How to Select the Right AI Projects

Artificial intelligence has great potential to support digital business growth by spurring experimentation and innovation and helping organizations operate more efficiently and effectively. But AI is no magic wand. This leaves many executives wondering: Why isn’t AI delivering on all that IT promised it would?

What is probably slowing your AI strategy down is that now, to get the greatest value from AI, businesses need to invest in strategy, not “adhocracy.”

Yes: Ten years ago, we said you should initiate AI right away, make mistakes and stumble instead of waiting to watch someone else try to keep their footing in the spooky new space. But now, we’re telling executives to slow down and first ask the questions that will define whether an AI project will fit the larger business strategy or serve as the standard that sets it. IT and business leaders must establish who is in charge, what they need, and how AI will set them up for a successful future.

Here are three key questions that executives should consider when they are approached with new ideas for AI. Data and analytics leaders should be ready to answer these questions, and maybe even pose the answers before the questions are asked.

1. Who is going to sponsor this AI project and make sure it matters to the organization?

When the answer is “a CxO” then success is much more likely. C-suite executives have access to sources of funding and influence that may provide critical. When inevitable obstacles to an AI project’s success -- such as integration costs, staff availability and security concerns -- pop up, leadership in the executive suite can get done what needs doing.

CxOs also know how to turn the CEO’s ambitions for growth or innovation into project relevancy. We talk to IT executives who understandably want to pursue AI projects that deliver results -- but results are not always enough. Value is measured in impact to the aspects of the business that get attention. For example, one client shared that they used AI to categorize millions of images, rather than having humans do it at year end. However, this task was not particularly important to the business, so no one treated the IT team that automated it as the heroes they deserved to be seen as.

2. Will this decision result in better skills, better data, and a better direction?

AI analogies are easy to come by. Let’s go to TV shows: You don’t want to be in a Twilight Zone situation with AI, where every story is new, and each episode might or might not keep you in the armchair for the full three acts. No: You want to be Star Trek, where the episodes -- or in our case, projects -- interlock thematically.

Executives should insist on enterprise-wide strategies for AI. They have already confirmed that any given project will be setting the organization up for strategic impact, so one can assume that more than one department will be committed to each initiative’s success. But workers and business leaders should also be able to see that path into a more effective future.

AI demands commitments from data leaders (management and quality), IT leaders (integration and security) and business leaders (staff impacts and value). Invest accordingly in the promise of a narrative that interlocks with others. When you care what happens in Deep Space, you also care about the Next Generation. Cross-timeline interactions are the best.

3. Is this really something we need to use AI for?

This last question is tough. Some clients tell us they use AI when they want to experiment with something familiar using a new set of skills. Some just do small tasks with AI to try and get started. But regardless of where organizations are on their AI journey, it can continue to pose a challenge.

The average AI initiative that reaches production takes 7.3 months to get there, and 10% of initiatives take at least a year (but less than two years), according to the 2021 Gartner AI in Organizations Survey. By the same token, half of such initiatives take less than six months.

We recommend that executives at least ask: Is there another way we could do this, without using AI? If the answer is no, and if the project is strategic, then it’s time to get started.

If the answer is that yes, the project can be done another way, then the experimental mindset that AI demands should be treated as even more important than usual. Measurements related to the project should include questions about its resource cost, the challenge of getting it started and accepted, and any ongoing effort that you could expect. When AI is elective, you want to be sure it is advancing the rest of the organization’s story.

By using these questions to frame and assess AI projects, IT leaders will not only have a better chance of being successful -- but they will also gain stronger support from key stakeholders within and outside of the organization, from employees to Board members to customers. Some of these questions may require research and data analysis to answer, but this preparation work will ensure that only the best AI use cases are pursued, supporting a virtuous cycle of AI investment.

Whit Andrews is a Distinguished Vice President Analyst at Gartner, Inc. researching organizational impacts, use cases and business opportunities for AI. Additional analysis on data and analytics trends, including AI, are being presented during the Gartner Data & Analytics Summit, taking place August 22-24 in Orlando, Florida.