When most people hear "ROI," they picture spreadsheets and finance teams. But if you're just trying to use AI to save time, work smarter, or stop doing the boring parts of your job — ROI is a much simpler question:
"Is the time and energy I put into this going to come back to me — and then some?"
The answer depends less on how clever your AI build is, and more on where you are in your AI journey. That's why we think about adoption in three stages:
Crawl → Walk → Run.
Start at Crawl. Always.
The biggest mistake people make with AI is trying to Run on day one — building some autonomous agent for a niche problem before they've even gotten comfortable with the basics. That's the fastest way to get frustrated and conclude "AI doesn't work for me."
Crawl isn't a warm-up you skip. Crawl is where you build the muscle memory and foundation that makes everything after it possible.
🐢 CRAWL — Build the foundation
Use cases that have immediate time-to-value and work for most of what you do:
Web research
Working with files (summarize, extract, rewrite)
Image generation
Brainstorming
Basic CRM integration / lookups
Enhancing existing processes you already do
ROI logic: Low effort, low-to-medium impact per task — but the returns show up immediately, and more importantly, you're learning. You're figuring out what AI is actually good at for your work, what prompts get you what you want, and where the real bottlenecks in your day live. You can't identify a good Walk or Run use case until you've spent real time in Crawl.
Don't rush through this stage. The instinct to jump ahead to "the cool stuff" is exactly what kills ROI later.
🚶 WALK — Stitching it together for greater ROI
Once the basics are second nature, you start chaining things together.
Use cases that are similar to Crawl but require customization and a bit more hand-holding:
Using multiple tools in one flow
Early agentic work (AI taking a few steps on its own)
Workflows + triggers ("when X happens, do Y")
Greater ROI — the same setup pays you back over and over
ROI logic: A bit more effort to configure, but now one build keeps working for you every day or every week. This is where AI stops feeling like a novelty and starts feeling like leverage. The reason you can do this well is because you crawled first — you already know which repetitive task is worth automating.
🏃 RUN — AI-first processes and autonomous agents
Use cases unique to a specific person or business unit — high impact, but they need specialty implementation:
Autonomous agents that handle a job end-to-end
Brand-new processes built AI-first (not just speeding up an old one)
ROI logic: High effort, high impact. Reserve Run for the handful of problems where it really matters. And you'll know which ones those are — because Crawl and Walk taught you.
The Impact vs. Effort Matrix
When you're staring at a list of "things I could build with AI," sort them into four boxes:
| Low Impact | High Impact |
Low Effort | 🐢 CRAWL — start here, build the habit | 🚶 WALK — your sweet spot for ROI |
High Effort | ❌ AVOID — not worth it | 🏃 RUN — plan for it later |
Three rules:
Start in Crawl, not Walk. Yes, Walk has better per-build ROI — but without Crawl, you don't know which Walk use cases to pick.
Avoid the bottom-left at all costs. A clever AI build for a problem nobody has is the #1 way people decide "AI didn't work for us."
Don't camp at Crawl forever, either. The foundation only pays off when you use it to move up.
A Simple Self-Check Before You Build
Before you spend an hour (or a week) on an AI build, ask:
How often do I do this? Daily > weekly > monthly. Repetition is where ROI compounds.
How long does it take me today? If it's 2 minutes once a quarter, skip it.
Is this a Crawl task I haven't mastered yet? If so, do that first — don't skip ahead.
Can I reuse the setup? A one-off prompt dies after one use. A saved workflow, app, or agent keeps paying you back.
Who else could use this? A workflow that helps you and three teammates is 4x the return for the same build time.
The Mindset Shift
The average AI user measures ROI by asking "did this one prompt save me 5 minutes?" That framing will always undersell AI.
The better question:
"Am I building the muscle memory today that lets me unlock 10x bigger wins next month?"
That's the Crawl → Walk → Run payoff:
Crawl builds the foundation and the habit.
Walk turns that foundation into real, repeatable leverage.
Run turns leverage into capabilities that didn't exist before.
Skip Crawl and you'll spend weeks building a Run use case that solves the wrong problem. Stay in Crawl forever and you'll never see the compounding returns. Move through the stages in order, and ROI stops being something you have to justify — it shows up on its own.