Over the past month, I’ve analyzed 200 trend reports for 2025 using AI. The one trend that stood out across the board? AI agents. Every day I come across ambitious plans, bold promises, and major predictions in the news. But as a professional, can you actually do something with them right now? In this article, I dive into that question.

Automation is nothing new—we’ve been doing it for centuries. Think of the first windmills in the Middle Ages, allowing farmers to grind grain without manual labor. Or the Industrial Revolution, when machines like the loom and the steam locomotive drastically sped up production and transport. Automation has always been about one thing: making repetitive tasks smarter and more efficient.

Today, AI agents are taking this to the next level—not just replacing physical tasks, but also cognitive ones like responding to emails, analyzing data, and even supporting creative processes. Unlike traditional software that always needs human input, AI agents can interpret information and learn from it.

“The IT department of every company will become the HR department for AI agents.” – Jen-Hsun Huang, CEO of Nvidia

More than that, they can execute tasks independently, make decisions, and even negotiate with other agents on our behalf. And unlike traditional technological transformations that required years of infrastructure building, these AI agents are relatively easy to build and implement.

The Predictions

And the forecasts are striking. According to McKinsey, AI agents could automate 60 to 70 percent of employee time across many industries. Globally, Goldman Sachs predicts agents could boost GDP by 7%—or $7 trillion. Deloitte says that by next year, half of all companies will be using AI agents, with a quarter already using them this year. It’s expected to be the killer app of AI. Gartner projects that by 2028, at least 15% of daily business decisions will be made autonomously by agentic AI.

Your next colleague? It might very well be an AI agent. Research suggests that agents can collaborate in ways far beyond human capabilities. As one researcher put it:

“It’s incredibly promising that they can bring together different viewpoints and reach consensus far faster than we can—and with more diverse perspectives.”

“This isn’t just an evolution of technology. It’s a revolution that will fundamentally redefine how humans work, live, and connect with one another from this point forward.” – Marc Benioff, CEO of Salesforce

But let’s be real: Most bold predictions come from parties who stand to benefit—consultants and AI vendors. These are the same folks who, back in 2020, said crypto would replace all banks by 2025, that all art would be NFTs, and that we’d be working full-time in the metaverse by now.

Less Talk, More Action

For 2025, I’m hearing a growing sense of pragmatism. Last week during an AI training, someone in the audience said:
“I read a lot about new features—like context windows, multilingualism, and resonance—but I keep missing the ‘why.’ It seems like we, as professionals, have to figure out how to use these tools… but shouldn’t it be the other way around?”

A great observation. We talk too much about the tech, and not enough about real, practical value. In my view, we need to talk less about wires and more about actions.

We also need to ask: What do we really want AI to take over? I’m genuinely enthusiastic about this technology—I use different tools for 1–2 hours every workday. It makes me smarter, better, more efficient, and even makes work more fun. But I’m also becoming more critical of how these tools function. For instance, AI agents can book an entire vacation for me. I’ve used AI to plan three trips in the past year.

But the results still had flaws—like suggesting places that didn’t exist. Would you trust an agent with your credit card to book your flight and hotel? Plus, I actually enjoy planning my own trips—it sparks new ideas and inspiration.

Are We Really Ready?

The same goes for writing. Sure, I use ChatGPT to check my tone and style—makes the editors at Frankwatching happy. And yes, people ask if my latest book was written with AI (it wasn’t—I spent three full years writing it myself). I could have written it in a day using one of the 320 tools that auto-generate books.

But would I want that? Tech giants like Microsoft and TikTok now run AI-powered publishing houses. But is there really demand for AI-generated books, music, or videos—stuff made without effort, soul, or passion?

I can see how fast AI is evolving—just look at how good tools like GPT and Midjourney have become in under two years. Research shows hallucinations in large models have dropped from 10–15% to nearly 1%. That risk will likely be negligible soon.

In his 1888 novel Looking Backward, Edward Bellamy imagined a future where art and literature flourish once automation frees people from menial work. Ironically, the opposite seems to be happening now.

The Mainstream Moment

According to OpenAI, 2025 will be the year agents go mainstream. Every major player is launching agent products—OpenAI, Microsoft, Google, Salesforce, Anthropic, Meta. Just like AI quietly entered our lives long before GPT, agents are already embedded in countless systems—from ING and Bol.com’s customer service, to Fitbits, robot vacuums, and spam filters.

