How AI Is Changing Work

Learning Objectives

• Explain how AI automation differs from previous waves of technological displacement
• Identify the characteristics that make a task or role vulnerable to AI automation
• Articulate why augmentation — not wholesale replacement — is the most likely near-term outcome for most knowledge workers

The Shift That Is Already Underway

We are not waiting for AI to change work. It is already happening.

Software developers are using AI coding assistants that write, review, and debug code. Lawyers are using AI tools that review contracts in minutes instead of hours. Marketers are generating campaign copy and social content with a prompt. Customer service teams are deploying AI agents that handle the majority of tickets before a human sees them.

The examples are no longer hypothetical. Klarna's CEO announced in 2024 that their AI assistant was doing the work of 700 full-time customer service agents — the company reduced headcount from roughly 5,000 to 3,800 through attrition, explicitly crediting AI. Duolingo replaced contract content creators with AI in early 2025. Chegg, the education company, saw its stock collapse roughly 99% from its peak after CEO Dan Rosensweig directly attributed declining subscriber growth to ChatGPT.

The question is not whether AI will change your field. It is already changing most fields. The useful question is how — and what that means for you personally.

What AI Is Actually Good At

To understand which roles AI affects most, you need to understand what AI is actually good at right now.

AI excels at tasks that are:

• Pattern-based: Recognizing patterns in large datasets, text, images, audio — tasks where the answer can be inferred from examples
• Well-defined: Tasks with clear inputs and clear success criteria
• High-volume: Repeating the same type of task thousands or millions of times at low marginal cost
• Language-based: Generating, summarizing, translating, classifying, or extracting text

AI currently struggles with tasks that require:

• Novel physical dexterity: Unstructured physical manipulation in unpredictable environments
• Deep contextual judgment: Decisions involving complex ethical trade-offs, power dynamics, or unstated human needs
• Long-horizon accountability: Work where the consequences unfold over months or years and require sustained relationship management
• Genuine creativity: Truly original conceptual leaps (as opposed to recombining existing patterns — which AI does well)

Why This Wave Is Different

Previous waves of automation — factories, spreadsheets, computer-aided design — primarily affected manual labor or narrow, clearly defined computational tasks. A robot could weld a car frame; it could not write the press release about the car launch.

The difference with modern AI: It affects knowledge work directly.

Large language models can draft, summarize, translate, reason, code, and analyze. For the first time, tasks that required a university education and years of professional experience can be partially automated. This is new.

The scale of potential impact has been noted by major institutions:

• Goldman Sachs (2023): Estimated 300 million jobs globally could be "exposed to automation" — meaning AI could perform a significant share of their tasks, not that all 300 million jobs disappear. Follow-up research in 2024 shifted emphasis toward productivity gains, estimating AI could boost global GDP by 7% over a decade (roughly $7 trillion)
• WEF Future of Jobs Report (2025): Projects 170 million new jobs created and 92 million displaced by 2030 — a net gain of 78 million roles globally, though 59% of the workforce will need reskilling or upskilling. The fastest-declining roles include cashiers, administrative assistants, data entry clerks, and bank tellers
• McKinsey Global Institute: Estimates AI overall could deliver $13 trillion in additional global economic activity by 2030 (2018 estimate), with generative AI specifically adding $2.6–4.4 trillion annually on top of that (2023 estimate). By 2024, 72% of organizations had adopted AI in at least one business function
• IMF (2024): Estimates 40% of global jobs are exposed to AI; in advanced economies, the figure rises to 60%. About half of those exposed jobs may benefit from AI integration, while the other half could see reduced demand
• OECD (2024): Found that 27% of jobs across OECD countries are in occupations at high risk of AI automation — but emphasized that actual job losses have so far been limited, with the bigger near-term effect being task transformation

The Automation Risk Spectrum

Not all roles face equal risk. Think of it as a spectrum:

High automation risk — roles where AI can perform the majority of the core tasks with high accuracy:

