As expected, Gartner kicked off its annual Symposium/IT Expo this morning with a focus on AI. Specifically, the company discussed the difference between “AI-steady” and “AI-accelerated” companies, and how this affects their technology strategies.

In their presentation titled “Pacing Yourself in the AI Races,” Gartner analysts Hung LeHong and Mary Mesaglio explained that we are currently at the peak of the Gartner Hype Cycle for AI, but also at the trough. This is because there are two different races happening simultaneously. One is for tech vendors, with new foundation models being released every two and a half days. The other is for companies to deliver AI outcomes safely and at scale, which requires setting a sustainable pace.

Mesaglio noted that for years, only about 20% of CEOs believed that AI would impact their industry. However, this number has risen to 59% in 2023 and 74% in 2024. Both analysts emphasized that it is the AI outcome race that truly matters.

To determine whether their companies are “AI-steady” or “AI-accelerated,” tech executives should consider the business, technology, and behavioral outcomes they are trying to achieve. AI-steady companies typically focus on productivity gains, while AI-accelerated companies aim for other advancements in addition to productivity. Gartner predicts that 50% of the benefits from AI will come from productivity gains, 30% from process improvements, and 20% from changes to the business model. These companies often see AI as a way to reinvent their industries and have multiple proofs-of-concept in development.

However, LeHong cautioned that achieving productivity gains from AI is more challenging than it seems. Mesaglio shared that while 98% of employees are eager to try AI, 72% of IT leaders report that employees struggle to integrate it into their daily work. Nevertheless, those who do successfully incorporate AI report saving an average of four hours per week or 43 minutes per day. One issue is “productivity leakage,” where employees use the extra time for non-work-related activities, such as getting a latte.

LeHong also pointed out that the benefits of AI are not evenly distributed, unlike traditional automation. For example, in a call center, new employees may see a significant boost, while experienced employees may not. On the other hand, a junior lawyer may not benefit much from AI because they lack the necessary expertise, while a senior lawyer may see a significant improvement.

In general, AI is most beneficial for low-complexity jobs performed by less experienced workers, while it is more valuable for high-complexity jobs performed by more experienced workers. Therefore, companies should focus on their “deep productivity zones” to maximize the benefits of AI. LeHong cited examples such as Austin and Vancouver using AI to automate building permit requests, and how John Deere is using AI to optimize crop yields. However, it is crucial to ensure that key information is not lost in the process.