Read the result

Plot value against feasibility, and mark risk by colour.

  • High value, high feasibility. Do these first.
  • High value, low feasibility. Worth preparing for. Identify what blocks them.
  • Low value, high feasibility. Easy, and only worth doing if they build skills or goodwill cheaply.
  • Low value, low feasibility. Drop them, and say so plainly.

Then apply two checks. Anything with a low risk score needs human review designed in, or should wait. And the selected set should not all depend on the same few people.

Choose a balanced first set

A good first selection usually includes:

  • One quick win. High feasibility, low risk, visible within weeks. Internal drafting or summarising tasks are typical.
  • One project with real business value. It takes longer and justifies the programme.
  • One enabler. Cleaning a data set or building a knowledge base that several later ideas require.

This mix produces early evidence, a meaningful result and a foundation.

Weighting

If value matters more to you than ease, weight it. A common approach doubles the value score. Decide the weights before scoring, not after seeing which idea you hoped would win.

Keep the list alive

The ranking is a snapshot. Review it every six months:

  • Ideas blocked by data may now be ready.
  • Model capabilities change what is feasible.
  • Finished projects teach you how optimistic your scores were.

Compare predicted and actual value for completed projects. It improves the next round of scoring.

Common mistakes

  • Scoring by enthusiasm. A persuasive sponsor is not a measure of value.
  • Ignoring feasibility. The most valuable idea on paper stalls for a year on data.
  • Ignoring risk. A customer-facing agent launched without review damages trust in the whole programme.
  • Too many at once. Three done well beat ten started.
  • No owner. An idea without a named business owner should not be selected, whatever its score.

What you end with

One page: a ranked table with scores and reasons, a chart, the three chosen projects with owners and measures, and the blocked ideas with what would unblock them. People who did not get their project can see why and what would change it.

Running the collection and scoring workshops and producing this page is a standard part of our AI Consulting engagements. If you already have a list of ideas and need help ranking it, send it to us.

Summary

Collect ideas in a common form. Score each on value, feasibility and risk with the people who own the work. Start with a quick win, a project with real value and an enabler. Write down the reasons, name the owners, and revisit the ranking twice a year.

Frequently asked questions

How many AI use cases should we evaluate?

Collect widely, then score fifteen to thirty. Fewer than ten usually means only the loudest ideas were heard. More than forty makes the scoring session too long to do carefully. Group near-duplicates before scoring.

Who should take part in scoring AI use cases?

The people who own the processes, someone from IT or data who knows what the systems can provide, and someone from finance or risk. Scoring by a single function gives a biased ranking, and scores agreed together are easier to act on.

What if a high-value idea scores low on feasibility?

Break it down. Ask what makes it hard, such as missing data or an absent integration, and treat that as a project in its own right. Often a smaller version of the idea is feasible now and builds toward the larger one.

Author

Etzal Group Team

Etzal Group Team, helping businesses grow with technology. Contact us.

Related service

WhatsApp AI Chatbot

We build, train, and run a WhatsApp AI agent that answers customers 24/7, qualifies leads, and follows up.