NEW YORK — In enterprise AI, there's a gap between what's possible and what's expected — a disconnect that is hurting implementation.
The technology's appeal has led business leaders to support adoption, with more than half of executives moving to the scaling phase of the technology, according to Deloitte. As more businesses bring AI software into their expanding technology stacks, global revenue for the space will reach $62.5 billion in 2022, a year-over-year increase of 21.3%, Gartner projections show.
While the ultimate goal for AI adoption is a more efficient organization, sometimes the gap between the possible and the achievable is too big, according to Fabio Caversan, digital business and innovation VP at Stefanini US, speaking last week at the AI Summit in New York.
To deploy the right AI tool sets and technologies in the enterprise "you have to handle the expectations," said Caversan. Sometimes, the technology is ready to deliver, but the infrastructure and quality of data aren't quite there yet, according to Caversan.
The top barrier to AI adoption in the enterprise is access to talent, according to an IBM survey. But closely behind are the complexities of data and silos as well as a lack of tools for developing AI models.
One strategy to win the expectations game is to begin with simple steps, using AI to expand the capabilities of the workforce.
"Start with augmentation and then, if you're comfortable and your process allows, flip slowly to automation," said Caversan.
Enterprises can also miss the mark with their AI implementations by losing focus on their core business, according to Atalia Horenshtien, customer facing data scientist at DataRobot. It's up to business leaders to align the implementation of AI with their intent of becoming leaders in their space, said Horenshtien.
The talent component is also critical to achieve enterprise AI success, said Saira Kazmi, senior director, enterprise data engineering strategy and AI at CVS Health.
"When a business decides to use AI or algorithms for a need, the question is not just about the technology but about change management," said Kazmi. It's up to leaders to decide if and when to move the technology from experimentation to production.
"It all goes back to understanding the problem that you're trying to solve," Kazmi said.