95% of AI pilots return nothing. What the other 5% do differently
26 August 2025

95% of AI pilots return nothing. What the other 5% do differently

An MIT report this month found that 95% of enterprise generative AI pilots produced no measurable return.

The technology is not the problem. The way it is applied is.

The finding

The report, from MIT's NANDA initiative, looked at hundreds of corporate AI projects. Most stayed as pilots and never changed a number on the profit and loss account. A small group, around one in twenty, produced real results.

What the 5% do

  • They pick one process, not ten.
  • They connect AI to their own systems and data.
  • They buy proven tools where they exist instead of building everything.
  • They give the project an owner inside the business, not only in IT.
  • They measure before and after.

What this means for you

A general chat assistant on every desk is useful, but it rarely shows up in results. An AI that handles customer enquiries from start to finish, or plans maintenance jobs, does. That is the principle behind our own products, BizBot and Tessaro: AI placed inside a specific piece of work, where its effect can be counted.

Frequently Asked Questions

They are not connected to real processes and data, so they never move beyond a demonstration.
Long enough to measure one number properly. For most processes, sixty to ninety days.

Choose a pilot that will show up in your numbers.

Tell us about your business and we will show you where AI agents can take work off your team.

Talk to us