Executives are saving just 16 minutes per week using artificial intelligence after accounting for the time spent reviewing AI-generated work, according to new research from Foxit, raising questions about whether businesses are realizing meaningful productivity gains from their AI investments.
The secure PDF software company surveyed approximately 1,400 executives and end users for its AI and Document Work report, finding that widespread adoption has not necessarily translated into substantial time savings.
The results also highlight an opportunity for managed service providers (MSPs) to help customers evaluate AI tools based on measurable business outcomes rather than adoption alone.
AI output reviews erase much of the time savings
According to the research, executives reported net time savings of approximately 16 minutes per week after accounting for time spent reviewing AI-generated outputs.
In an interview with Evan Reiss, SVP of Marketing at Foxit, Reiss explained that the survey asked how much is being measured and how much is being seen in time savings, along with how much time is actually being spent reviewing the quality and accuracy of outputs.
Reiss sees reducing the time spent reviewing AI outputs as a technology problem that technology companies must address. Specialized applications need to ground organizations’ AI outputs in accurate, up-to-date source documentation, alongside a centralized, structured, and well-maintained knowledge system to provide reliable context for AI.
“That’s where you see a lot of these organizations start feeling the negative impacts of AI – where it really is like a dirty data in, dirty data out situation,” Reiss said. “They’re realizing ‘we just bought all of this AI, all of these AI-enabled tools, and we completely missed from an IT standpoint.”
Executives use AI more frequently than employees
Further, the research found that 85% of executives reported using AI daily, compared with 68% of end users.
Reiss suggests that the gap may be attributable to executives’ experience in using AI to brainstorm, ask questions, and generate specific outputs, while end users may lack training, established connectors, or an understanding of how to incorporate AI into their workflows.
Organizations should help employees learn how to use AI effectively while avoiding excessive dependence on it, Reiss notes. He advocates for policies that preserve critical thinking and require employees to produce original work before using AI to further develop it.
“I think it puts the burden on organizations, ops teams, and IT teams to really define what type of people are emerging in the workforce,” Reiss said.
Trust and training gaps complicate AI adoption
The gap in AI usage extends beyond mere use to trust, learning, and AI-generated work.
Reiss noted a trust gap between executives’ perceptions of AI use and end users’ actual experiences. End users may be less trusting of AI outputs than executives believe, the report found.
According to Reiss, producing original work is part of the learning process, particularly for knowledge workers and creative professionals. Reiss argues that employees need to understand and develop information themselves rather than simply relying on AI to generate it.
“If you’re in a role where you have to be a steward of knowledge and information, part of the knowledge process and part of the learning process is the critical thinking to produce a document or produce some sort of knowledge or information for the first time,” said Reiss. “Even if you’re in design, right? [Critical thinking is needed] to produce the atomic elements and assets of creativity to create something new.”
Measuring AI’s impact beyond productivity
Reiss recommends evaluating AI through “return on employee” (ROE), not just traditional return on investment.
Time savings alone may not demonstrate value when employees must spend additional time validating outputs – as demonstrated by the reported 16 minutes per week of net time savings figure.
Organizations should identify specific workflows where AI can replace or accelerate parts of the process. For example, Foxit has used AI to rapidly prototype website designs and reduce the UI/UX time required to move pages into production.
For individual knowledge workers, organizations should also consider employee satisfaction, job effectiveness, and access to new information, even when these benefits cannot be quantified in financial terms.
“It has to be looked at from sort of a holistic return on employees,” Reiss explained. “Are the employees happier? Do they perceive that they’re enabled to do their job better, and have they figured out how to access or get information that they would otherwise not have? Or are they just using it simply as a planning tool and it’s faster than putting things in their calendar on their own?”
Why AI outsourcing might hurt talent pipelines
While AI is taking over some research, information gathering, and reporting tasks traditionally assigned to coordinators, analysts, and junior employees, Reiss argued that these tasks are crucial to helping employees develop a deeper understanding of business data and organizational trends.
Organizations risk weakening their future talent pipeline if they stop investing in entry-level roles and outsourcing these tasks to AI.
Managers and executives need to make intentional decisions about hiring, employee development, and success planning. Without those investments, organizations could struggle to transfer knowledge between employees and across the business.
“If we just use AI for everything … we’ll eventually end up with a workforce that isn’t playing ball for the future,” said Reiss.
What the findings mean for MSPs and IT partners
MSPs should understand clients’ entire technology stacks and identify overlapping or shadow AI capabilities. Partners need to distinguish between general-purpose AI tools and specialized AI-enabled applications.
Reiss used document technology as an example of how specialized applications can handle PDFs, complex document layouts, and unstructured data differently from general-purpose models.
“They have to look at the whole stack, and they have to make deliberate choices between: am I providing a generalist AI solution that enables the broader workforce? Or do I need to provide specialty technology solutions that are AI-enabled to enable very specific teams?” said Reiss.
MSPs should determine whether clients need broad AI capabilities for the workforce or specialized solutions for specific teams and tasks.
Reiss argues that executives may overestimate the extent to which AI improves their capabilities, and that organizations should identify hidden productivity and capability gaps that AI tools may not address.
Software should improve knowledge workers’ effectiveness and be grounded in a real business context. The focus should be on finding solutions that address actual business needs rather than assuming AI adoption automatically produces value.