Part 1 of 3 — The AI ROI Series
Recently, I presented an AI built Return on Investment (ROI) calculator at the IT Alliance Spring Collaborative in Indianapolis. The session was called, AI in Action. The feedback was very positive, so I turned some key points into a 3-part blog series on how to evaluate an AI project. I’ll start with the green flags, when AI is worth the conversation.
For context, I work with small and midsized companies that do not usually have the transaction volume that typically guarantees a positive ROI. Sometimes, I advise my clients not to proceed.
That probably sounds contradictory. It isn’t. The truth is that AI is genuinely transformative, but only for the right processes, at the right volume, under the right conditions. Getting those conditions wrong doesn’t just waste money. It poisons the well for the next AI project that would have been successful.
So, let’s talk about 4 key factors that make AI automation or agentic AI attractive.
High Volume
The single strongest predictor of AI ROI is volume. If a task happens hundreds of times a month, the fixed cost of the AI tool gets spread thin across thousands of instances. The math starts working in your favor almost immediately.
If that same task happens a dozen times a month, the fixed cost never amortizes. You’re paying enterprise-level tooling for a problem that a spreadsheet and a focused employee could solve for a fraction of the price.
The question isn’t just “can AI do this?” It’s “do we do this often enough for AI to make financial sense?”
Consistency
AI excels at structured, predictable, repeatable work. If every instance of a task looks roughly the same, with the same inputs, decision logic, and output format, AI will handle it reliably and, at scale.
The green flag here isn’t just consistency in easy cases. It’s consistency in the exceptions too. If your edge cases are well-defined they canbe handled by rules, AI can be trained to handle them. If your edge cases require genuine human judgment every time, that’s a different story — and we’ll cover it in Post 2.
Low Error Stakes
When AI makes a mistake on a low-stakes task, someone catches it, fixes it, and moves on. The cost of the error is minimal. The human oversight required is light.
That’s a green flag because it means your savings rate, the percentage of human time AI replaces, stays high. You’re not burning your efficiency gains on an elaborate review process designed to catch every AI mistake before it causes real damage.
Low error stakes don’t mean no oversight. They mean the oversight is proportionate, and the math still works.
Labor Cost
This one should be straightforward. If the fully loaded cost of the human time doing this task, including salary, benefits, payroll taxes, and overhead, significantly exceeds the monthly cost of the AI tool at a realistic savings rate, you have a viable business case.
The key word is realistic. Not the savings rate in the vendor’s case study. There will still be some labor; review time, exception handling, and the management overhead of running the tool. However, for large volume or time-consuming tasks, AI will easily create value.
The Green Flag Checklist
Before any AI conversation with a client, I run through these four questions:
- Volume: Does this happen enough times per month to justify fixed tooling costs?
- Consistency: Is the task structured and predictable enough for AI to handle reliably?
- Error tolerance: If AI gets it wrong, is the cost of that mistake manageable?
- Labor vs. tool cost: Does the labor savings at a realistic savings rate clearly beat what the tool costs?
Four greens and you have a conversation worth having. Before signing on the dotted line, Post 2, Red Flags is required reading first.
Next up — Post 2: Red Flags: When AI Will Cost You More Than It Saves