
Introduction: Generative AI and the Future of Work
Generative AI has moved beyond drafting shortcuts. It is reshaping workflows, decision-making, roles, and what employees expect from their careers. Many workers are asking a blunt question: do you still need me?
NIST defines generative AI as models that emulate the structure of input data to produce new content such as text, images, and audio. In practice, it shows up in three ways:
- A tool that supports work, such as a drafting assistant
- A system that automates defined tasks, such as routing requests
- An agent that runs multi-step processes with limited human input
What shapes the future of work is how leaders redesign roles, assign responsibility, build skills, and keep people essential to the mission.
Key Takeaways
- Generative AI changes tasks and skills within jobs more often than it erases whole occupations.
- Judgment, empathy, and accountability matter more as AI handles execution.
- Workflow redesign, training, and manager support decide whether adoption works.
- Measure trust, workload, and quality, not just tool usage.
How Generative AI Is Changing the Way We Work
From task automation to human-AI collaboration
Generative AI drafts, summarizes, translates, writes code, retrieves information, and generates ideas. Outputs still need context and review.
"AI at work" is not one experience. The ILO distinguishes augmentation, which combines AI with human intelligence, from automation, which replaces tasks suited to full automation. Agentic systems add a third layer: autonomous decisions inside a workflow.
The same technology plays out differently by setting:
- Insurance claims: CorVel announced AI summaries of medical documents to support adjusters and nurses, with human oversight.
- HR: IBM's AskHR classifies and routes questions, while human advisers handle complex cases.
- Healthcare administration: In a UC San Diego trial, AI-drafted patient replies raised physician reading time by 21.8%. Reply time fell 5.9%, which was not statistically significant.
- Customer service: A study of 5,172 agents found AI assistance raised issues resolved per hour by 15%, with larger gains for less-experienced workers.
Results are task-specific. They are not organization-wide savings.
Redesigning workflows instead of simply adding tools
Buying a tool is not redesign. Map the workflow end to end:
- List every step.
- Mark what AI performs, what needs human judgment, and where review or escalation occurs.
- Test the map with the people who do the work.
Work shifts from producing everything from scratch to directing, curating, and validating. That has a cost. A 2025 CHI study of 319 knowledge workers found people moved from gathering information toward verification, and higher confidence in AI went with less critical-thinking effort.
Conditions matter too. In a Harvard/BCG experiment with 758 consultants, AI users finished 12.2% more tasks within the model's capability. On a task outside it, they were 19% less likely to get a correct answer. Replacing repetitive work can also bury people in a flood of higher-order decisions.

Clarifying responsibility in AI-enabled work
Every AI-supported process needs a named owner who reviews outputs, catches errors, handles exceptions, and makes the final call.
Consider an AI system that recommends denying a claim. A qualified adjuster checks the recommendation against policy language, the customer's circumstances, and professional standards. She overrides it when the file shows something the model missed. The AI proposed; a person decided and answered for it.
How Jobs and Roles May Evolve
Why tasks, not job titles, are the right starting point
A job is a bundle of tasks that serve outcomes. Automating one task does not erase the role.
Ask these questions of each task:
- How repetitive is it, and is the data available?
- How ambiguous is it?
- What happens if the AI is wrong?
- Does it depend on a relationship?
- Does it require contextual judgment?
The ILO estimates 24% of global employment has some generative AI exposure, but only 3.3% falls into the highest-exposure category. Most occupations keep tasks that need human input, so ILO's exposure index points toward transformation more than replacement. Exposure is not a layoff forecast.
Which responsibilities may become more valuable
Employees may spend less time processing routine information and more on interpretation, problem framing, exception handling, relationships, and quality control.
New duties also appear inside existing jobs:
- Validating outputs
- Documenting decisions
- Monitoring risk
- Explaining AI-supported recommendations
The World Economic Forum proposes two useful roles:
- AI work architect: clarifies the problem and success criteria, splits work into delegated, augmented, and human-led pieces, and sets decision rights and escalation paths.
- AI steward: validates outputs against domain knowledge and frontline context, and decides whether to accept, modify, reject, stop, or escalate.
These are responsibilities, not standard job titles. Many organizations will embed them in current roles.
Avoiding false certainty about job replacement
Some tasks will shrink, some roles will be redesigned, and some new work will emerge. Effects will land unevenly across workers and industries.
Productivity gains do not guarantee better jobs. A 2025 randomized study of 7,137 knowledge workers found less email time and less after-hours work, yet tool access alone did not change task quantity. Outcomes depend on whether organizations reinvest that time in learning, service, and development.
Leaders should say what is known, what is uncertain, and how affected employees will be supported. Citi's decision to train 175,000 employees to work with AI before roles change is one visible attempt to do that.
Human Skills, Reskilling, and Career Resilience
Skills that support durable human contribution
No skill or occupation is permanently AI-proof. Still, some work leans heavily on trust, accountability, and judgment. Five broad categories look more resilient:
- Relationship and care work that depends on trust and empathy
- High-stakes judgment where someone must be accountable
- Physical, hands-on trades that adapt to unpredictable settings
- Leadership and coaching that guide people through change
- Domain expertise paired with AI oversight, including spotting exceptions that data misses
The International Labour Organization (ILO) finds lower exposure among craft and trades workers, plant and machine operators, and agricultural workers, partly because of physical tasks. Lower exposure is not immunity to organizational change.
Whatever the exposure level, employers still hire for skills that travel with people. The World Economic Forum (WEF) 2025 skills outlook says analytical thinking is essential to 7 in 10 employers, with empathy, active listening, and lifelong learning among core skills.
Building an effective reskilling approach
Pair AI literacy with role-specific knowledge, process understanding, risk awareness, and practice on real work scenarios. Then back it with time and people:
- Protected learning time
- Coaching and feedback
- Safe space to experiment
- Manager support
One-time tool demos fall short. Microsoft's 2024 survey of knowledge workers found only 39% of AI users reported employer-provided AI training.
BCG's 2025 survey linked regular use to at least five hours of training and in-person coaching, though that association does not prove causation.
Moderna offers an applied example. It paired its AI Academy with an internal GenAI tool and functional workflows, and the company reports nearly 65% active use. That is company-reported adoption, not proof of career-mobility gains.

