How AI is affecting the job market

Jack Virag

September 3, 2025

Table of Contents

AI isn’t taking everyone’s job—but it is rewriting job descriptions at scale.

Everyone feels something about AI's ubiquity in the workplace today. Some compare it to other paradigm-shifting inventions like the telephone, the internet, and mobile devices. Others see it as a plagiarism machine being used to displace real humans at scale.

While replacement use cases, e.g. "Instead of hiring freelance writers I just use ChatGPT," aren't abnormal, they're not necessarily the direction we're all heading.

Here are some stats and trends we surfaced to show exactly how AI is impacting the job market in 2025:

Adoption is broad, impact is concentrated

AI has shifted from experiment to enterprise. According to McKinsey, 78% of organizations now use AI in at least one function, and 71% use generative AI regularly. That’s a sharp rise from just two years ago when adoption was still exploratory.

But the impact isn’t evenly distributed. Certain functions have become hotspots for AI use because embedding AI in those workflows is relatively straightforward:

Marketing and sales: Personalization, content generation, and targeting.
Service operations: Customer support automation, chat assistants, workflow optimization.
Engineering and IT: Code generation, bug detection, knowledge base building.

Even with this widespread uptake, only 17% of companies report that generative AI has contributed at least 5% of their EBIT. The message is clear: adoption is high, but monetization is still catching up.

Jobs vs tasks—what exposure really means

The public debate often swings between “AI will take all our jobs” and “AI will only make us more productive.” The truth lies somewhere in between. Research from the IMF estimates that 40% of global jobs are exposed to AI, and in advanced economies that figure jumps closer to 60%.

It’s important to separate jobs from tasks. Exposure doesn’t always mean elimination:

Complementation: About half of exposed tasks are augmented by AI—think drafting, summarizing, or analyzing data that still requires human judgment.
Substitution: The other half risks automation, particularly routine cognitive or clerical work.

The imbalance creates a structural challenge. Higher-skilled workers tend to benefit as their output is scaled, while lower-skilled or routine roles face substitution pressures. Without deliberate intervention, this dynamic risks widening inequality.

What job postings tell us

Hiring data offers a leading indicator of how organizations think about skills. Mentions of generative AI in US job ads rose 170% year over year, but still account for only ~0.3% of total postings. That’s a small base, but the growth trajectory is steep.

The skill expectations are also changing. According to LinkedIn’s work change report, jobs listing “AI literacy” skills grew more than sixfold in the past year, though they remain rare—roughly one in 500 postings. What’s notable is the type of roles asking for these skills: not just data scientists or AI engineers, but consultants, marketers, and managers.

This signals a shift: companies aren’t only hiring people to build AI—they’re hiring people to implement AI. In other words, the differentiator isn’t technical fluency alone, but the ability to apply AI tools inside traditional workflows.

Productivity evidence, not just hype

For years, talk about AI’s productivity boost was mostly anecdotal. Now we have hard data. In a randomized trial with customer support agents, access to generative AI lifted productivity by 14% overall. The gains were even larger for less-experienced agents—25 to 35% improvements—because the AI diffused best practices from top performers.

In software development, the story is similar. A controlled experiment showed that developers using a Copilot-style tool completed tasks about 56% faster than those working solo.

The signal is consistent: AI narrows the gap between novices and experts. For employers, that means training and onboarding cycles can compress. For workers, it means your “floor” performance improves quickly—but the ceiling still depends on creativity and judgment.

Wages, premiums, and inequality

Workers who add AI skills to their toolkit are already seeing outsized returns. According to the PwC AI jobs barometer, those with AI skills earn about 56% more than peers in the same roles. Skill change is also happening 66% faster in AI-exposed jobs, a sign that these roles evolve more quickly.

But there’s a warning baked into the data. The IMF cautions that without targeted policy or employer-led programs, AI could widen income inequality in advanced economies. High-skill workers will reap the premiums, while lower-skill workers risk displacement.

For employers, the path forward is twofold: budget for premiums in roles that demand AI literacy, and build equitable training programs so the benefits don’t just concentrate at the top.

Where jobs are growing vs shrinking

The long-term outlook is more positive than the headlines suggest. By 2030, the World Economic Forum projects about 170 million jobs will be created and 92 million displaced—a net gain of 78 million roles.

Growth is strongest in areas like:

AI and machine learning specialists
Data analysts and scientists
Green transition roles (climate, energy)
Care economy roles (health, aging populations)

On the decline are clerical and routine jobs such as bank tellers, data entry clerks, and postal workers. These roles are heavily exposed to substitution risk and offer fewer opportunities for AI complement.

The skills mix is shifting fast: about 39% of today’s skills will be obsolete or fundamentally changed by 2030, and nearly 60% of workers will need some form of retraining. That makes workforce planning less about static roles and more about building durable, adaptable skills.

Regulation fuels compliance jobs

Regulation is quickly becoming a driver of AI-related hiring. In Europe, the EU AI Act entered into force in 2024 and begins phasing in obligations through 2027. By February 2025, prohibited uses and literacy requirements take effect. By August 2025, rules for general-purpose models apply. By 2026–2027, all high-risk systems must comply.

In the United States, the OMB M-24-10 memorandum requires federal agencies to inventory AI models, appoint a chief AI officer, and publish risk management practices. That’s creating demand for governance, compliance, and risk roles—not just in government, but in every vendor and contractor that touches federal data.

The takeaway: compliance itself is becoming a growth industry. Organizations need AI governance capacity whether or not they’re building models in-house.

What workers should do right now

AI literacy is no longer optional.

Individual workers can future-proof their roles by taking a structured sprint approach:

Spot opportunities: Identify three tasks in your role that could benefit from a copilot—anything repetitive, data-heavy, or communication-based.
Build credentials: Earn a recognized AI literacy certificate and immediately apply it to a weekly deliverable.
Show evidence: Create a small portfolio artifact that demonstrates measurable gains, such as a 20% time reduction or noticeable improvement in quality.

By making these steps visible, workers signal adaptability to employers and build resilience in an evolving labor market.

Going forward

AI is no longer a “future of work” conversation—it’s reshaping jobs right now. Organizations that experiment early and build literacy across the board will capture the upside faster and with less disruption.

If your go-to-market team wants to turn AI into real outcomes—from pipeline growth to customer retention—Tavus can help you scale personalized video where it matters most.