Written by Stephan Forseilles, CTO at Easyfairs, and Chair of the UFI Digital Innovation Working Group.

Let’s be honest—AI doesn’t wow us quite like it used to. Once we get used to chatting with ChatGPT or Claude, generating images with Midjourney, creating videos with Runway, or making deepfakes with HeyGen, we’re left wondering: “So, what’s next?”

Don’t get me wrong—these tools are incredibly useful. But since the peak of AI hype one or two years ago, we haven’t seen a truly life-changing breakthrough. We’re stuck in a loop: ask a question, get a result, rephrase it, get a better result, tweak the prompt again, get something slightly improved… then overdo it and end up with something unusable. Start over.

For those of us who use LLMs in our daily routine—summarizing documents, comparing contracts, extracting to-dos from meeting transcripts, or responding politely to emails when we’d rather send invectives—it’s all become… mundane. Like driving to get ice cream. Useful, yes. But exciting? Not really. LLMs generate solid content, but they don’t act for us.

Enter AI Agents

AI agents are autonomous, adaptable, goal-driven AI systems that work independently on our behalf. They move beyond the “conversational” model of LLMs and can actually do things.

Take customer service, for example. We’ve had chatbots on websites for years. But modern agents—think of them as chatbots on steroids—have vastly superior reasoning, contextual understanding, and even creativity. They can take meaningful actions, like booking hotels, resetting passwords, or making online purchases.

I recently used an AI agent called DoBrowser, a Chrome extension, to register for an event. I asked it to: “Register me for event XXX, but make up all data except this email address…” It filled in plausible fake details, clicked the “I’m not a robot” box, and completed the registration. Cute, effective—and a data quality nightmare.

DoBrowser functions as a personal assistant by taking over your browser. It “sees” the web page and interacts with it, much like a human (although technically, it reads the HTML directly for efficiency). Other tools like Adept.ai, Rewind.ai, and PixieBrix do similar things—automating repetitive tasks either in the browser or across your whole system (some can control your screen, keyboard, and mouse).

Simpler Agents: ChatGPT Tasks and Custom GPTs

OpenAI recently introduced ChatGPT Tasks, allowing you to schedule AI-driven tasks. For example, you can ask it to search the web daily for news on a specific topic and send you a digest—no more digging through newsletters or RSS feeds. Great for staying informed or finding blog inspiration (and yes, it can write those too… but that’s another story).

Agents can still be conversational—they just need to be specialized and goal-driven. This is what Custom GPTs are all about. You give ChatGPT a role and set of instructions (e.g., “You are a helpdesk agent for software XXX. Guide users using this documentation…”). It then acts as an expert, and can continuously improve based on interactions.

Other platforms offer similar custom agent capabilities are Anthropic’s Claude, Google’s Gemini Agents, Microsoft Copilot Studio and Amazon Bedrock Agents.

Each allows you to create agents tailored to specific tasks or knowledge domains.

No-Code Agents: Zapier, Make, n8n

An easy way to build simple agents is through no-code platforms like Zapier, Make, or n8n. These tools let you trigger AI workflows based on time-based schedules, events (like form submissions or spreadsheet updates) or external calls (aka “webhooks”).

For instance, when a new exhibitor is added to my Google Sheet, ChatGPT can:

  1. Fetch info about their product from their website
  2. Write a personalized “Thank You” message
  3. Generate a video using my AI avatar

Another step can easily be added to send the video via email, WhatsApp, or another channel. All without writing a single line of code.

This is just one example. Far more advanced automations are possible. And you can do all of this for under €100/month in tools.

Why Agents Matter

This is why I’m so excited about AI agents—and why I keep testing and deploying new ones. Sometimes I forget I’ve set one up, and I’m pleasantly surprised when it completes a task I didn’t remember assigning!

You probably already have dozens of use cases in mind—both professional and personal—where agents could save time, boost productivity, or simply remove repetitive friction from your day.

AI and LLMs aren’t going away. If anything, they’re just getting started. Investing even a bit of time each week to learn and experiment with agents is well worth it.

Want to dive in? YouTube is full of amazing, free tutorials on building and using agents. And of course, your favorite LLM—whether it’s ChatGPT, Claude, DeepSeek, Gemini, Grok, or others—can act as your personal tutor. Just ask!