How Leaders Can Prepare Organizations for AI Adoption
Build a people roadmap alongside the technology roadmap
Start with the business problem and the employee impact. Map affected workflows, roles, and skills, and name likely sources of uncertainty before choosing or expanding a tool.
Involve frontline staff and subject-matter experts. They know the exceptions, customer needs, and unintended consequences.
Communication should cover:
- Why AI is being introduced
- What will change and what stays human-led
- How performance will be evaluated
- Where people can raise concerns
Give managers clarity, ownership, and practical support
Middle managers carry adoption into daily work. They translate decisions, field questions, and absorb what the plan misses. They need:
- Clear decision rights
- Language for job-security conversations
- Guidance on delegation and review
- Time to coach
McKinsey's analysis names managers' judgment, empathy, and work-redesign skills as central to adoption. Treat manager enablement as an operating requirement, not an optional memo.
Christina Muller's Carrying the Rollout™ workshops, offered as half-day or full-day sessions, focus on ownership, delegation, responsibilities, and decision rights during AI change.
Establish measures and feedback loops
Track more than logins. Measure these areas:
| Area | What to watch |
|---|---|
| Quality | Accuracy, errors, subgroup performance |
| Workload | Review time, after-hours work |
| Confidence and trust | Confidence, verification behavior |
| Learning and retention | Skill progress, attrition |
| Feedback | User reports, near misses, incident reviews |
NIST's voluntary GenAI Profile is a useful governance reference. Skip invented benchmarks.
The COUNTED at work™ Score, a free five-minute diagnostic, examines leader clarity, ownership, and perceived support during AI adoption. Use it as a baseline, alongside pilots, employee listening, and regular workflow adjustments, before scaling.

Ethical and Human Considerations
NIST's GenAI Profile documents the main risks: confident false outputs, harmful bias, privacy leakage, and overreliance. Others include unequal access to training, unclear accountability, surveillance, and automating unfair existing practices.
Trust depends on transparency: where AI is used, who reviews outputs, how people can challenge decisions, and how errors get fixed.
High-stakes settings need special care, and the rules vary:
- Hiring: The EEOC says existing anti-discrimination duties apply to AI in employment decisions. NYC Local Law 144 requires bias audits for covered tools.
- Healthcare: HIPAA safeguards and business associate agreements apply to cloud providers handling patient data.
- Insurance: The NAIC model bulletin calls for written AI governance and appropriate human involvement.
- Lending: Regulation B still requires specific reasons for adverse action.
State law is moving. Colorado enacted replacement legislation in 2026, and federal policy continues to shift. Check current status with counsel.
Rules alone do not protect people at work. How leaders handle efficiency gains, role changes, and pushback shapes whether AI adoption builds trust or erodes it.
Protect wellbeing during adoption:
- Avoid turning efficiency gains into constant overload
- Don't remove meaningful work without a transition plan
- Treat resistance as a signal to investigate, not a character flaw
A Berkeley Haas study of a roughly 200-person company described task expansion and work spilling into evenings, though the research was qualitative. That pattern is worth watching as tools speed individual output.
Frequently Asked Questions
Can AI replace jobs in the future?
AI will automate some tasks, reshape many roles, and create new responsibilities. How much a job changes depends on the work, the industry, human oversight, and organizational choices.
What 5 jobs will survive AI?
No occupation is guaranteed to stay unchanged. Five broad categories look more resilient: care and relationship work, high-stakes judgment roles, skilled trades, leadership and coaching, and domain experts who oversee AI.
Where will AI be in 5 years?
Expect no single outcome. Likely directions include deeper workflow integration, more capable assistants and agents, shifting skill requirements, and stronger expectations for governance and human oversight.
How should organizations prepare employees for generative AI?
Offer role-specific AI literacy, clear expectations, and protected learning time. Add manager support, employee involvement, and honest communication about changing work.
What skills will be most valuable in an AI-powered workplace?
Critical thinking, domain expertise, communication, empathy, and ethical reasoning stand out. So do creativity, adaptability, and the ability to evaluate and responsibly apply AI outputs.