Recruiting giant Adecco processes over 300 million job applications annually—and could only respond to less than 5%. With AI agents, they now qualify candidates automatically and can respond to everyone within 24 hours. Google published a list of over 300 agent use cases across industries.

Crypto AI Agents

Another booming field for AI agents: crypto. I’ve been in Bitcoin since 2013, mainly advising startups and building companies—not trading (I’m terrible at it). Most people lose money in crypto because of emotional decision-making. But crypto AI agents like Virtuals now fully automate trading—analyzing trends, making decisions—up to 10,000 times per day. Just set it up and yes, theoretically, make money in your sleep (if the agent is well-configured).

I also love the quirky agent experiments people are doing. One agency lets ultra-realistic virtual influencers create content for OnlyFans—and makes a fortune. Another, Altera, ran a wild experiment: they unleashed 1,000 autonomous AI agents on a Minecraft server.

Led by ex-MIT professor Robert (Guangyu) Yang, “Project Sid” explored if these agents could collaborate better than humans. The result? The agents built a trade hub, voted on a constitution using Google Docs, spread a religion (Pastafarianism!) through bribery, and even helped a lost villager find his way home using torches.

“Agents will eat the world.”

Opportunities… and Risks

As with any major shift, the rise of agents brings clear challenges and understandable fears. The biggest fear I hear? Our jobs. In five years, we might all be on permanent vacation thanks to agents.

But let’s zoom out. History shows how new tech transforms industries—think planes, satellites, the internet, smartphones, green energy. While jobs may disappear, new opportunities always emerge.

Look at the U.S.: In 1950, 43 million people worked. By 2020, it was over 152 million. That’s 100+ million new jobs—mostly in fields that didn’t exist in 1950.

One reason I’m so excited about AI: Everywhere I look, staff shortages are a growing issue. In some regions and sectors, the workforce is stagnating or shrinking. At university, I learned that productivity—not headcount—will drive future economic growth, especially in the service sector, which employs 80% of people in the Netherlands. AI agents will play a key role here.

Killer Apps? Or Killer Robots?

How risky this tech becomes depends on how we use it, what we want from it, and how well we understand it. AI agents can make poor decisions if their goals don’t align with users or organizations. This can lead to bad—and sometimes dangerous—outcomes. Sounds far-fetched? Ukraine is already experimenting with autonomous weapons—yes, weapons that decide whether to fire.

Agents can also behave unpredictably, with unintended consequences. Bias in AI is a growing problem, and if decisions (like hiring) are made based on that, it can go horribly wrong.

Another risk: over-reliance. People may stop thinking critically. I heard a powerful story recently from a hospital board I coach. On their oncology ward, they now use AI to diagnose cases incredibly quickly—cutting down the waiting time for patients drastically.

Amazing, I thought. A former colleague of mine had to wait two months for her next appointment during a tough breast cancer treatment. Now, new doctors are trained with AI. But senior oncologists warn that young doctors are losing the finesse needed to read mammograms—finesse that AI still lacks.

Start Using AI Agents Today

Technology isn’t inherently good or bad—it depends on how we use it. Without proper oversight and the right data, agents can make choices that conflict with human values—like prioritizing profit over safety or unintentionally discriminating.

I’ve built and tested a few agents myself. I started by identifying repetitive tasks I still do that I’d love to automate. One example is SEO—analyzing trends and data to find relevant keywords and then turning those into SEO-friendly blogs (with meta descriptions for Yoast, etc.). My SEOBot automates all of that in 50 languages.

I’m also building an app for my new AI startup—largely developed with AI. I’m amazed by how many tools now let you build high-quality apps with little effort: Databutton, Replit, and others.

Getting Started in a Few Steps

Want to experiment? Start with AgentGPT. Or try GenFuse—great for non-tech users. You can easily create your own agent. Amazon also launched Lex, which lets you build a conversational AI interface in just a few steps. I also see positive buzz in communities around Rasa OS. I’m a big fan of Botpress (though it gets expensive fast). Finally, check out Dify—it connects your agent to a wide range of other powerful AI tools.