Artificial intelligence (AI) is playing an increasingly prominent — and fast-evolving — role at large U.S. firms. While the use of AI has become nearly universal across enterprises, current implementations of generative AI consistently require human involvement from start to finish. This suggests that the promise of agentic AI, a next-level iteration that has the potential to fully automate core business functions, including high-risk tasks, without human supervision, remains a distant prospect for many enterprises. Additionally, large companies are deploying GenAI in narrow, use-specific cases, not in integrated, connected modes across the entire organization. For both AI types to have a transformational impact, companies must think of AI as a team player — even when that team doesn’t include other humans.
GenAI Technology Not Unbound
While AI adoption has become nearly universal across enterprises, current implementations of GenAI consistently require human oversight. Most tasks that chief operating officers (COOs) perform using GenAI still necessitate human operators for prompting and assessment; in other words, from start to finish. Such tasks include writing reports and software code, generating data, creating content and providing customer support.
Full automation using GenAI is currently limited to narrow and industry-specific use cases. For example, services firms tend to automate software code generation, while technology companies lean toward automating fraud detection.
Enterprises with high levels of automation see their concerns about return on investment (ROI) in AI dissipate, with none of them worried about a limited ROI. But there’s a trade-off: Data security and privacy concerns are significantly more pronounced for high-automation firms, with 80% citing these worries, compared to 39% of low-automated firms.
Paradoxically, the more a company uses GenAI to automate tasks, the more it reports unfamiliarity with the tool’s inner workings and a lack of skilled labor to guide it. Less-automated firms are less likely to report these drawbacks, suggesting that higher levels of human involvement reduce the perceived challenges of using the technology.
Conclusion
Large U.S. firms are diving into GenAI, but with buoys. Despite the nearly universal adoption of the fast-evolving technology, its current uses consistently involve human oversight throughout the entire process. While full automation via GenAI is restricted to specific and narrow use cases, increased levels of automation lead to greater concerns about data security and privacy, alongside reported unfamiliarity with the technology’s inner workings and a scarcity of skilled labor. Ultimately, for AI to achieve its transformational potential and move beyond its current narrow applications, companies must view it as a trusted collaborative partner, even as the vision of fully agentic AI remains a distant prospect for most organizations.
