AI is already changing up the way we handle our daily tasks, with constant ways to apply the ever-growing list of models to discover what works best.
Imagine screening a hundred resumes to find the best candidate and even asking the AI to interview the ones with the most potential? Or possibly asking an AI to assemble all of your expenses from the year into one report for taxes, itemized and categorized exactly how you want them.
This concept of having AI do work for you without having to type in prompts each time is not just on the horizon — if you believe the company behind the new Manus AI, it’s already here.
If you've not come across Manus before, let me explain what it is, what it can do and what it means for the field of artificial intelligence. Specifically, how it's attempting to advance work automation and why some experts are a little skeptical.
Introducing Manus AI
Agentic AI means there is AI that can do work for you autonomously — like an agent for an NFL star or a celebrity. The AI has “agency” in that you are empowering and equipping the AI to do work for you even if you are not constantly monitoring the progress. That's what Manus AI is positioning itself as.
The best example of this comes from the official video on the Manus.im website. In a product demo conducted by Yichao 'Peak' Ji, the Manus co-founder and chief scientist.
In the demo, we can see how the bot can be used to scan resumes and then summarize the best candidates.
What can it do?
On the right side of the display when using the tool, you see the results in a panel labelled “Manus’s Computer” and can track along with what the AI is doing.
The idea is that you don’t have to monitor everything the agent does. Let’s say you ask Manus to read 100 books and summarize each one. Presumably, that would take quite some time, but you could let the “program” run and come back to it later. (By the way, we might have to come up with new terms for these things. For starters, is Manus really a “computer”? Is summarizing a bunch of resumes a “program”? Not really.)
Later in the demo, Manus is instructed to analyze stock prices for three companies over three years. The bot is then asked to build a website showing the results in a dashboard.
My thoughts went a few steps beyond what Manus can do today — that someday, we might have a personal assistant that runs like a cloud computer, analyzing data for us, writing content, developing presentations, and doing other work after providing simple instructions.
This agentic assistant might constantly monitor our stock profile and make changes for our benefit, increasing our net worth on a daily basis. It might look for fraud in our financial dealings and block them or challenge charges on our behalf.
Manus is making a splash right now similar to when DeepSeek debuted because it seems novel, unique, and powerful.
Expert reaction to Manus
Manus is mostly a proof of concept because it is not widely available yet. To test it, you need an invite code which might be hard to find. (Tom’s Guide made a request at the Manus website and through a media request but have not heard back on either yet.)
Yet, there’s some serious potential here. Experts who are tracking the AI field see Manus is an evolutionary step in the right direction.
“Manus will jumpstart the conversation about agents since there is very little in existing AI offerings that can actually function as a ‘personal assistant’ without a large degree of training,” says HP Newquist, the Executive Director of The Relayer Group and author of the book The Brain Makers: Genius, Ego & Greed In The Quest For Machines That Think.
“Manus does open the door to out-of-the-box AI assistance. US companies developing agent AI will have to acknowledge that the race is officially on.”
“From what I have learned, Manus isn’t your typical AI tool — it’s taking a meaningful step toward autonomy,” adds Pronnoy Goswami, an Engineering Leader at Workday and book author who studies AI infrastructure.
From what I have learned, Manus isn’t your typical AI tool — it’s taking a meaningful step toward autonomy.
Pronnoy Goswami
“Manus actively initiates tasks in a more agentic way, rather than passively responding to tasks. This evolves from a copilot to a coworker.”
Goswami says Manus can remember context and can think independently. “It can identify tasks, set goals, and act without continuous human prompts more or less autonomously,” he says.
That type of coworker is a bit different from well-known bots such as ChatGPT and Claude. Those bots are incredibly helpful and powerful, but they act almost like an MS-DOS prompt from decades ago in that you have to type in a prompt and wait for a response.
Reservations about Manus AI
Yet, as with the DeepSeek chatbot, there are concerns about Manus mostly because of where the AI originates. It was developed by a company called Monica based in China.
That means there’s very little information to help US users know how their information is stored, if it could be used for nefarious purposes, or even if there’s any security at all.
“Manus faces the same obstacles that DeepSeek faces, at least in the U.S.,” says Newquist.
“Since it is a Chinese product that accesses private data, there is the concern about how well that data is protected—if at all. This is a concern for all AI, regardless of where it is based, but China’s disregard for protecting private data is quite troubling to discerning users in the US.”
Apart from that, it’s hard to know how capable Manus will be, especially since the agentic AI is not publicly available yet. It could be extremely powerful and even game-changing, but without testing and evaluating more complex prompts, it could be mostly an elaborate demo.
And then there’s a more ethical question for each individual about whether they even want to allow so much AI agency into their lives.
“This leap in technology comes with its challenges such as: how comfortable are we to hand over significant control to an AI agent?” says Goswami.
“Can an autonomous agent like Manus, be trusted with mission-critical decisions? I highly doubt that it would be the case.”
More from Tom's Guide
- 5 tips to get the most out of ChatGPT from someone who uses it every day
- I ditched to-do lists for ChatGPT Tasks — here's 5 ways it's changed everything
- I test AI chatbots for a living — 7 common glitches and what to do when they happen
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.
