[Workflow in the AI Era: The Line Between 'Delegate' and 'Non-Negotiable'] Over 70% Say Approval/Decision-Making 'Should Not Be Left to AI,' with 62.8% Concerned About Black-Box Judgment

ATLED, Inc., a leading company in workflow systems, announces the results of a survey on the potential for AI to replace workflows, conducted among 109 IT, general affairs, and DX promotion managers who have experience using generative AI in their work and are involved in the implementation and operation of workflow systems.

Key findings:
01｜Over 70% of IT, general affairs, and DX managers believe the 'approval/decision-making process itself' in workflows should not be delegated to AI.
02｜Top reasons to avoid AI delegation: 'judgment basis is a black box' (62.8%) and 'human responsibility' (57.7%).
03｜67.9% feel that the value of workflow systems is 'increasing' with the evolution of AI.

■Survey Overview
Survey Name: Survey on the Potential for AI to Replace Workflows
Method: Internet survey planned by IDEATECH's research marketing service 'Risapī®︎'
Period: May 15, 2026, to May 16, 2026
Valid Respondents: 109 IT, general affairs, and DX promotion managers with experience using generative AI and involved in workflow system implementation/operation.

■Most positively viewed workflow function for AI replacement: 'Automatic check and error detection of application content' at 58.7%.
In Q1, when asked which workflow functions they think are fine to be replaced or automated by AI, 'Automatic check and error detection of application content' was the top answer at 58.7%, followed by 'Automatic data entry and transfer of application data' at 46.8%, and 'Search and reference of past similar cases' at 45.9%.

■Over 70% of managers believe the 'approval/decision-making process itself' should not be left to AI.
In Q2, when asked if they think the 'approval/decision-making process itself' should not be delegated to AI, 25.7% said 'Strongly agree' and 45.9% said 'Somewhat agree,' totaling 71.6% showing a cautious stance on AI-based approvals.

■Top reason for caution on AI decisions: 'AI's judgment basis is a black box and cannot be explained' at 62.8%.
In Q3, when asked for reasons why the approval/decision-making process should not be left to AI, 'Because the AI's judgment basis is a black box and cannot be explained' was the top reason at 62.8%, followed by 'Because approval/decision-making entails human responsibility' at 57.7%, and 'Because the locus of responsibility is unclear if AI makes an erroneous approval' at 37.2%.

■Other comments include 'Human confirmation is necessary unless there is a 100% guarantee of no mistakes' and 'Risk of fraud by exploiting AI algorithms.'
In the open-ended responses for Q4, comments included, 'As long as there is no 100% guarantee of no mistakes, human confirmation is essential,' and 'There is a possibility of fraud by exploiting AI algorithms.'

■67.9% of managers feel the value of workflow systems is 'increasing' with the evolution of AI.
In Q5, when asked how the value of workflow systems has changed with the evolution of AI, 19.3% said 'It has increased significantly' and 48.6% said 'It has increased somewhat,' with a total of 67.9% feeling the value has increased.

■Reason for increased value: 'It prevents departmental inconsistencies as a company-wide approval platform' cited by over half.
In Q6, when asked for reasons for the increased value, 'Because it can prevent departmental inconsistencies as a company-wide approval platform' was the top answer at 52.7%, followed by 'Because governance and control mechanisms become more important as AI utilization advances' at 44.6%.

■Reason for decreased value: 'Application/approval processes themselves are simplified by AI' ranks high.
In Q7, when asked for reasons for the perceived decrease in value (n=6), 'Because the application and approval processes themselves are being simplified and reduced through AI utilization' was at 50.0%.

■Over half of managers want 'automatic application content check and error detection function' in AI-era workflows.
In Q8, when asked what evolution they consider important for workflow systems in the AI era, 'AI-based automatic check and error detection function for application content' was the top answer at 54.1%.

■46.8% of managers revealed to be 'not yet started' on integrating workflow and generative AI.
In Q9, when asked about the stage of integration and utilization of workflow systems and generative AI, 14.7% said 'Not considering integration' and 32.1% said 'Interested but not yet started,' totaling 46.8% who have not yet begun.

■Desirable future of AI utilization: 'AI auto-approves low-risk routine cases, humans perform post-checks' tops at 43.1%.
In Q10, when asked what they think is the desirable future relationship between workflow systems and AI, 'AI should automatically approve only low-risk, routine cases, and humans should only perform post-checks and handle exceptions' was the top answer at 43.1%.

■Conclusion
This survey revealed that while 71.6% believe the approval/decision-making process should not be delegated to AI, 67.9% feel that the value of workflow systems is 'increasing' in the AI era. This suggests that while AI is utilized to assist with routine tasks, the importance of workflow systems is growing as a platform for managing processes where final judgment is made by humans.