Manus AI Review – My First-Hand Experience of using the General AI Agent

I’m excited to kick off a brand-new season of the Inspiring Tech Leaders podcast!

Over the past four seasons, I’ve explored the breakthroughs shaping our digital world, and this season, I hope to go even further.  Expect fresh perspectives, insightful conversations, and the latest in tech leadership and innovation.

This latest episode is hopefully not going to disappoint!  A few weeks ago, I covered the launch of Manus AI, a general AI agent developed by Butterfly Effect, and since then I’ve been given private beta access to test it first-hand.

In this episode, I share:
💡 My honest experience using Manus AI
💡 How it stacks up against ChatGPT, Gemini, and Claude
💡 Where it shines and where it falls short
💡 What it means for the future of AI-powered productivity

Manus isn’t just another chatbot.  It acts like a self-sufficient intern conducting research, planning, and executing tasks autonomously.  I tested it on everything from SEO optimisation to generating slide decks, and even complex policy documents.  Some results were truly impressive.

But it’s not all smooth sailing.  There are questions about data privacy, pricing, and stability.  So, I dive into the good, the bad, and the potential.

Is Manus AI the future of autonomous agents?  Tune in and decide for yourself. If you enjoy it, please like, comment, share, and don’t forget to subscribe wherever you get your podcasts.

Transcript

Welcome to the Inspiring Tech Leaders podcast, with me Dave Roberts.  I’m excited to kick off a brand-new season of Inspiring Tech Leaders!  Over the past four seasons, we’ve explored the groundbreaking innovations, leadership strategies, and transformative technologies shaping our world.  This season, we’re diving even deeper, bringing you expert insights, thought-provoking discussions, and the latest trends in tech leadership.

Whether you’ve been with us from the start or you’re tuning in for the first time, thank you for joining this journey.  Get ready for another season of inspiration, innovation, and impactful conversations.  Let’s get started!

A few weeks ago, I released an episode of the podcast focused on the release of Manus, a new general AI agent that has been making headlines across the world.  Following the episode, Manus kindly provided me with access to the private beta programme to enable me to test the AI agent first-hand.

In today’s episode I’m going to share with you my first impressions and experience of using Manus and how it compares with other similar AI platforms.

I will be exploring its strengths, its weaknesses, and whether it truly represents the next leap in AI-driven productivity. 

But let’s do a quick recap on that Manus AI actually is.

Manus isn’t just a conversational tool, it is an autonomous AI agent that can break down complex tasks, navigate the web, and even execute projects with minimal human input.

It has been developed by Butterfly Effect, a Wuhan-based startup, and it claims to be the world’s first true general AI agent.  Unlike AI chatbots that rely on a single large language model, Manus integrates multiple AI models, along with a network of autonomous agents that collaborate to complete tasks.

But what does that mean in practical terms?  Essentially, Manus isn’t just an advanced chatbot, it is designed to take action, break down tasks, and execute them autonomously.  Imagine working with a highly intelligent, self-sufficient intern who can research, plan, and even perform web-based tasks on your behalf.

However, access to Manus has been extremely limited.  It is thought that fewer than 1% of those on the waitlist have received an invite code.  Thanks to the team at Manus, I have been one of those lucky enough to test it out.

So, what does it look like and how do you get started?

Once you enter an invite code, you are directed to a landing page that looks quite similar to ChatGPT or DeepSeek, with past conversations on the left and a central chat window for interaction. The clean, minimalist interface gives it a professional feel.

One unique feature is Manus’s Computer, a window that lets you see the AI’s thought process in real time, which is super cool.  You can even intervene and adjust its approach, making the experience much more transparent than any other AI tools that I have tested.

I started off with asking Manus to conduct a review of the Inspiring Tech Leaders podcast and the PriceRoberts.com website.  It provided a very detailed report outlining how I could optimise my website and improve SEO, along with a strategy and roadmap to follow.  The recommendations also provided snippets of actual code to use.  Implementing the recommendations from the report, I was able to upscale the SEO on my web page from a B+ to an A+ On-Page SEO rating.  I was also able to follow recommendations to improve the site performance from being graded as F to now being B+.  It was very impressive to make such a significant difference so quickly.

What I found out though is that I quickly reached my daily maximum usage allowance, so I needed to think more carefully about the types of questions or tasks that I asked Manus to complete.

When my daily quota was reset, I could continue with the testing.  I had Manus create a PowerPoint slide deck on the topic of Secure by Design principles, something that I have been spending time working on recently, so I was curious to see what would be produced.  The result was a comprehensive slide deck outlining the core principles with appropriate images and meeting notes including references, all packaged up as a downloadable and ready to go PPT file.  Similar results were achieved with other requests in downloadable Word and Excel formats, which was really convenient and saved my editing time.

