Most companies have launched AI initiatives that have failed to move beyond the pilot stage. Nitin Jayakrishnan, chief executive officer of Freehand, explains why industry is stuck.

So many companies can’t get past the pilot stage of their artificial intelligence implementations. Why not?

The problem, says Jayakrishnan, is that a lot of companies are viewing AI as the next technology shift, “when in reality it’s the next talent or organizational model shift.”

Companies should think about AI adoption in the same way they go about hiring and training people, he says. The hiring process begins with onboarding, during which the new employee is provided with context about how the business operates in such areas as supply chain, customer relations, networks and logistics. At the outside, they’re given specific tasks that tend to be less strategic or impactful, with trust building over time, at which point the company can expand the scope of their work and build teams around them.

“That’s not how we’ve implemented software,” Jayakrishnan says.

AI requires a “paradigm shift” under which it needs to be provided with context so that it can carry out specific tasks. Large language models are trained on data on the internet, Jayakrishnan points out, but what they don’t learn is how humans react to given situations, such as how the procurement function chooses between suppliers or regions.

Much of the structured data contained in software systems is of a “point-in-time” nature. It details what happened, but not why. That’s the kind of intelligence that only humans can convey, and it can’t be found on the internet, Jayakrishnan says.

AI does present the opportunity to free humans of certain unnecessary tasks within a workflow, and carry them out in a manner that far more productive, he says.