If most of our work happens in the digital domain, why should our assistants only be flesh and blood analogues, asks Satyen K. Bordoloi
Back in 2021, when I began claiming in my writings that assistants like Jarvis in Iron Man are a matter of ‘when,’ not ‘if,’ people would laugh. “All that remains is for us to have affairs with a digital girlfriend like in Her,” their reply would mock. My earnest answer of “yes” would leave them in stitches. I don’t blame them. Generative AI hadn’t yet hit the market, and our AImagination was in the pits.
However, by early 2023, when I wrote a trilogy of articles coining the term “PAI” – Personal AI assistants – and predicting it would be a trillion-dollar business, no one laughed.
Two years later, and with many more articles on the subject published by Sify, the first ‘Jarvis’ of its kind, Manus AI from China, has entered our lives. Manus is the first step into exactly the kind of AI butler I’ve been ranting about for years. In technical terms, the company refers to itself as the world’s first general AI agent (GAI?).
The Dawn of PAI-Manus AI’s Magic: Imagine a digital servant that doesn’t just take orders—it autonomously takes care of it. Need a plan for a Goa vacation? A stock analysis of Tata Motors? A 10-page report on India’s green energy boom? Give it to Manus, and it won’t just assist; it’ll execute. There’s no need to micromanage or ask follow-up questions.
Take an example. Type “Find me a 2BHK flat in Mumbai under ₹50k, pet-friendly, near a metro station”, and Manus will quietly scour multiple housing portals, filter options using Python scripts, generate a comparison spreadsheet, and even WhatsApp the top five brokers—all while you sip nimboo pani, haggle with your cabbie, or binge the latest season of Paatal Lok. It’s like hiring a digital peon that learns your quirks and can even leverage other AI tools in the pursuit of the tasks it’s given.
From Sci-Fi to Dabbawala’s Efficiency-How Manus AI Works: Launched on March 6 this year by Wuhan-based Butterfly Effect (a Tencent-backed startup), Manus broke the internet in China in less than a day. People were stunned that, unlike other AI, Manus AI feels like a team of specialised agents working clockwork like Mumbai’s legendary dabbawalas. There’s a Planner (the project manager), a Researcher (Google scrawler), a Coder (who writes Python), and a Designer (who creates charts, memes, and summaries). Together, they tackle tasks with precision.
Take another example. Ask Manus to analyse Reliance Industries’ stock, and here’s what happens: The Planner maps out steps, the Researcher downloads BSE filings and scrapes Twitter/X for Ambani-related buzz, the Coder runs correlations between refinery outputs and share prices, and the Designer builds a slick dashboard for you to regale at. What you eventually get is a polished report and a shareable link— human intern’s nagging excluded.
Manus runs tasks asynchronously. This means you begin a task during your morning coffee, forget about it, and by lunch, your phone pings: “Your stock analysis is ready. Also, I booked your Uber to Juhu Circle.” And if you want to peek into what your assistant is doing, you can catch its live feed showing which agents are working, what code they’re writing, and which websites they’re browsing. It’s like having CCTV for your AI—perfect for paranoid bosses who micromanage their employees.
To see a demo of how it works for yourself, visit https://manus.im/ and scroll to the “Use case gallery”. Click on one of the examples and watch it replay the task on your screen. This is more like a recording than the real thing, but you’ll get the gist.
The magic of a system like Manus AI lies in its ability to process and generate various types of data, including text, images, and code. This versatility means it can be used across various sectors, including software development, content creation, and research. Think of its ability to seamlessly integrate with tools such as web browsers, code editors, and database management systems as an octopus at work, its digital arms reaching out to gather the relevant bits of information or data.
MIT Technology Review writer Caiwei Chen took it on a spin (I’m on the waitlist for one) and wrote: “..using it feels like collaborating with a highly intelligent and efficient intern: While it occasionally lacks understanding of what it’s being asked to do, makes incorrect assumptions, or cuts corners to expedite tasks, it explains its reasoning clearly, is remarkably adaptable, and can improve substantially when provided with detailed instructions or feedback.”
What I Predicted: I define Personal AI Assistants (PAIs) as AI programs designed to autonomously handle a wide range of digital tasks with minimal user supervision and errors. These PAI, now known as autonomous agents, can make decisions based on natural language commands and interact with various applications to execute multi-step tasks. Kind of what we expect from our interns or personal assistants, except this is in the digital domain. I envisioned a future where such AI would seamlessly integrate into our daily lives, automating tasks to enhance our productivity. Back in 2021, I had even painted some detailed scenarios.
You set your PAI to make the presentation you are to deliver in the office, to talk to the PAI of your evening date to fix the time and mutually convenient restaurant, pay your electricity bill, buy some stocks and sell a few, and even match your shirt to your tie and trousers all while you brush, bathe, breakfast and head for office and all using just voice commands.
I had also predicted that this PAI business would become a trillion-dollar business. It turns out I was right in every detail (take that, naysayers) except one. I had imagined that such a system would emerge in the second half of the decade, reaching a cumulative trillion-dollar valuation by 2030. Seems like it’s happening much sooner.
Criticisms and Concerns: Manus became one of the first to demonstrate its abilities publicly, but it is not the only multi-agent system in town. As I had written in a previous Sify article, there are Google’s Project Jarvis and Vertex AI; Anthropic’s Agent, which can control a PC and automate tasks; Microsoft’s Windows Agent Arena; Open AI’s AI agents; and many other startups, such as Relay and Induced AI, all trying to achieve similar goals.
All of them, especially those like Manus, will face issues related to the transparency and explainability of tasks, as well as debates about human skill erosion resulting from overreliance on AI.
However, the biggest concern is about privacy and security. As an AI agent that can access and process sensitive information, concerns arise about data breaches and the misuse of personal data. We should be cautious about entrusting private information to an AI system, especially given the potential for unintentional fallouts. Manus AI’s ability to make autonomous decisions also raises ethical questions.
The lack of human oversight in decision-making could lead to unintended consequences, particularly in critical areas such as finance and healthcare. Hence, there’s an urgent need for robust oversight and ethical guidelines to ensure that AI agents make decisions that align with human values.
As you’d have guessed, despite the concerns, I’m biased towards PAI. I’ve spent years evangelising PAIs at Mumbai tech meetups and Guwahati family dinners. But beyond my partiality for PAI, Manus AI feels like the first real glimpse of a future where AI isn’t just a tool—it’s a companion that can get into your jugaad spirit. Will it replace jobs? Probably some.
Will it save us time? Absolutely. And honestly, I can’t wait for the day I can yell: PAI, cancel my meetings and plan a Goa trip! while it quietly files my GST returns as I stand in the flight boarding queue. So, here’s to Manus AI—the clunky, brilliant, infinitely promising start of our Jarvis era. Oh, and to everyone who laughed at my predictions—enjoy eating your words.
In case you missed:
- The Rise of Personal AI Assistants: Jarvis to ‘Agent Smith’
- Collaboration, Complexity, & Innovation: Understanding Multi-Agent Systems
- Forget Smart Homes – Welcome to Your ‘Feeling’ Home
- Anthropomorphisation of AI: Why Can’t We Stop Believing AI Will End the World?
- Generative AI With Memory: Boosting Personal Assistants & Advertising
- Rufus & Metis Tell Tales of Amazon’s Delayed AI Entry
- Hey Marvel, Just Admit You’re Using AI – We All Are!
- Apple Intelligence – Steve Jobs’ Company Finally Bites the AI Apple
- The Path to AGI is Through AMIs Connected by APIs
- OpenAI’s Secret Project Strawberry Points to Last AI Hurdle: Reasoning
