Insights

Buy or build? The MIT study says buying works about twice as often

Everyone quotes the 95% number. Fewer people read the part about who succeeds.

  • Ravi R. Singh
  • 2 min read

What the report says

The MIT NANDA report, The GenAI Divide, was published in mid-2025. It found that most generative AI pilots never reach scaled adoption with a measurable effect on profit and loss. Its lead author said the problem was less the quality of the models and more how organisations try to use them.

The report found that AI tools bought from outside suppliers or partners succeed about two-thirds of the time, while systems built internally succeed about one-third of the time.

It also reported that about half of AI budgets go to sales and marketing, while the biggest returns appeared in back-office automation.

Read it with care

The headline 95% comes from a small base. Coverage describes about 150 interviews with business leaders, a survey of 350 employees and a review of 300 public AI deployments. Outlets also describe the 95% differently. Treat it as a direction, not a precise rate.

What it means for a small team

If you are a small business, building your own AI system is rarely the right first move. Start with a tool that is already built, give it to one team and one workflow, and have the line manager own the result, not a central group.

Then check the second part. Is the money going where the return is likely to be? A team that spends most of its AI budget on marketing content, while invoicing and reporting stay manual, may be repeating the pattern in the report.

Sources

See the four advisory offersAll articles