Strategies to successfully deploy AI in the enterprise

More enterprise use of AI does not correlate to success of the business.

On Oct. 21, Deloitte released the fourth edition of its "State of AI in the enterprise" survey. The survey examined the practices of 2,875 executives from 11 countries. This year Deloitte looked at how successful enterprises were in deploying AI .

Deloitte found that while more and more enterprises are deploying AI, not all are successful.

In this Q&A about deploying AI in the enterprise, Beena Ammanath, executive director at Deloitte, says the success of AI in the enterprise relies on a number of factors, chief among them the team that's leading the enterprise's application of AI.

Why was one group of executives able to see such good outcomes despite their low deployment of AI in the enterprise?

Beena Ammanath: [They] clearly are the ones everybody wants to learn from. A few things that we started seeing was having an enterprise-wide AI strategy. They made AI a key element of business differentiation and success. The strategy came from the business leaders and not from the data scientists or the IT teams. Having the business leaders drive the strategy was one of the best practices.

The other one was really the focus on culture and change management. [Some executives] had invested heavily into change management because working with AI leads to new ways of working. That needs a cultural change, that needs some level of change management, and you need to be able to get all your employees on board, to be able to succeed with this new technology. So, culture and change management and putting a focus on that upfront was another we saw in this group.

The third one was really leaning into ecosystems and partnerships. The companies that leaned more heavily into the AI ecosystem, whether it's startups or academia or partners or vendors … they're ahead.

Why does it matter for enterprises to have business leaders take charge of strategies for AI in the enterprise?

Beena Ammanath

Ammanath: Even a few years ago when AI was just getting started and new, most of the ideas of how AI should be used within a business came from the data scientists. They understood the technology and they felt that they could guide on what should be done with the technology. It was very tempting for the business leaders to let the data scientists lead the application of the technology. Now we see more and more pivot into putting the business requirements first. Even a few years ago, the question would be, 'Oh I have all this data, what insights can I get from it?' The question has to change into: 'You know I have this business problem, how can I solve it with AI, and data?' So, starting with the business problem, starting with the business requirements, and then figuring out which technology to use, that's the nuance behind it.

Can this lead to problems?

Ammanath: At the end of the day, there may not be a technology solution right for the business problems. It becomes a partnership between the technology team and the business leader but being led by the business leader. And I've personally been in scenarios where there is a business problem but it's not something that can be solved with data and AI. The technology is just not mature enough. The most successful companies are the ones that have that close partnership between the technology team and the business team.

Most companies still see AI as a competitive edge, and if they don't invest in AI they'll be left behind.

For enterprises that are underachievers, and their large-scale deployment of AI did not result in positive outcomes, does this discourage their use of AI in general?

Ammanath: I guess it's a bit too early still. Most companies still see AI as a competitive edge, and if they don't invest in AI, they'll be left behind. So, across the board they are planning to continue to invest in AI, and some of the best practices that we're putting out hopefully help them move up from underachiever status. They are continuing to invest, and they do see AI as a competitive edge so it's not, they're not disillusioned yet.

For enterprises that are just starting, what is the reason for their late deployment of AI?

Ammanath: They just did not start early enough. They're also the group that's least likely to demonstrate some of the best practice behaviors. They will average about 1.6 out of 10 possible full-scale deployments of different types of AI. They are purely lagging. And what we've seen is the late start has impacted their ability and they must do a bit of catchup.

I think we're going to see, based on the best practices that's coming out and where the field is, they might be able to accelerate. Even though they got a late start, they're more likely to accelerate because the industry has matured. The best practices are coming out more frequently so they hopefully don't have to experiment as much, they just have to get their foundation base set up, and then they can move faster.

Editor's note: This interview has been edited for clarity and conciseness.