Meta has launched Muse, a personal AI agent that goes beyond answering questions. It can open a browser, use apps, fill out forms, send emails, book travel, make purchases, and keep working on a task even after you close the app.
This is what interests me most. I’ve seen AI grow from just generating text to reasoning over information, writing code, analyzing documents, and using tools. But there’s a big difference between giving advice and actually doing the work for me.
Muse brings AI closer to actually doing tasks for us.
So, what can Muse really do?
Meta says Muse is a personal AI agent that works with the apps you already use. Instead of giving it step-by-step instructions, you just set a goal.
For example, instead of asking an AI to find flights and then booking them yourself, you can tell Muse what you want and let it handle everything.
Muse can browse websites, fill out forms, manage schedules, handle emails, shop online, and work on long-term goals. Meta says Muse keeps working in the background and checks in if something changes or needs your approval.
This is why the system behind Muse is important. Muse runs in its own virtual machine with a built-in browser and security controls. Meta also created a separate monitoring system called Sentinel, which can approve or block actions. Users decide which apps Muse can access.
There are still important questions about reliability and security. Reuters reported that internal tests found both helpful results and serious problems, like reliability issues and security risks. Meta says it delayed the launch to make the system safer before releasing it.
This is something I’d watch closely. If an AI just gives a wrong answer, it’s annoying. But if it takes the wrong action, it can cause real problems.
Why this is more important than just another chatbot
This is where the conversation about jobs gets more interesting. For years, we’ve talked about AI taking over tasks like writing emails, making reports, summarizing documents, or answering questions.
But most jobs aren’t just one task. They involve hundreds of small digital actions.
For example, an administrative worker might check emails, update spreadsheets, schedule meetings, enter information, follow up with people, prepare documents, and update a CRM.
A customer support worker might read a request, look up an account, check a policy, issue a refund, update a ticket, and reply.
A sales worker might research leads, update the CRM, send emails, schedule meetings, prepare proposals, and follow up.
AI agents can link all these actions together. That’s the real change I notice.
If you want to learn how LLMs and AI agents are built, I recommend Hands-on GenAI, LLMs and AI Agents. It’s a practical guide that teaches these technologies through real projects.
Which jobs are most likely to change?
I don’t think AI will wipe out every job right away. Instead, it will take over large parts of some jobs.
Administrative work is a clear example. Tasks like scheduling, data entry, processing documents, managing emails, filling out forms, and routine coordination are things an agent can do.
Customer support is another big area. An agent could eventually handle a support request from start to finish, instead of just suggesting a reply to a human worker.
Virtual assistants and executive assistants could also see big changes.
Travel is another good example. An agent could compare options, watch prices, coordinate calendars, make reservations, and adjust plans if something changes.
Sales could change too. Imagine an agent researching prospects, updating the CRM, preparing personalized messages, scheduling calls, and following up automatically.
Personal productivity is another area. Instead of an AI just telling me, “Here are five ways to organize your week,” I can now imagine an AI that actually organizes my week for me.
That difference might seem small, but it’s not.
What new jobs might appear?
When I talk to learners about AI and jobs, I avoid saying that automation only destroys jobs. New types of work often appear with new technology.
As AI agents get better, I think there will be more demand for people who can design, supervise, evaluate, secure, and connect these agents.
This could mean jobs like AI agent engineers, workflow designers, AI operations specialists, evaluation engineers, AI security engineers, automation architects, and people who connect agents to business systems.
There will also be more need for people who understand the business processes being automated.
I’ve seen this many times in AI projects: knowing how a model works is helpful, but understanding the real workflow often decides if an AI system is valuable.
If You Want to Build This Future, Start Here
When I talk to learners about AI agents, I always suggest building something yourself instead of just reading about what companies like Meta are doing.
Muse shows where the industry is going: agents need to reason, use tools, remember context, find information, and take action. If you want to see where AI engineering is headed, here are four projects I recommend:
- Your first AI Agent with CrewAI
- Integrating Memory into AI Agents
- Build an End-to-End Agentic RAG System
- Build a Voice AI Agent From Scratch
I recommend building these projects in order. You’ll move from multi-agent workflows to memory, then to agentic retrieval, and finally to multimodal interaction. This skill set is much more useful for building real AI agents than just learning to write better prompts.
Meta’s Muse works on a much bigger scale, with browser access, app integrations, security controls, background work, and long-running tasks. But the main lesson is simple: you don’t have to build Muse to learn the engineering principles behind it.
The Takeaway
I think the biggest change won’t be humans versus AI. It will be people who use agents versus those who don’t.
Someone with an AI agent that handles research, scheduling, customer communication, admin tasks, and routine work can get much more done than someone doing it all by hand.
Learn how AI agents work. Get to know APIs, tool use, browser automation, databases, authentication, evaluation, security, and workflow design. Most importantly, learn how real businesses work.
I hope you enjoyed this article about Muse, Meta’s new personal AI agent that does more than just answer questions. For more tips on AI and machine learning, feel free to follow me on Instagram.