Overall, Manus performed well with these types of tasks and would refine its approach like a human assistant would when given feedback.

I found you needed to be careful with the words you use with Manus.  When using tools like ChatGPT I find I need to ask for a detailed response to a particular question to get the level of detail required.  I used the same prompt command with Manus and managed to burn through my daily quota on one question, which it then failed to deliver any results.  Even though I found that frustrating, I did accept that I needed to be more careful crafting my prompts.  

Manus feels like it is a more thorough tool than other AI platforms that I have used.  It doesn’t provide quick results, it spends time gathering information from various sources, comparing and compiling.  Using ChatGPT or Copilot, you get your answer or output quickly, whereas Manus feels like it is actually researching and analysing before it provides you with your end result.  This enables you to get on with other tasks in your day, while Manus works as your assistant in the background.  There were times when I would set Manus a task and then go and make a coffee or take a shower, but the results feel like it’s worth the wait.

My testing highlighted that Manus did fail more than once when presented with heavy workloads.  An early bug that I think has now been fixed, is that if you were running a number of different tasks and you hit your daily quota, you could sometimes piggyback off another task in the left-hand panel and continue asking questions in those sessions.

However, Manus has now moved to a Beta Subscription model.  The costs are not insignificant either!  To upgrade to Manus Pro (Beta) it will now cost $199 per month.  This provides you with 19,900 credits per month and the ability to run up to 5 tasks concurrently.  The Pro version provides access to the high-effort mode with enhanced stability and dedicated resources.  It will enable you to extend context length and also provide priority access during peak hours.

There is also a Manus Starter package available at a more reasonable $39 per month.  This provides you with 3,900 credits per month and enables you to run up to 2 tasks concurrently.  This package also comes with the promise of enhanced stability with dedicated resources, extended context length and priority access during peak hours.

To give you an idea of credits, I asked Manus to create a best practices document on a complex area of technology that I’m currently researching.  The result was a well thought out 9-page paper outlining the areas of interest to me and used 429 credits or the equivalent of $4.29.

Manus does learn from user feedback and improves the responses as it understands what you are wanting to achieve.  I love the level of transparency you get with Manus as it shows its reasoning and allows for user intervention if needed.

During the initial beta period, I did try asking Manus to create images, but this was not a strong point for Manus compared with tools like Midjourney or DALL-E.  I have now got limited credits available, so I’m being more thoughtful on what I’m asking Manus to do. 

The key advantage of Manus is its ability to combine research and execution.  It doesn’t just gather data; it actively works on projects with a level of autonomy that other tools lack.

Manus is just part of a collection of AI Assistants like ChatGPT, Google Gemini and Claude where there is a growing concern about how these tools collect, store, and use our personal information.  Manus, along with every major AI assistant retains data using long-term memory to profile users over time.  It remembers what you like, how you think, and how you work.  This might sound useful, but’s also a form of deep behavioural tracking.

We still don’t have public detail on where data is being stored and who might have access to it.  Your data could be used for internal training, sold to partners, or handed over to governments.  So, you really need to think about how you want to use these tools.  Manus is new and hasn’t had chance yet to work out its approach to privacy concerns.  Especially when compared to an AI Assistant like Claude, which is privacy focused by design and doesn’t retain data by default.

As summarised by Freddie Quek, fundamentally the challenge AI systems have is how they demonstrate trust and transparency. People need to have greater understanding and comfort around these two critical ethical principles of AI to help build trust. 

So, what are my final thoughts and is Manus the future?

After a few weeks of testing, I can confidently say that Manus is one of the most exciting AI tools I have used.  It is not perfect, far from it.  It still requires user oversight, struggles with some types of content, and occasionally crashes.  But its potential is undeniable.

Manus could become an essential AI companion for professionals, researchers, and even everyday users but perhaps at either a lower price point or with more credits for your subscription.

Right now, Manus is like a highly intelligent, eager intern, it does a lot of great work but sometimes needs a bit of support and guidance.  Whether it can evolve into a true AI assistant that handles complex tasks without intervention remains to be seen.

Well, that's all for today's episode!  But what do you think?  Is Manus AI the next big step for autonomous agents? 

Let me know your thoughts!

If you've enjoyed this episode of the Inspiring Tech Leaders podcast, don't forget to subscribe and leave a review.  Thanks again for listening, and until next time, stay curious, stay connected, and keep pushing the boundaries of what's possible in tech.