Adding artificial intelligence to the workplace sounds simple: give employees an AI tool, automate a few repetitive tasks, and enjoy higher productivity.
In reality, meaningful transformation is much more complicated.
When AI can research information, generate documents, analyze datasets, write code, coordinate workflows, and complete increasingly complex tasks, companies eventually need to rethink the jobs surrounding those capabilities.
The question shifts from “How can AI help this employee?” to “How should this work be organized now that humans and AI can perform different parts of it?”
That is why companies increasingly need to redesign jobs around AI-assisted workflows rather than simply automate isolated activities.
McKinsey reported in September 2026 that an estimated 57% of current work activities could be performed using existing AI technology, while emphasizing that this does not mean the same percentage of jobs will disappear.
Roles are more likely to change as tasks are redistributed between people and technology.
The real opportunity is job redesign: identifying what machines should execute, where humans should make decisions, and how both can work together more effectively.
Start by Redesigning Tasks, Not Eliminating Jobs
Organizations often begin AI planning with job titles.
They ask whether accountants, recruiters, marketers, analysts, or programmers can be automated.
That is usually the wrong starting point.
Jobs contain collections of tasks, and those tasks have very different levels of automation potential.
A recruiter, for example, might write job descriptions, screen applications, schedule interviews, speak with candidates, evaluate cultural fit, negotiate offers, and advise hiring managers.
AI could potentially handle much of the scheduling, initial information processing, and document preparation.
That does not mean the entire recruiter role disappears.
Instead, the job may shift toward candidate relationships, hiring judgment, workforce planning, and advising business leaders.
Deloitte’s workforce tools are specifically designed around assessing AI’s impact at the task and skill level, reflecting how companies increasingly need to examine workflows more granularly.
Map the Workflow Before Adding AI
A common mistake is automating an inefficient process without questioning why the process exists.
Imagine a financial report moving through six departments.
One person gathers information. Another cleans it. Someone else analyzes it. A fourth employee creates the presentation. Two managers then review everything.
If AI can perform several of those steps, simply inserting automation into each department may preserve unnecessary handoffs.
McKinsey’s 2026 research on workflow redesign argues that major AI opportunities can exist between workflow stages, where work moves across teams and systems.
It reported that 80% of respondents said AI improved individual productivity, while only 37% attributed any EBIT impact to organizational AI use.
That gap is important.
Individual productivity does not automatically translate into enterprise value.
Companies need to map the entire workflow and ask whether steps can be combined, removed, reordered, or handled by AI before redesigning individual jobs.
Separate Human Tasks From Machine Tasks
Once companies understand the workflow, they can decide which activities belong primarily with humans and which can shift toward AI.
Machine-friendly activities often include repetitive processing, summarization, basic content generation, data transformation, scheduling, monitoring, and standardized analysis.
Human-heavy work tends to involve accountability, difficult trade-offs, relationship management, ethical decisions, ambiguous problems, and strategic judgment.
The division is rarely absolute.
AI might prepare a pricing analysis while a commercial manager decides whether the recommendation fits customer relationships and competitive strategy.
An AI system could screen customer-support conversations, while an employee handles situations involving anger, exceptions, or reputational risk.
This creates a hybrid job architecture.
The objective is not maximizing the percentage of work assigned to machines.
It is allocating each activity to whichever combination of human and AI capability produces the best outcome.
Redefine Roles Around Higher-Value Human Work
Automation should ideally remove low-value friction rather than simply make people work faster.
Consider a manager who spends hours each week preparing reports, updating systems, scheduling meetings, and consolidating performance information.
AI can potentially reduce much of that administrative work.
Deloitte’s 2025 Human Capital Trends analysis notes that AI can free managers from administrative activity so they can spend more time coaching employees, solving strategic problems, developing talent, and managing important relationships.
That is genuine job redesign.
The manager is not simply doing the same job faster.
The job itself becomes different.
Companies should ask what employees could do with the capacity automation creates.
Perhaps analysts spend less time gathering data and more time advising decision-makers.
Recruiters spend less time scheduling and more time assessing talent.
Customer-service teams handle fewer basic inquiries and more complex customer problems.
The value shifts from execution toward interpretation and judgement.
Design Clear Human-AI Handoffs
Hybrid workflows require clear handoffs.
Employees need to know when AI is allowed to act independently and when human review is mandatory.
Imagine an insurance company using AI to evaluate claims.
The system might automatically approve straightforward cases that meet predefined criteria.
More unusual claims could be routed to human specialists.
High-value or potentially fraudulent claims might require additional review.
Without these rules, organizations risk two problems.
The first is under-automation, where employees manually check everything and AI produces little value.
The second is over-automation, where technology makes decisions that should involve human accountability.
Companies therefore need escalation points.
AI handles routine activity.
Humans intervene when uncertainty, financial exposure, ethical considerations, or customer consequences cross predefined thresholds.
This creates controlled autonomy rather than uncontrolled automation.
Rewrite Decision Rights Alongside Job Descriptions
AI changes more than who performs tasks.
It can also change who makes decisions.
When employees can access advanced analysis instantly, decisions that previously moved upward through several management layers may potentially happen closer to customers or operations.
This requires companies to rethink decision rights.
Who is allowed to act on an AI recommendation?
Who can override it?
Which decisions need executive approval?
Who is responsible when AI-supported decisions go wrong?
