Agentic AI Is Reshaping Jobs Across Multiple Sectors

Fears that AI will simply replace workers are overstated, according to AI transformation researcher Einav Peretz Andersson at Jönköping University. Instead, she argues, roles will evolve as AI improves productivity and quality. The biggest shift comes from agentic AI, systems that can act toward goals rather than only answer questions.

Why this AI researcher says jobs will change, not vanish

Andersson’s core point reflects what many organizations are seeing in early deployments of generative AI: the technology tends to unbundle jobs into tasks. Routine, repeatable tasks are automated first, while human work shifts toward oversight, judgment, relationship-building, and domain-specific decision-making.

In practice, that means fewer “before-and-after” job eliminations and more redefinition of responsibilities. A marketer may spend less time drafting first versions of copy and more time on brand strategy and performance analysis. A customer support agent may handle fewer basic queries and more complex cases that require empathy, negotiation, or policy interpretation.

This transition is also driven by the economics of AI adoption. Many companies are not trying to remove entire functions; they are trying to increase throughput, reduce cycle time, and raise quality. That aligns with Andersson’s view that AI can make work “more effective, more productive” and improve output quality.

Agentic AI: from chatbots to systems that take action

The source highlights agentic AI as a key inflection point. Unlike a traditional chatbot that waits for prompts, an agentic system can execute multi-step workflows toward a goal, make intermediate decisions, and adapt when conditions change.

In enterprise settings, this often looks like:

• Planning: breaking a goal into tasks and sub-tasks
• Tool use: calling APIs, searching internal knowledge bases, writing code, or updating tickets
• Monitoring: checking results, handling errors, and iterating
• Handoffs: escalating to humans when confidence is low or approvals are required

This matters because the productivity gains from AI are typically limited when the model only generates text. The impact grows when AI can connect to business systems (CRM, ERP, ticketing, analytics) and complete work end-to-end. That is also where risk increases, since an autonomous action can create real-world consequences.

Industries most exposed to near-term workflow change

While the source does not list specific sectors in text, the pattern of disruption is consistent across industries where work is information-heavy, process-driven, and measurable. The following areas are among the most likely to change quickly as agentic AI matures:

• Customer service and contact centers: AI can draft responses, summarize conversations, route cases, and propose next-best actions. Humans increasingly handle escalations and sensitive scenarios.
• Marketing and communications: AI accelerates ideation, localization, A/B variant creation, and performance reporting. The role shifts toward governance, differentiation, and channel strategy.
• Software development and IT operations: coding assistants already speed up routine tasks; agentic tools can open pull requests, run tests, and propose fixes. Engineers focus more on architecture, security, and product decisions.
• Finance and accounting: AI can reconcile transactions, flag anomalies, and draft narratives for reports. Professionals concentrate on controls, compliance, and interpretation.
• HR and recruiting: AI can screen for skills, draft job descriptions, and schedule workflows. Human judgment remains critical for fairness, culture fit, and final decisions.
• Healthcare administration: documentation support, coding assistance, and scheduling optimization can reduce clerical burden, though clinical decisions require strict safeguards.
• Legal and compliance: AI can review contracts, extract clauses, and generate first drafts, while lawyers focus on negotiation, risk, and accountability.

What changes inside roles: skills, governance, and accountability

As AI becomes embedded in daily workflows, organizations will demand different skills. Prompting is only a small part of the shift. More valuable capabilities include:

• AI literacy: understanding model limits, hallucinations, and evaluation
• Data fluency: knowing what data is reliable and how to measure outcomes
• Process design: mapping workflows so AI can safely automate steps
• Human-in-the-loop oversight: defining approvals, escalation paths, and audit trails

Agentic AI also raises governance questions. If an AI system can act, companies need clear policies on permissions, logging, and accountability. This is where regulatory frameworks and standards become relevant. In Europe, the EU AI Act pushes risk-based controls, and many firms are adopting internal model-risk management practices similar to those used in finance.

Why this matters now for the AI industry

The shift from “AI that writes” to “AI that does” is a competitive battleground for major AI vendors and cloud platforms. Model providers are racing to pair large language models with tool-use, memory, and orchestration layers, while GPU-heavy infrastructure from companies like NVIDIA underpins the compute demands of training and inference.

For employers, the headline is not mass job disappearance but accelerating change management. The organizations that benefit most will be those that redesign processes, invest in training, and implement guardrails for autonomous actions. Andersson’s message is ultimately a pragmatic one: AI is less a replacement for people than a catalyst that reshapes how work gets done.