Hey there, 

Over the past six months, Meta, Google, OpenAI, and Anthropic have all launched their own AI agents, each one promising to handle tasks for you in the background. Four of the biggest AI companies releasing the same kind of product within months of each other is worth paying attention to. But since each company describes its agent a little differently, it can be tricky to tell how they compare or how they are different from the chatbots you already use. So this week, we’ll explore what makes these agents different and how they compare.

What Makes These Agents Different From Regular Chatbots

ChatGPT, Claude, and Gemini can already search the web, connect to your apps, and take actions for you, so the difference isn’t whether the AI can do things. The big difference with these newer agents is that they can keep working after you leave the conversation. They can monitor connected apps, handle ongoing tasks, and reach out when something needs your attention, all while running on their own computers in the cloud. So in other words, a chatbot is like a capable helper you work with in the moment, while these agents are more like an assistant you can hand a job to, who texts you when it’s done or when something comes up.

For a real example, Anthropic shared a story from a brand strategist testing its agent. It noticed a call with a potential client on her calendar, prepared a brief for it, and then pointed out that the call fell right before her flight and offered to move it, something she hadn’t caught herself. And because agents work without you watching, they’re generally designed to check with you before doing anything important, like sending a message or making a payment.

Meet the Four AI Agents

Muse (Meta). Meta’s agent, Muse, launched in September. It can send emails, shop online, book travel, and help turn long-term goals into action plans, and you can talk to it in its own app or through WhatsApp. Each Muse runs on its own secure virtual machine that holds both the agent and your data. It’s free for most tasks, with paid plans for heavier use.

Spark (Google). Gemini Spark is what Google calls a 24/7 personal AI agent, built to work on tasks on your behalf. On a Mac, it can organize folders, use your local files to build documents, and handle multi-step work across Google Workspace. It’s currently available to Google AI Ultra subscribers.

Dots (OpenAI). OpenAI’s dots are always-on agents inside ChatGPT that keep working between your conversations. Each dot gets its own cloud computer, connects to your apps, and brings results back for you to review. Dots are available on ChatGPT’s Pro plan, and as of last week, you can even text your dot through iMessage or RCS in beta, and it can text you first.

Claude Duet (Anthropic). Anthropic’s Claude Duet arrived in October and is built for professionals and small business owners with a lot on their plates. Each morning it sends you a short brief on what matters across your connected apps, checks in every few hours, and messages you only when something needs your attention. Before it sends a message, makes a payment, signs a document, or deletes anything, it shows you exactly what it’s about to do and waits for your approval. It’s currently a waitlist research preview for Pro, Max, and Ultra subscribers.

If You Want to Try One

If you’d like to see how an agent works without paying for a subscription, Muse is currently the only one of the four with a free option. You can use it through its own app or on WhatsApp. If you do try it, starting with something low-stakes, like planning a weekend dinner, is a good way to get a feel for how it works.

My Thoughts

Agents are one of the more interesting applications of AI, because they actually do work for people instead of just answering questions. But for agents to become anything more than a glorified chatbot that handles small, low-stakes tasks, there needs to be a real level of trust, and I don’t think we’re there yet. Personally, I wouldn’t hand an agent anything mission critical and let it go off on its own without thoroughly double checking everything it did. And if I have to review every step, it’s fair to ask whether it’s actually saving me any time. Still, every major AI company launching an agent within a few months tells you where things are headed, and I think the idea has real merit as AI keeps improving. So the real question is how quickly people will trust them, and I think that will come down to their own experiences and whether agents keep getting things right.