Deloitte’s 2026 Human Capital Trends research argues that organizations intentionally redesigning roles, workflows, and decision-making around human-AI collaboration are more likely to outperform expectations from their technology investments.
This is why updating job descriptions alone is not enough.
Governance needs to change with the work.
Broader Roles May Replace Narrow Job Boundaries
AI can make employees capable of performing tasks that previously belonged to other functions.
OpenAI research analyzing more than 800,000 work-related messages found that 43.5% of occupation-specific AI usage involved tasks associated with another occupation.
A salesperson might analyze data, while a marketer might troubleshoot a website without waiting for a specialist.
That can reshape traditional jobs.
Roles may become broader because AI lowers the technical barrier to adjacent work.
A marketing manager might perform basic analytics independently.
A finance professional could automate simple programming tasks.
An entrepreneur might handle marketing, financial analysis, and document preparation with a much smaller team.
Organizations therefore need to reconsider rigid functional boundaries.
Specialists will still matter, particularly for difficult or high-risk work.
But some routine handoffs between departments may become unneccessary.
Reskill Employees for AI-Augmented Roles
Job redesign works only when employees can operate effectively inside the new system.
That requires new capabilities.
Technical AI literacy is one.
Employees need to understand what AI can do, how to provide effective instructions, and how to verify outputs.
But human capabilities are equally important.
Critical thinking, problem solving, change management, communication, creativity, and adaptability become more valuable when routine tasks are automated.
Deloitte’s workforce redesign work has highlighted these capabilities alongside AI-assisted workflows, particularly as workers shift toward more strategic and forward-looking responsibilities.
The World Economic Forum reported that 77% of employers expected to upskill workers in response to AI-related change, while almost half expected to redeploy employees from AI-disrupted roles into other parts of their organizations.
Training therefore needs to accompany automation rather than arrive afterward.
Redesign Entry-Level Work Carefully
Junior jobs deserve particular attention.
Many entry-level roles historically involved repetitive work that doubled as training.
New consultants prepared basic analyses.
Junior developers handled simpler programming.
Young accountants reconciled information manually.
AI can now perform portions of this work.
That creates a paradox.
Removing repetitive tasks can make jobs more interesting, but it can also remove the experiences through which beginners learn professional fundamentals.
Companies need to redesign early-career development intentionally.
A junior employee might spend less time producing basic outputs but receive earlier exposure to reviewing AI work, attending customer discussions, solving problems, and making recommendations.
The training pathway becomes different.
Organizations cannot assume employees will somehow develop senior judgment after eliminating every junior task that previously helped create it.
Measure Outcomes Instead of Automation Rates
Companies sometimes become fascinated by automation percentages.
“We automated 40% of this workflow” sounds impressive.
But it says little about whether the business actually improved.
A better measurement system looks at outcomes.
Did cycle time decrease?
Did error rates improve?
Are customers happier?
Did employee capacity increase?
Did revenue grow or operating costs fall?
McKinsey’s finding that individual AI productivity gains are significantly more widespread than measurable EBIT impact illustrates why this distinction matters.
Automation is not the goal.
Business performance is.
A workflow where AI completes only 20% of tasks but dramatically improves quality may be better than one where 80% is automated but customers receive worse service.
Agentic AI Will Push Job Redesign Further
The next phase may be even more significant because AI is moving from answering prompts toward completing longer workflows.
OpenAI reported in June 2026 that AI agents are increasingly being used for tasks estimated to require more than an hour of human work, with users moving from isolated interactions toward delegated, longer-horizon assignments.
That changes job design again.
Employees may increasingly supervise several AI agents rather than use one AI tool for individual tasks.
A project manager could delegate market research to one agent, data analysis to another, and draft preparation to a third.
Their role becomes orchestration.
The valuable skill shifts toward breaking complex goals into tasks, assigning them intelligently, checking results, managing exceptions, and integrating outputs into a final decision.
Companies may eventually design roles around the amount of human-plus-machine capacity a person can coordinate rather than simply the work they personally produce.
Protect the Human Experience of Work
There is also a human risk to badly designed automation.
If companies use AI only to increase output targets, employees may become busier rather than more productive.
Deloitte warns that poorly integrated AI can create heavier workloads, reduced autonomy, and greater isolation rather than creating more meaningful work.
Good redesign should therefore consider employee experience.
Does AI remove frustrating work?
Does it improve autonomy?
Can employees focus on more meaningful problems?
Do they understand why responsibilities are changing?
Are performance expectations being adjusted realistically?
Companies need a human-AI value proposition, not merely an automation plan.
Technology creates sustainable value when employees understand how it improves both business outcomes and the quality of work.
Companies cannot capture the full value of AI simply by placing new tools inside old jobs.
They need to redesign work itself.
That means mapping workflows, separating human and machine activities, defining clear handoffs, changing decision rights, broadening some roles, reskilling employees, and protecting the development pathways of junior workers.
As agentic AI becomes more capable, professionals may increasingly move from performing every step themselves toward supervising and orchestrating automated work.
The best organizations will not ask how many employees or tasks they can replace.
They will ask how humans and AI can create better outcomes together.
If your company is introducing AI, start with the workflow rather than the technology. Identify where friction exists, decide where human judgment matters most, and redesign roles around the new capabilites available.
That is how AI adoption becomes real organizational transformation.
