Here’s a scene I’ve watched play out in a dozen conference rooms. Leadership announces an AI initiative. Someone schedules a brainstorm. Ninety minutes later there’s a whiteboard with forty sticky notes ranging from “chatbot for HR questions” to “predict customer churn with machine learning,” and no way to choose among them. So the organization either chases the shiniest idea (usually the hardest one) or tries to do everything, and ships nothing.
The brainstorm isn’t the problem. The missing step is prioritization, and the tool I use for it in every engagement is deliberately old-school: a 2×2 matrix. Effort on one axis. Impact on the other.
First, generate better ideas
A prioritization matrix is only as good as what you feed it. Before scoring anything, source ideas from three places.
- 1
SOURCE 1
Start with the work people don't love.
Walk the workflow and ask each person: what takes you away from being strategic? What do you procrastinate on? The monthly report that takes three days to assemble. The proposal rebuilt from scratch every time instead of from a template. The meeting notes nobody writes up. Status updates, data re-entry between systems, first drafts of routine emails. Your team will hand you this list gladly, because you're offering to take away the part of the job they like least. Notice what that does for adoption: nobody resists a tool that eliminates their most-hated task.
- 2
SOURCE 2
Reclaim the tools you're already paying for.
Audit your software subscriptions. If a single team is paying for a project tool like Smartsheet or Monday, or a lightweight CRM, ask the question nobody asked five years ago: could we build this in-house now? With modern AI-assisted build tools, a working internal app customized to exactly your workflow is genuinely within reach, often for less than a year of license fees.
One firm caution: this applies to single-team tools with contained data and a handful of users. It does not apply to your ERP, payroll, company-wide CRM, or anything holding regulated data. Replace the team-level subscription; respect the enterprise platform.
- 3
SOURCE 3
Borrow from the proven starter categories.
You don't need to invent from scratch. Across the research on early AI wins, the same categories show up again and again: meeting capture, first drafts of routine documents, recurring reports and data digests, internal knowledge Q&A, inbox triage, and research synthesis. What they share: high frequency, low risk, easy to evaluate, and a human stays in the loop. They make excellent first moves.
More on Starter Categories
That third source is worth making concrete. The starter categories that show up again and again across the research are:
- Meeting capture and follow-through: summaries, action items, decisions, a searchable archive. Nobody has to be the note-taker anymore.
- First drafts of routine documents: reports, SOWs, job descriptions, internal updates, customer responses. AI writes the draft; humans make it right.
- Recurring reports and data digests: the weekly metrics email, the monthly board pack, trend summaries pulled from spreadsheets.
- Internal knowledge Q&A: an assistant grounded in your policies and process docs, so the same five questions stop landing in the same inbox.
- Email and inbox triage: categorize, summarize, draft replies for review. Benign failure modes, visible value within a week.
- Research and synthesis: competitor digests, market scans, literature reviews that used to eat an analyst’s afternoon.
Starter categories drawn from Grant Thornton, Tech River, and UC Today.
Then, score honestly
Take every candidate and place it on the matrix. Two axes, four quadrants.
The Effort / Impact Matrix
Score every candidate on two axes, then start where value is high and effort is low.
LOW EFFORT · HIGH IMPACT
Quick Wins
Visible value in weeks, with a human in the loop.
Do these first, two or three at a time.
HIGH EFFORT · HIGH IMPACT
Big Bets
Workflow redesigns and custom builds that change the business.
Sequence after a quick win has proven the team.
LOW EFFORT · LOW IMPACT
Fill-Ins
Nice to have, easy to do, not where progress lives.
Batch them when capacity allows.
HIGH EFFORT · LOW IMPACT
Money Pits
Impressive in a demo, far from a business outcome.
Decline politely. Revisit only if conditions change.
Quick Wins aren’t about the value of any single automation; they’re about momentum. A visible win in 30 days buys you the political capital for everything that follows.
Big Bets are the workflow redesigns and custom builds that actually change the business: plan them deliberately, resource them fully, and sequence them after a quick win has proven the team can deliver.
Fill-Ins are fine to batch when capacity allows; just don’t confuse activity here with progress.
Money Pits are where AI initiatives go to die, usually because the idea was impressive in a demo but far removed from a business outcome.
When you score effort, include more than build time: data readiness, integration complexity, security review, training burden, and the one everyone forgets, change effort. How many people have to work differently for this to pay off? A technically trivial tool that requires 500 people to change their habits is not low effort. When you score impact, anchor to numbers: hours saved per month times people affected, error reduction, cycle-time improvement, revenue influence. If you can’t articulate the impact in a sentence with a number in it, score it lower.
Download the template
To make this easy to run with your own team, we’ve packaged the whole exercise into a free worksheet: a use-case inventory, guided effort and impact scoring (including change effort and data readiness), automatic quadrant assignment, and a one-page summary you can put in front of leadership.
Where this fits in the bigger picture
Use-case prioritization is the heart of our Adoption by Design™ Envision workshop, the second of four sessions, where teams move from ideas to a sequenced roadmap with owners, metrics, and a 30-day quick win locked in. If your whiteboard already has forty sticky notes on it, that’s a great starting point.
The hardest part of AI isn’t the technology. It’s the transition, and the transition starts with choosing well.