Understand what AI replaces, not whom

There is a quiet test I use when thinking about AI and the future of work: "How do you actually spend your day at work?"

Most people answer with a job title. The more revealing answer is the list of tasks that make up the role.

That distinction matters. AI is not transforming work one job at a time; it is reshaping activities within jobs. Some tasks are highly structured and data-intensive. Others depend on judgment, creativity, expertise and human connection. Understanding how those activities are changing provides a much clearer view of workforce transformation than focusing on jobs alone, especially because these shifts often occur long before the organizational chart is adjusted or a job disappears.

For boards and executive teams, this evolution raises questions that extend beyond cost optimization and productivity to workforce performance, organizational agility, employee experience, risk and long-term value creation.

The data is striking

WTW’s AI Workforce Transformation solution moves the automation debate from “which jobs will vanish?” to “which activities will change?” Applying its WorkVue Agent to deconstruct more than 900 occupation profiles representing over 50,000 jobs from the U.S. Bureau of Labor Statistics’ O*NET database at the task level, it provides evidence organizations can actually act on.

The core finding: 20% to 35% of tasks in professional and judgment-intensive roles are amenable to automation and AI augmentation. This proportion can be as high as 60% to 70% of tasks within highly structured roles like operations, administrative and clerical functions. These are not marginal jobs. They include accountants, loan officers and compliance professionals, the analytical backbone of organizations around the world.

The value of this analysis is not the automation percentage alone. It gives organizations a fact base for deciding where to automate, where to augment human capability and where to redesign work to improve performance.

The wrong unit of analysis

For decades, workforce planning has treated the “job” as its primary unit: hire for jobs, benchmark jobs, eliminate jobs. With AI, that reflex produces the wrong question: “Which jobs will it replace?” The more relevant framing is to look at tasks and the skills required to do those tasks.

Within almost every role there is a mix of tasks. Some activities demand judgment, creativity and relational trust while others are repetitive and data-intensive — exactly where automation excels. The future workforce will not be defined by job elimination but by task reconfiguration, skill alignment and human-machine integration at scale.

This requires organizations to deconstruct roles into component tasks, determine where AI can automate work or augment human capability, and reconstruct roles, workflows and teams around the intended business outcomes. Skills, rewards, governance and change priorities must then be aligned to make that redesign sustainable.

Indeed, the partnerships between HR and technology functions, between CHROs and CIOs have never been stronger or more critical. In the AI era, people and technology are inextricably linked.

AI changes tasks within jobs

AI can automate some activities, augment others and create entirely new forms of work. In operational settings, AI-enabled tools can improve inspection, planning and scheduling. In professional roles, language models can accelerate research, analysis and drafting. In both cases, people remain responsible for applying judgment, context and accountability.

Across the trades, robotics and wearables are extending human capability while improving safety. Exoskeletons can reduce physical strain on employees, while simultaneously generating performance data that feeds optimization models.

Whether the work is physical, operational or professional, the relevant question is not which roles are immune. Rather, which tasks require the skill, dexterity and improvisation that make automation unreliable and inefficient today. That calculus may very well change tomorrow.

The cost-trap can prove expensive

The business case for automating knowledge and physical work is immediate in high-cost labor markets (United States, Western Europe, Singapore, Australia). Discovering that more than half the task content in structured operational roles is susceptible to automation and AI augmentation is not an abstract risk; it is a budget optimization decision waiting to be made. But cost-only logic may prove the proverbial penny-wise and pound-foolish.

In one instructive case, a company running a back-office BPO in South Asia automated a significant portion of its processing workforce with large language models. When AI token costs rose, the calculation reversed, making human processing cheaper. But the workforce had been dismantled and institutional knowledge had dispersed, so rebuilding proved expensive.

The experience of Swedish fintech Klarna offers a similar lesson. Between 2022 and 2024, the company replaced approximately 700 customer service agents with an AI-powered assistant, reporting a $40 million profit improvement.

By mid-2025, the company was rehiring. The AI handled volume efficiently but could not manage the nuanced, emotionally charged interactions that define financial services at its margins. CEO Sebastian Siemiatkowski acknowledged it directly when he said, “We focused too much on efficiency and cost. The result was lower quality, and that’s not sustainable.” By early 2026, Klarna had rebuilt around a hybrid model: AI for the routine, humans for the complex.

Neither case argues against AI adoption, but both argue against treating it purely as a cost arbitrage play.

The more durable rationale is sustainable performance — designing the right combination of people, skills, technology and work so the organization can adapt as capabilities and economics change.

This is a board question

The board’s fiduciary responsibility includes oversight of all enterprise-wide intelligence whether natural or artificial. As AI becomes embedded in work, that oversight increasingly extends to workforce capability, skills development, organizational change and the management of operational, ethical and reputational risks associated with AI adoption.

WTW’s research on the expanding mandate of remuneration/compensation committees points to a structural shift already underway. Similar to a wave of board committee changes in the U.S., other regions are expanding the terms of reference to include talent management, succession planning, organization development and, increasingly, workforce transformation.

An important question is whether boards have sufficient visibility into AI workforce transformation. This includes understanding not only headcount efficiency, but also skills investments (i.e., which capabilities are we building for the next five years?). Indeed, progressive compensation and talent committees are asking HR teams to prepare a task-level AI and automation exposure map for mission-critical roles as well as top roles by headcount cost.

While “How much will AI save us?” is a relevant question, it is equally important for boards to understand “What will we lose if we get this wrong?”

Back to the task at hand

Bill Gates popularized an observation that we overestimate what technology can do in a year and underestimate what it will do in a decade. On AI and workforce transformation, many organizations are committing both errors simultaneously, overestimating the time they have available to act and underestimating how quickly task-level disruption compounds. When such a large portion of knowledge-based roles face task-level disintermediation and substitution, the window for proactive governance is not a distant horizon; it may be as imminent as the next budget meeting.

The what-do-you-actually-do question was always intended as a provocation, not a prophecy.

Across sectors, labor markets and cost structures, AI is changing how tasks are performed, connected and governed. What endures across every sector, every labor market and every cost structure is the premium on judgment, relational intelligence and accountability; work that machines still struggle to perform reliably, cheaply and safely at scale.

Companies need to understand their task-level exposure and use that insight to design the right combination of people, skills and technology. By reconstructing roles and workflows around where AI and people each add the greatest value, organizations can turn AI ambition into better performance, stronger workforce outcomes and measurable business value.