Why headcount no longer captures what people and AI deliver

Headcount no longer describes all the work a business can deliver

A business with 500 employees needs more than an employee count to explain how much work it can deliver. In this hypothetical organisation, some people work without AI assistance, others use AI to increase their capacity, and autonomous systems complete defined tasks independently. Headcount still describes employment, but delivery capacity depends on how these contributions fit together.

Angela Vidler, Managing Director & Co-Founder at Hyvid, predicts that this wider view of capacity will reshape workforce decisions in 2027 and beyond. Her forecast asks leaders to consider how people and machines contribute, how work is assigned, and who answers for the outcome. Hyvid has a commercial stake in these decisions, so her forecast reflects that interest.

Vidler’s proposed shift starts with the work a business needs to deliver. Leaders can then identify the capacity required and decide how people and AI will contribute. Adding agents to an employee inventory leaves these decisions unresolved, because introducing AI into a job changes individual tasks while redesigning work changes their allocation, sequence, supervision and evaluation.

Agent use is growing faster than process readiness

These decisions are becoming more urgent as businesses adopt AI agents: systems that carry out defined tasks, sometimes across several steps, with a specified degree of human oversight. Microsoft’s 2026 Work Trend Index found that active agents across the Microsoft 365 ecosystem had increased 15x year-on-year, rising to 18x in large enterprises. Microsoft sells the ecosystem in which it measures this growth, so greater agent use also supports its business.

Within that ecosystem, Microsoft’s research links greater AI value to deliberate changes in how work gets done. Businesses obtaining the greatest value reshape workflows and decide which activities people and AI should perform. Among its most advanced AI users, workflow redesign and agents handling multi-step work are becoming established practices. The findings concern how organisations arrange work as well as which tools they provide.

Deloitte’s latest 2026 research examines how leaders expect that arrangement to change. Of those surveyed, 74% expect nearly half of their business processes to be redesigned or rebuilt around AI agents within four years, while 61% expect most agents they use to operate with general autonomy under human oversight. Deloitte also has a commercial interest in demand for AI adoption and organisational change, which gives its research a business context.

Yet current readiness is much lower: only 5% of organisations in Deloitte’s research describe their processes as highly prepared for agents. Expectations about future redesign, anticipated autonomy and present preparedness answer different questions, so they cannot be combined into a single adoption measure. They do, however, reveal a planning problem within Deloitte’s findings: leaders expect substantial delegation while reporting limited preparation for it.

Microsoft and Deloitte therefore inform different parts of the planning decision. Microsoft describes agent growth and reported practices among advanced and high-value adopters, while Deloitte records leaders’ expectations and their assessment of preparedness. Leaders can use the former to examine how work is being reorganised and the latter to consider what preparation their own plans for delegation require.

Workflow redesign changes human responsibilities

Preparing for delegation starts with the question leaders ask about a process. After several years of discussion about generative AI, AI used to produce content, Vidler expects attention to move from “Where can we use AI?” to “How should this work be redesigned now that AI can do part of it?” The decision shifts from finding individual uses to examining how an entire business outcome is produced.

Examining that outcome means deciding which tasks remain with people, which receive AI assistance, and which can be delegated to agents. Vidler’s proposed redesign then requires leaders to specify when work returns to a person and how completed output will be assessed and governed. Each allocation changes the responsibilities around it: more machine execution creates work for people supervising, validating and managing the result.

Vidler illustrates this change through a person who oversees, checks and manages substantially more work as AI handles a larger share of repetitive execution. The person supplies judgement, context, accountability and responses to exceptions. This predicted division of responsibilities explains why Vidler rejects “AI replaces the employee” as an adequate description of the transition.

Deloitte’s findings offer a separate view of the value executives place on this division of work: 75% of those surveyed say human–agent collaboration creates more value than agent automation alone. Their preference concerns the contribution of people working with agents. It values the person who judges whether an output is suitable and decides when to intervene, even as agents perform more of the execution.

Supporting those responsibilities also requires changes around the employee. Microsoft’s research found that organisational factors, including culture, manager support and talent practices, account for more than twice the reported AI impact of individual factors. In Microsoft’s findings, the environment in which people use AI matters substantially, so workflow redesign extends to management expectations and support. Widely available models provide a capability; how a business organises their use can become a stronger source of differentiation.

Boards need to govern delegated AI work

The allocation of work also determines what boards need to oversee. For decades, employee oversight has covered headcount, remuneration, capability, succession and organisational structure, while AI and automation have generally sat within the technology remit. Vidler predicts that this separation will become less useful as agents take on business tasks, because their execution contributes directly to what the organisation delivers.

That contribution changes the demands of oversight when an agent moves from a trial into recurring work. A trial may be assessed for technical promise, but recurring responsibility requires an owner, agreed limits and a way to judge performance. The agent’s permitted decisions and access to information shape the outcome, so boards need visibility into delegated AI work alongside work assigned to employees.

