Subscribe
Sign in
General
p/generalShare and discuss tech, products, business, startups, or product recommendations
Start new thread
Sayoni Dutta Roy

Do we really need another AI agent builder?

by•

These days, whenever I tell someone that I am building a no-code tool for creating AI agent automations, they roll their eyes. People usually follow up with polite comments, but I can sense them thinking, “Another AI agent builder? Aren’t there enough of those already?” So why am I building this? The short answer: We believe existing no-code agent builders fall short of what we’re trying to achieve at GenFuse AI. Here’s why: 1️⃣ Most AI automation tools only work for simple, rigid workflows. They can generate text in response to a query but cannot make decisions such as when to search Google or scrape a website for a given task. 2️⃣ Existing no-code platforms are too developer-focused, leaving non-technical domain experts out of the loop. If AI is supposed to communicate in natural language, why should building AI agents require coding expertise? We’re committed to building a platform that’s both simple to use and powerful, enabling ANYONE to automate complex tasks without needing technical skills. We want to make AI accessible to ALL. It’s a big dream, but we’re making steady progress. Have a different view? I’d love to hear your thoughts. Interested in GenFuse AI? Join the waitlist here https://www.producthunt.com/prod...

Add a comment

Replies

Best
Nha Hyerin
It depends on the value it offers. If a new AI agent builder brings unique features, better usability, or addresses specific market gaps, then yes, it's needed. Otherwise, it may struggle to stand out.
Sayoni Dutta Roy
@nha_hyerin Totally agree. We feel that there are some major gaps in the market right now (specifically around usability of tools) and we aim to address them with GenFuse. But I would love to get your feedback if you could try GenFuse once it launchs
Tony Tong
I don’t think we really need another drag-and-drop AI agent builder. With the advancements in LLMs, a natural language-driven approach feels like a more intuitive and scalable solution. Instead of manually arranging workflows, being able to simply describe the task in plain language and have AI generate the necessary code could offer a much more flexible and powerful experience. That said, it’s great to see more tools emerging in this space—there’s definitely value in offering different approaches. One thing though, the term "no code agent builder" might not be the best choice for marketing purposes. It sounds a bit abstract, and it could benefit from a clearer, more engaging name that speaks directly to the user’s needs and the value proposition.
Sayoni Dutta Roy
@tonyabracadabra Great points Tony. I totally agree that the ultimate future for AI agent builders will be using just natural language to describe a task and let the AI create the workflow. But I don't think the LLMs are there yet, especially for more complex, multi-step tasks. We ultimately will create an AI that can create this workflow for users from natural language prompts, which the user can tweak a bit and run. I was also thinking about "Create AI agents to automate business operations without writing any code". Does that sound better than "no code agent builder" or is it too long?
Tony Tong
@sayoni_dutta_roy from a UX perspective, incorporating a natural language interface alongside the drag-and-drop builder could significantly differentiate your app from other agent platforms. This approach would cater to users who are less familiar with workflow-based systems, making the app more accessible and intuitive. While the NL interface might not handle complex workflows perfectly, that's not a concern at this stage since it would serve as a complementary feature. As you're just starting out, this could be a strategic way to stand out while enhancing usability
Anton Osipov
I'm actually surprised that people think there are enough AI agent builders. Every single industry could use many different types of agents, and this expansion has barely started. The scope of work is enormous and the more options there will be, the better. On top of this, this field is too new to identify the best building approaches. There is so much more space to grow: example-based training, template compilation, user-guiding interfaces, collaborative building, intent-based design, etc. I guess in the future, we'll have all kinds of agent platforms, unified approaches, and industry standards. Maybe at that time, it'll be the time to doubt if there is a need for a new builder, but not now.
Sayoni Dutta Roy
@anton_osipov Great points. Yes I also believe we're too early in the game right now. There are quite a few solutions though but it's still too early to say who is the clear winner, both in terms of company and the approach to building AI agents. I would love to hear more about intent-based design and how you're thinking about it
Anton Osipov
@sayoni_dutta_roy As far as I understand, the intent-based design is actually what @tonyabracadabra suggested: a plain language description of the desired result when a system decides how to solve the task. What I like about this term is how it shifted my point of view on agent design. When I checked GenFuse AI, it looked simple and flexible. Nevertheless, I started wondering how it might fit into the intent-based approach. After getting a user's request, the system could check if this request fits in one of the predefined templates of typical tasks. If it fits, a starting flow could be created automatically. LLM aggregators and prompt writers already exist, so setting up different agents inside the flow should be possible. The system could also have a library of how to interact with different services. As a result, the system could build a draft of a complex agent that will be finalized through the drag-and-drop interface by a user. Of course, all of this is just a theory since I'm a product designer and surely missed some important points.
Rafael Krueger
The idea of removing barriers for non-coders resonates with me.
Sayoni Dutta Roy
@rafael_krueger Glad to hear that. Do try out GenFuse once it launches. :)