63% of Large-Company Executives Believe Agentic AI Will Take 3 to 7 Years to Deliver a Return

NLX, the AI platform powering advanced conversational experiences for the world’s most admired brands, today announced the results of a commissioned survey of U.S. executives in companies with 5,000 or more employees. The survey was conducted by QuestionPro, an independent online research, insights, and experience company.

With agentic AI being the latest topic of conversation in enterprise circles, NLX is interested in when executives in large organizations are expecting agentic AI to deliver a positive return on investment. The survey results show a longer horizon for ROI than NLX would anticipate and much more muted expectations than media and investor hype around agentic AI would suggest. 

“It’s true that we’re early days implementing and scaling agentic AI, but even so, return on investment should happen within 12 months for enterprise organizations,” said Andrei Papancea, co-founder and CEO of NLX. “If it takes longer than that, companies are choosing the wrong problems to solve.”

Interestingly, while 57% of executives say they will have customer-facing agentic AI experiences in 2025, most aren’t expecting the expense to offset the benefits for at least three years, and most don’t plan to reduce headcount for at least two years.

The largest cohort of executives, 45%, believe that agentic AI will deliver a return on investment within three to five years. Twenty-two percent are more optimistic, expecting a payoff within two years. Interestingly, 33% of executives are thinking agentic AI won’t deliver a return for at least five years, with 18% saying five to seven years, and 15% seven years or more.

The sentiment on ROI on agentic AI aligns with expectations on reducing headcount, with 30% expecting to replace humans with agentic AI in the next two years, 50% saying they won’t reduce headcount in that time, and 20% being unsure.

“Choosing to focus on solving small problems with AI often seems safer, but at the end of the day, the business impact isn’t there, so it’s a failure,” continued Papancea. “It’s actually better to choose a large problem that’s well suited to be solved using AI. You can still deploy the application in a way that minimizes risk, but the business impact will be there.”

This survey was conducted by QuestionPro in February 2025 of employees of U.S.-based companies. Of the 100 responses, 45 employees worked at companies with 5,000 to 10,000 employees, and 55 at companies with 10,000 or more employees. All of the respondents had a director-level title or higher.

NLX is an enterprise AI platform for building and managing chat, voice, and multimodal applications at scale. As human interactions with technology become increasingly conversation-based, NLX makes it practical for organizations to create, deploy, and manage conversational experiences for any industry or use case; working in concert with any large language model and technology stack. NLX enables the world’s biggest brands, including Comcast, Red Bull, and United Airlines, among others, to invest in a future where interactions with technology mirror the natural ebb and flow of people’s day-to-day decision-making. Follow NLX on LinkedIn.