Why you may be falling behind in the AI race

In their 2000 book, “The Knowing-Doing Gap,” Stanford business professors Jeffrey Pfeffer and Bob Sutton examined one of the epic struggles of running a modern company—turning talk into action.?

A quarter-century later, AI seems to be putting executives under the same scrutiny: They talk a big game and report about how much is riding on AI for their long-term success. But walking the talk—especially with generative AI—so far seems to be a different story.?

In a recent AWS survey about generative AI with over 300 chief data officers, 80% of CDOs said they believe gen AI has the potential to transform their business. How are they acting on those ambitions? Here’s a quick reality check:

• Only 19% said they had developed business-unit-level gen AI projects?
• 11% said they had developed organization-wide projects
• 6% had deployed gen AI in production

The biggest roadblocks

What explains this new knowing-doing disconnect? In his latest report for The Works, journalist Dan Tynan explains what’s slowing down many companies: To begin with, many have not yet defined or identified the right use cases, which is why individual experimentation with ChatGPT for personal productivity remains the dominant use case at most companies.?

Second is insufficient data quality or quantity to help generative AI projects get off the launchpad. (One chief data officer we interviewed put it bluntly: “Most enterprises are missing the data quality, privacy, security, and governance needed to do trustworthy AI.”)?

Lastly is an aversion to risk and experimentation; the speed at which gen AI is evolving means risk-averse slowpokes are at risk themselves of being lapped by competitors. Read Tynan’s reporting about what companies need to do to close the knowing-doing gap with data.

Fast-track strategies?

Fortunately, longtime MIT researcher Andrew McAfee offers some practical advice for accelerating generative AI pilot projects. In an interview with reporter Kristin Burnham, McAfee lays out some specific strategies, including:?

1. Conduct an audit of existing jobs. This will help leaders determine where generative AI can offer the biggest value. “A good ground rule is the more tasks a worker does with language, the more generative AI can assist that person,” says McAfee.
2. Consider off-the-shelf AI. Publicly available large language models (LLMs) can serve in many organizations as a handy “naive assistant,” writing code or creating project management plans.
3. Prioritize high-impact projects. This helps solve one issue—lack of clear use cases—that CDOs noted in the AWS study. McAfee adds: “Success means having a clearer idea of where the big potential benefits are to be found.”

“The more tasks a worker does with language, the more generative AI can assist that person.” — Andrew McAfee

Extra points: Building an AI center of excellence

For companies already walking the talk with generative AI, bestselling author and enterprise tech guru Tom Davenport encourages them to lock in a long-term competitive weapon: building an internal AI “center of excellence.” In his interview with The Works reporter Howard Rabinowitz, Davenport discusses the initial steps companies can take to get started.

Enjoy what you’ve read? Don’t forget to subscribe to receive future Freshworks Connect editions.