Companies are under pressure to demonstrate returns on AI investments quickly. Speeding up a single task, however, may not improve the outcome of the overall process. Ee Lyn Khoo argues that managers need to redefine workflow dependencies, accountability and how people collaborate with automated systems.

Ee Lyn Khoo argues that pressure for rapid ROI pushes companies toward a simple approach: take an existing process, add an agent and count a faster individual step as success. That does not address how the work is organized end to end or who owns the outcome produced jointly by a person and an automated system.

Citing PwC’s 2026 Global CEO Survey, the author notes that 56 percent of CEOs say they have not yet seen significant financial benefit from AI in the form of revenue growth or cost reduction. She also cites a study commissioned by Writer in which roughly two-thirds of companies report spending at least $1 million on AI technology. Investment alone therefore does not guarantee a visible financial result.

The costs also extend beyond the model price. Project economics include token usage, security and governance reviews, data cleanup and employee training. If the company leaves the original process unchanged, it may simply place a faster tool inside a system that still contains unnecessary approvals, waiting and manual checks.

The author also points to apparent savings. A system may produce a draft document in 30 seconds, while an employee then spends an hour correcting it and issuing more prompts. If management measures only one task rather than the full workflow, the case may be reported as automation even though little real capacity was released.

Even when time is genuinely saved, management must decide what happens to the reclaimed capacity. A team can handle more volume, work on previously deferred tasks or stop doing an activity altogether. Without such a decision, time savings may never appear in the profit-and-loss statement.

Khoo therefore proposes treating the person-agent combination as a single unit of accountability. A manager should track what that combination produced and what the company did with the capacity it freed. Measurement then shifts from the number of automated tasks toward outcomes such as speed to resolution or customer impact.

For managers, this creates new demands on judgment. They need to know when to trust a system’s output and when to interrogate it, when employees are spending more time supervising a tool than doing the work, and where a decision must remain clearly human.

The author distinguishes between redesigning work and reimagining it. Redesign improves an existing process, while reimagining starts with the outcome and treats the process as negotiable. Her product-discovery example contrasts using AI merely to summarize customer-support input with using agents to turn customer and partner feedback more directly into working prototypes.

Such disruption is not appropriate for every process at once. Khoo recommends choosing a small number of activities where the outcome truly matters and where the old way of working strongly constrains performance. The competitive advantage then comes not from buying the same tool as everyone else, but from how the company rebuilds work, accountability and decision-making around it.

KEY TERMS

• Workflow design: The arrangement of steps, responsibilities and dependencies through which a work result is produced.
• Automation: Transfer of selected activities to a technological system with less manual work.
• Accountability: Clear assignment of the person or role ultimately responsible for the result.
• Productivity: The relationship between the output produced and the resources, time or labor used.