• Data entry and data processing roles
• Basic customer service (FAQ-type queries, ticket routing, form processing)
• Standard legal document review (contract redlining, discovery document review)
• Medical transcription and clinical documentation
• Basic financial analysis (variance reports, standard ratio analysis)
• SEO content writing targeting keyword density rather than insight

Being transformed/augmented — roles where AI handles a significant portion of the work, but human judgment, relationship, and creativity remain central:

• Software engineering (AI writes code; humans design systems, make architectural decisions, debug complex issues)
• Medicine (AI assists with diagnosis, documentation, triage; physicians maintain clinical judgment and patient relationships)
• Law (AI handles research and drafting; lawyers exercise judgment, strategy, and advocacy)
• Marketing (AI generates content at scale; marketers set strategy, manage brand voice, build relationships)
• Education (AI personalizes learning, handles grading; teachers facilitate human development)

Growing because of AI — roles that did not exist a few years ago or are dramatically expanding:

• AI Engineer (building and deploying AI systems in production)
• AI Agent Developer (building multi-step autonomous agent systems — one of the hottest roles in 2025-2026)
• AI Product Manager (bridging technical AI capabilities and business needs)
• AI Safety / Evaluation Specialist (alignment, red-teaming, benchmarking — driven by EU AI Act and corporate risk management)
• Chief AI Officer (CAIO) — a C-suite role adopted by 20-25% of Fortune 500 companies by early 2025, up from under 5% in 2022
• AI Integration Specialist (deploying AI tools into existing enterprise workflows)
• LLM Ops / MLOps Engineer (deployment, monitoring, cost optimization for AI systems)
• AI Trainer / RLHF Specialist (providing expert feedback to improve model outputs — Scale AI, Outlier, and internal AI labs are major employers)
• Prompt Engineer (evolving from a standalone role into an integrated skill expected of many technical positions; specialized prompt engineering for enterprise deployments and red-teaming remains in demand)

Resilient Role Characteristics

Some characteristics make a role more resilient to automation. If your role has several of these, you are in a stronger position:

Human judgment in high-stakes decisions: Courts, medical diagnoses, hiring decisions, investment calls — society is still not comfortable with AI making these decisions without human accountability. This creates durable demand for humans.

Physical presence in unpredictable environments: Plumbers, electricians, nurses, surgeons, childcare workers — physical manipulation in variable environments remains difficult to automate.

Emotional labor and relationship-building: Therapists, coaches, teachers, salespeople with deep client relationships, social workers — the human connection itself is often the service.

Creativity and original conceptual work: Not all creative work is equally safe, but original conceptual thinking, aesthetic judgment, and brand identity work at the highest level remains human-driven.

Interdisciplinary synthesis across contexts: Connecting insights from law + technology + human behavior + organizational dynamics — this kind of broad contextual integration remains a human strength.

The Likely Outcome: Augmented Professionals

The realistic near-term outcome for most knowledge workers is not replacement — it is a new kind of role where being effective requires AI fluency.

The lawyer who uses Harvey AI to do research in 20 minutes instead of 4 hours does not lose their job. They become a more productive lawyer who can serve more clients, work on more complex matters, and bill for judgment rather than research time. Harvey AI now serves roughly 100,000 lawyers across 50 of the top 100 Am Law firms — and the lawyers who refuse to engage with these tools increasingly look uncompetitive against those who do.

This creates a two-tier dynamic: AI-fluent professionals and AI-avoidant professionals — and the gap between them will widen over time.

Key Takeaways

• AI is transforming knowledge work now — the shift is already underway, not hypothetical
• AI automation targets pattern-based, well-defined, high-volume, language-based tasks first
• Most near-term change will be augmentation (AI + human) rather than wholesale replacement
• Roles with human judgment, physical presence, emotional labor, and genuine creativity are more resilient
• The emerging career risk is not "AI takes my job" but "a human using AI takes my job"