2 Cents from six50: Why 70% of “AI Projects” Never Pay Off

TL;DR

Most AI initiatives fail because they automate activity, not outcomes. Projects tied to a measurable metric (hours, dollars, or error rate) are 3× more likely to deliver ROI within 90 days.

The Take

AI isn’t failing businesses.
Bad project selection is.

Too many teams start with:
“What can AI do for us?”

Instead of:
“Where are we losing time or money today?”

When AI is treated like a feature instead of a fix, it becomes shelfware.

The Numbers

What we see most often:

• 70% of AI pilots stall after the proof-of-concept
• 2–3 tools added with no owner or success metric
• 0 defined ROI beyond “efficiency”
• 90+ days before leadership questions the investment

AI doesn’t fail fast. It fails quietly.

What This Means for Operators

If you can’t answer:

• What gets better?
• By how much?
• By when?

…then the project is already at risk.

The best AI wins start small, focused, and boring.

Our 2 Cents

Before approving any AI initiative:

• Tie it to one business metric
• Define success in 30, 60, or 90 days
• Assign a single owner

No metric = no momentum.

Figures in this article are illustrative estimates for discussion, not guarantees or audited results. Outcomes vary by business. This article is general information, not accounting, tax, or legal advice.