Two years ago, "AI" for most people meant a chatbot: you type, it replies. In 2026, the conversation has moved to AI agents — AI that doesn't just answer, but takes action and finishes tasks on its own. "Agentic AI" is everywhere: startup decks, company meetings, your feed. But behind the hype, some of it is real and some isn't proven yet. Here's the breakdown.

Three things to know

  • An agent = AI that acts, not just answers. You give it a goal; it plans the steps, uses tools (a browser, email, code, APIs), and executes until it's done.
  • It's officially the top trend. Research firm Gartner named agentic AI the number-one strategic technology trend for 2025.
  • The hype is ahead of the evidence — for now. Many "agents" are just chatbots with a new label, and a chunk of projects are forecast to fail. Healthy skepticism applies.

So what is an "AI agent"?

Think of a chatbot as an advisor: you ask, it answers, you do the work. An AI agent is more like an assistant you hand a task: "find these, compare them, then lay it out in a table" — and it breaks that into steps, uses the tools it needs, checks the result, and corrects itself when something misses. As IBM puts it, the key is the ability to plan and act toward a goal, not just reply to a single message.

Why did it explode in 2026?

A few things matured at once:

  • Language models got better at step-by-step "reasoning", making them more dependable on multi-step tasks.
  • "Tool use" became standard — AI is now routinely wired into real apps (calendars, databases, browsers), not just a chat box.
  • Big vendors (OpenAI, Anthropic, Google, and others) shipped agent features through 2024–2025, so the idea spread fast.

The numbers people quote (and their status)

  • Estimate: Gartner projects that by 2028 around 33% of enterprise software applications will include agentic AI, up from under 1% in 2024.
  • Estimate: Gartner also expects that by 2028 at least 15% of day-to-day work decisions could be made autonomously through agentic AI.
  • Warning: on the other side, Gartner predicts more than 40% of agentic AI projects will be scrapped by the end of 2027 over cost, unclear value, or risk.

These are projections, not things that have already happened — the numbers can change. Use them as a sense of direction, not certainty.

What's still debated

  • "Agent washing". Plenty of products get the "agent" label when they're really just automation or a plain chatbot. Check: does it actually act on its own, or just answer?
  • Reliability. More steps means more chances to go wrong. For high-stakes tasks (money, sensitive data), human oversight is still a must.
  • Security and accountability. If an agent gets access to your email or bank, one mistake can be expensive. Who's responsible when it errs?

Why this matters to you

Because "AI agent" is about to get used to sell you things. If you understand the difference between a real agent and a marketing label, you won't fall for it — and you can judge a tool by real results: did the task finish faster, or did it just look impressive? (One local example: Jakarta studio Codiosity is building AI agents you can rent or buy — Codiosity sponsors Respawn Society.)

Quick recall

Without looking up: in one sentence, what's the difference between a chatbot and an AI agent? And why does "agent washing" mean you should be careful?

Try it out

Find one ad for a product that claims to be an "AI agent." What three questions would you ask to check whether it's a real agent or just a rebranded chatbot?

Still curious?

  • How is an AI agent different from ordinary automation or scripts?
  • Which tasks are safe to hand to an agent today, and which aren't?
  • What safeguards make an agent worth giving access to important data?
  • Why are so many agentic AI projects predicted to fail?

Sources

  • Gartner Newsroom — "Gartner Identifies the Top 10 Strategic Technology Trends for 2025" (agentic AI as the number-one trend) and its agentic-AI adoption projections for 2028. Figures are predictions.
  • IBM — "What are AI agents?" — an agent's definition: planning and acting toward a goal, not just replying to a message.