A few years ago, the state of Utah had a serious problem. Construction projects were blowing through budgets. They consumed contingency funds and then some. And there were no obvious reasons why.

So the state enlisted the services of Richard Byfield to serve as the new director of construction management. He looked at the problem and targeted subjective bias in procurement as an issue. He thought artificial intelligence might be a more objective buyer. So he tried it on a project.

The results were dramatic. The AI approach produced one of the best construction projects in the state of Utah. It finished on time, on budget and with no change orders.

The surprise to this story, published in the Journal of Construction Engineering and Management (paywall), isn’t that AI solved a business problem. Today, we expect it. It’s that the story is from 2002. And that you’ll find few similar or more recent examples of AI in procurement in the literature. (See for yourself.)

Why? Machine learning is a great fit for procurement. Purchasing generates lots of data to learn from. And there’s a clear path to ROI.

But today, procurement leaders prioritize baseline analytics over AI. Challenges with data quality, integration and availability inhibit more advanced machine learning. Yet analytics can only identify low-hanging fruit for savings. Not high-hanging fruit, like subjective bias.

There’s hope, however. Because AI isn’t only for novel insights from reliable, robust data. It can also help get your data that way in the first place.

‘Almost Nonexistent’