Why AI projects miss ROI in mid-sized companies
Most mid-sized companies are using AI, but few can show a return, because the tool was added to a workflow that was never fixed. The return comes from redesigning the work and naming an owner, not from buying more software.
Adoption is not the problem. Readiness and redesign are.
91% use it. One in four have integrated it.
91% of middle market firms surveyed say they use generative AI, yet only one in four of those users say it is fully integrated into core operations. RSM 2025 Middle Market AI Survey (966 firms, Feb–Mar 2025).
61% of U.S. firms were not fully prepared.
The top reasons were lack of in-house expertise (39%), no clear AI strategy (34%) and data quality (32%). Same RSM survey.
About 6% see a real financial return.
About 6% of organizations qualify as AI high performers, with at least 5% of EBIT from AI. Roughly 75% of them fundamentally redesigned workflows, versus about 25% of everyone else. McKinsey, The State of AI (1,719 respondents, May–Jun 2026).
Four reasons the return never shows up.
1. The workflow underneath is broken
AI speeds up a bad process and makes its problems arrive faster.
2. No one owns the number
Without a named owner and a baseline, "time saved" never becomes margin or cash.
3. The constraint is elsewhere
Improving a step that is not the bottleneck does not move throughput.
4. The data and skills are not ready
Data quality was the most common rollout problem (41%) in the RSM survey.
Before you fund the next initiative.
- Name the one business result the initiative must move (margin, cycle time, cash conversion).
- Measure the current baseline for that result.
- Find the constraint in the workflow and confirm the initiative touches it.
- Assign one owner and a 90-day check on the baseline.
Fix the workflow first, then apply AI where it pays.
That sequence is the core of Workflow Performance. Anyone can buy the same AI tools. A workflow that runs, with a named owner, is the moat that makes you defensible.
Common questions.
Why do most AI projects fail to deliver ROI?
Usually the process, ownership or data was not ready. Tools are easy to buy; redesigned work and clear accountability are what produce a return.
How long until AI shows a return in a mid-sized company?
There is no reliable industry figure. A defined baseline and a 90-day checkpoint on one workflow will tell you faster than a company-wide rollout.
Should we pause AI spending?
Not necessarily. Pause new initiatives until the current ones have a named owner, a baseline and a measured result.
Do we need a consultant to do this?
70% of generative AI users in the RSM survey saw a need for outside support. Whether you use one depends on in-house expertise, not company size.
Sources: RSM US, 2025 Middle Market AI Survey (June 11, 2025); McKinsey & Company, The State of AI (published August 25, 2026).
Find the constraint. Then decide what to fund.
Start with a free 12-minute Constraint Assessment.