WENCHANG, CHINA – MAY 29: A Long March-7 Y3 carrier rocket carrying the Tianzhou-2 cargo spacecraft … More
Another “Sputnik moment”? Another “DeepSeek breakthrough”? Hardly. The rise of Chinese AI — DeepSeek, Groku, now Manus — has created much media attention. But Manus AI, the latest AI from China, is far from novel. Unlike DeepSeek, which introduced meaningful advancements, Manus offers nothing revolutionary. It claims autonomy, but in reality, it’s just another large language model executing scripted workflows. The only novelty? It’s from China.
Manus Delivers Averages, Not Intelligence
Like many other tools, Manus relies on LLMs that generate plans based on statistical averages. If I ask ChatGPT, Gemini or DeepSeek for help with an article about Manus, they’ll generate an outline aligned with the dominant market narrative — emphasizing themes like “breakthrough,” “universal hand” and “China winning.” Why? Because that’s what the internet says. LLMs generate statistical averages, not insights.
Yes, Manus (like any other tool) can create a travel itinerary, but will it be good enough? In my eCornell certificate program, I introduce the concept of the minimum quality product. Does Manus meet the mark? It depends. Yes, if you treat this summary travel guide like any other. But for fully autonomous travel booking? I don’t think we’re there yet.
Manus Can’t Make Decisions
Decision-making requires prioritization and agency — something AI lacks. Take Benedict Evans’ analysis of OpenAI’s DeepResearch model for phone data analysis. The AI produced plausible-sounding but fundamentally flawed insights. Why? Because it couldn’t discern what data mattered most. Manus will face the same limitation.
Restricted Workflows — Key Success For AI
AI succeeds when it operates within structured, well-defined workflows. Companies that focus on domain-specific applications — not generic, do-it-all agents — will create real value. Reducing complexity, essentially narrow use-cases, allows AI to function effectively. Here are three examples:
- Octomind: Their AI agents automate end-to-end testing by analyzing web applications, leveraging AI and GenAI to interpret code, simulate user behavior, and generate precise test cases.
- Flank’s AI agents autonomously handle tasks such as document generation, negotiation support, and compliance form completion by analyzing legal requests, simulating expert decision-making, and producing precise outputs.
- r2decide’s AI agents enhance the eCommerce with genAI content by creating personalized sales advice integrated into the shopping journey.
All three companies leverage GenAI, but their applications are not general. Their AI operates within specific workflows, such as web testing, legal assistance, or eCommerce shopping. Within these narrow, well-defined use cases, AI can make decisions and automate tasks effectively.
Human Oversight Missing With Manus
The future of AI will be — wait for it — humans. Human oversight is critical. Most companies struggle with integrating AI into decision-making effectively. Here’s a video where I discuss human and AI interaction in healthcare. While AI can enhance scalability, human involvement remains essential, like Benedict Evans showed. OpenAI could research data and draft a report, but that will be wrong. The future will be companies that surface open questions where the AI lacks confidence. However that’s not what Manus is doing.
My eCornell certificate “Designing and Building AI Solutions” is a no-code course open to everyone. To support every student — regardless of coding experience — I had to built an AI co-instructor. Students equipped with an AI agent then explore and develop products. What do students tell me after the course? AI is great at summarization and coding. An AI can help being a dynamic checklist and memory. Where does it fall short? Guiding students in choosing what to prioritize. Human decision-making remains irreplaceable, no matter how “generalizable” Manu claims to be.
AI product design must emphasize two elements: quality and control. AI must be steerable. True AI success isn’t about replacing humans; it’s about augmenting them.
Integration of AI: The Real Challenge
For AI to be truly transformative, it must integrate effortlessly with external apps, databases and services. Manus’ demos of multi-app usage on social media tools are a step in the right direction. Today’s automation platforms — like IFTTT and Zapier — currently offer basic workflow integration, but they will follow suite.
Take Amazon. It recently announced that Alexa will soon be chat-ready, embedding deeply into existing tool ecosystems. That’s where where the market will move. Can Manus reach that level of integration anytime soon? I have my doubts.
Enterprise First, Consumer Later
Forget fully autonomous AI assistants for consumers anytime soon. Enterprises will adopt AI agents first because they can control deployment, manage risks and ensure ROI. Leading consulting firms like Accenture and BCG are already rolling out task-specific AI solutions. Consumers will follow — GPT didn’t take the world by storm overnight, and AI assistants still need multiple iterations. If we are there, Google and OpenAI will likely dominate this space — or an completely new user flow — but not Manus with same-old same-old LLMs.
AI Success Is Based On Application, Not ‘Made In China’
Manus AI’s biggest strength isn’t its technology — it’s its “Made in China” branding. The future of AI isn’t about flashy demos; it’s about consistent, secure and well-integrated solutions. Companies that refine workflows, integrate AI effectively and keep humans in the loop will define the industry. Those chasing flashy marketing will be left behind.
Disclaimer: I may have investments in some of the companies mentioned in this post. This content is for informational purposes only and should not be considered investment advice.
