Most organizations use AI, but few see it on the bottom line. Research from McKinsey and MIT points to the same fix: build AI into real workflows.

AI adoption is no longer the hard part. In McKinsey's 2026 State of AI global survey, nearly nine in ten respondents said their organizations regularly use AI in at least one business function, and 80% said AI has improved their individual productivity.

Turning that into business results is a different story. Only 37% of respondents attributed any impact on earnings (EBIT) to AI, about the same share as the year before. People are getting more done, but in most companies that has not yet shown up in the numbers.

The pilot problem

A 2025 report from MIT's NANDA initiative, reported by Fortune, put the gap more starkly. Based on 150 interviews with leaders, a survey of 350 employees, and an analysis of 300 public AI deployments, it found that only about 5% of generative AI pilots achieved rapid revenue acceleration. Most stalled with little measurable effect on profit and loss.

The researchers said the main problem was not the quality of AI models. It was a "learning gap": generic tools are flexible for individuals, but they do not learn from or adapt to how a specific company works. Unless AI is integrated into existing systems and processes, pilots rarely scale.

The returns are often in the back office

The same report found that more than half of generative AI budgets went to sales and marketing tools, yet the biggest returns came from back-office automation: cutting outsourced processing work, reducing outside agency costs, and streamlining operations.

For a small or mid-sized business, the back office means data entry, invoicing, scheduling, document handling, and reconciliation. This work is repetitive and measurable, and it is often where a team loses the most hours.

Integration beats experimentation

The MIT report also found that AI tools purchased from specialized vendors or built through partnerships succeeded about 67% of the time, while internal builds succeeded only about a third as often. McKinsey's survey points the same way: the challenge is shifting from encouraging individual AI use to redesigning workflows so those individual gains add up to company-level results.

What the projects that pay off have in common

Across this research, the AI efforts that deliver real results tend to share a few traits:

  • They target a specific, repetitive workflow rather than "using AI" in general.
  • They connect to the systems where the work already happens, such as accounting, scheduling, and customer records, instead of sitting in a separate chat window.
  • They have a clear before-and-after metric, such as hours per week, error rate, or days to invoice.
  • They keep a person reviewing the output where mistakes are costly, because AI performance is uneven across tasks.
  • They start small, prove value, and expand from there.

The takeaway

Buying an AI subscription for everyone is easy. Getting a return requires deciding which manual work should change, and building AI into that work so the time saved is captured by the business. Companies that treat AI as a workflow problem rather than a technology purchase are the ones seeing results.