To make that visibility practical, Vidler sets out six questions. They connect the agent’s assigned work to its authority, performance and human accountability:

• What is it responsible for?
• Who owns it?
• How is it performing?
• What decisions can it make?
• What data can it access?
• And who is ultimately accountable when it gets something wrong?

The answers give the board a way to assess what delegation means in practice. Assigned responsibility provides a basis for judging performance, while decision permissions establish the agent’s authority. Data access defines the information available during execution, and ownership identifies who must act when results fall outside expectations.

Vidler expresses the change through two imagined board questions: “Are we using AI responsibly?” and “Do we actually know what our AI workforce is doing?” The second turns a general commitment into a request to see what agents are doing. In her proposed model, people retain responsibility for judgement, validation and exceptions, so oversight must identify the accountable person or function whenever a system performs the task.

Global talent must be assessed for AI-enabled output

The same link between execution and human responsibility changes how leaders assess global teams. The conventional proposition has centred on access to skilled people at lower labour cost, making the cost of an individual employee a central measure of value. Vidler expects businesses to place greater weight on the output an AI-integrated team can deliver, changing how they design and source teams.

Over the next 12 months, Vidler predicts that an AI-enabled individual in a global team may take responsibility for substantially more work, extending the supervisory role described earlier to global teams. That prospect changes team economics because employment cost must be considered alongside the volume of work a person can manage with AI. It also makes skills for supervising execution and evaluating results more important.

Vidler therefore rejects “AI replaces global talent” as the defining account of the change. She expects people and technology to create greater capacity together, with consequences for responsibilities, skills and career paths. Labour cost remains relevant, while capability and output gain weight in the assessment: leaders comparing global-team options need to understand what each team can reliably deliver under its proposed division of responsibilities.

Workforce planning must combine skills, tasks, roles and capacity

Comparing teams on delivery capacity requires a wider set of workforce-planning inputs. For decades, planning has largely asked how many employees an organisation needs, which skills they require and where they belong in its structure. Those questions still guide employment decisions, but the hypothetical business described at the outset also needs to account for work performed by agents and the changing capacity of AI-assisted employees.

Deloitte identifies workforce planning as one of four priorities for adopting agentic AI, AI organised to carry out tasks through agents, alongside governance, strategy and trust. Its research on AI-enabled strategic workforce planning describes a move towards skills, tasks, roles and capacity. These inputs reveal different parts of the delivery model: roles describe responsibilities, tasks identify work that can be retained, assisted or delegated, and capacity connects those allocations to the required output.

Using this wider set of inputs, Vidler predicts more planning in which human roles, AI-augmented roles and digital workers share an operating model. Digital workers are systems assigned defined execution responsibilities. Planning these contributions together lets leaders examine how delegating a task changes demands elsewhere, including the human time needed for review, coordination and exception handling.

Vidler frames the allocation decision by moving from “Can AI do this?” to “What is the right combination of human judgement, AI capability and governance for this piece of work?” Starting with the work makes its required capacity, contributing roles and decision authority part of the same choice. The plan must then connect each contribution to a delivery responsibility and identify who answers for the result.

The extent of that redesign determines which of two adoption paths Vidler predicts a business will follow. She expects companies that add AI to existing jobs to obtain incremental productivity improvements, while those that reorganise work around people, augmented roles and agents adopt a different workforce model. Choosing between these approaches sets the scope of preparation, because changing task allocation also changes the supervision, skills and coordination the organisation needs.

As those requirements change, workforce planning needs a more frequent review cycle. Deloitte describes scenario planning becoming more continuous, allowing leaders to test different combinations of human effort, AI assistance and delegated execution as conditions develop. Each scenario can examine the resulting capacity, skills and responsibilities, then inform workforce decisions when the expected relationship between employee numbers and output changes.

Main highlights

• Plan around delivery capacity: Workforce planners need to map the output required across human roles, AI-assisted roles and autonomous systems. That assessment connects staffing decisions to the work the business can reliably deliver.
• Test readiness before expanding delegation: Deloitte reports that 74% of surveyed leaders expect substantial process redesign around agents, while only 5% describe their processes as highly prepared. Process owners need to assess supervision, handoffs and exception handling before assigning agents recurring responsibilities.
• Redesign human responsibilities alongside AI tasks: Workflow owners need to define who validates AI output, supplies context and handles exceptions. Managers then need to align skills, support and performance expectations with those responsibilities.
• Make delegated AI work visible to boards: Each agent needs an accountable owner, a defined remit, decision limits, data permissions and performance measures. Boards need visibility into these arrangements as agents take on recurring business work.
• Assess global teams by AI-enabled output: Hiring and sourcing decisions need to account for the work people can manage with AI, alongside employment cost. Vidler predicts that this shift will increase the value of judgement, supervision and output validation.
• Review workforce scenarios continuously: Workforce planners need to test combinations of skills, tasks, roles and AI capacity as capabilities change. Each scenario needs to include the human time required for review, coordination and exception handling.