AI productivity gains erased by document verification

Foxit has published research questioning whether artificial intelligence is delivering the productivity gains many organisations expect in document-focused work. The study found that once time spent checking AI outputs is included, executives gain an average of 16 minutes per week, while end users lose 14 minutes.

The findings are based on a survey of 1,000 desk-based end users and 400 senior executives responsible for AI implementation in the United Kingdom and the United States. Sapio Research conducted the independent fieldwork for Foxit's report on AI use in document workflows.

Most respondents said they felt more productive after adopting AI tools, but the survey suggests those gains are offset by time spent validating results. Foxit described this as a growing "verification burden" in day-to-day document work.

Executives estimated AI saved them 4.6 hours each week, but said they spent 4 hours and 20 minutes validating AI-generated outputs. End users reported saving 3.6 hours, while spending 3 hours and 50 minutes reviewing outputs.

The perception gap extended beyond time savings. Some 89% of executives said AI boosts productivity, compared with 79% of end users. Confidence levels also diverged sharply by role.

Confidence Gap

Sixty per cent of executives said they were highly confident in AI-generated outputs, compared with around a third of end users. Only one in 10 end users described themselves as "extremely confident" in AI accuracy, versus one in four executives.

Confidence also tracked with frequency of use, raising the risk that executive confidence and end-user caution coexist within the same workflows.

Respondents cited several barriers to wider AI adoption that centred on trust and risk rather than product features. Data privacy and security concerns ranked highest at 36%, followed by trust in AI output (34%) and concerns about response accuracy (25%).

Foxit characterised these as human-centric constraints on deployment at scale, arguing that concerns about privacy, trust and accuracy can outweigh demands for additional features or clearer financial justification.

The report argues that while AI can speed up drafting and content creation, it can also increase time spent on review and correction-shaping the net productivity impact across different employee groups.

"AI accelerates creation, but it introduces new layers of review, fact-checking and correction," said Evan Reiss, SVP of Marketing at Foxit.

"What we're seeing is a verification burden emerging inside document workflows. Time saved generating content is being absorbed by the time required to trust it. The next phase of document intelligence won't be defined by more AI features, but by embedding accurate, transparent intelligence directly into workflows - reducing validation time while keeping humans firmly in control," Reiss said.

Workforce Impact

The report suggests many organisations have already started reshaping work structures around AI. Some 68% of executives said adoption had triggered restructuring or headcount changes.

At the same time, 72% of executives cited retraining or upskilling existing employees as a high or top priority. Just over 20% anticipated that more than a quarter of roles would be affected.

End users reported lower concern about job security: only 12% said they were "very concerned". The report framed this as an awareness gap between leadership expectations and employee perceptions.

Critical Thinking

Both executives and end users raised concerns about AI's effect on judgement and problem-solving at work. Over-reliance on AI for decision-making ranked as the top concern for both groups, with executives rating it 10 percentage points higher than end users.

A majority in both groups said it was very or extremely important to preserve human problem-solving skills as AI use expands. Among people who use AI multiple times a day, nearly three-quarters rated human problem-solving skills as very or extremely important.

More than 90% of daily AI users said the technology had improved their productivity. Foxit said frequent users also showed stronger awareness of AI's limitations and the importance of human judgement in AI-enabled workflows.

Metrics Shift

Three-quarters of executives expect AI use in document workflows to increase over the next year. Asked what would make AI more valuable over the next two to three years, they prioritised reliable, accurate outputs, along with shorter validation times through improved training and better integration with existing tools and systems.

The report also points to changes in how organisations assess AI investments. It says "return on employee" measures are increasingly tracked alongside traditional return on investment metrics. In the survey, 93% of organisations said they track return-on-employee dimensions, defined as measures covering capability, confidence and satisfaction as well as productivity.

"The success of document intelligence depends as much on human confidence as on technical performance," Reiss said.

"Accuracy, transparency and clear human-in-the-loop design are the foundations of trust and therefore the foundations of adoption," he added.