[ad_1]
There is no shortage of startups trying to create AI assistants for enterprises. Less common is an AI assistant that can actually perform tasks across many of your business applications simultaneously. That’s the promise of Narada AI, a startup building an AI assistant based on new research from UC Berkeley.
Narada has been operating in stealth mode for two years, and made its debut on stage today as part of Startup Battlefield 20 at TechCrunch Disrupt 2024.
Two of its founders, Kurt Keutzer and Amir Gholami, co-authored a paper earlier this year on “LLM translators,These are artificial intelligence systems that perform multiple functions simultaneously. Their startup is largely based on this open source technology, and they believe it represents a key differentiator from the generic, general-purpose chatbots out there.
The startup’s co-founder and CEO, Dave Park, says his team used this as a basis for building a custom AI model that could use productivity tools. Park, who has a PhD in computer science from Stanford University and spent 24 years in enterprise sales, believes LLM Compiler and Narada’s ability to use websites without APIs is the company’s “secret sauce” to winning the enterprise dealer race.
The idea sounds promising, but how does an agent actually work? In practice, I found that the Assistant was successfully able to perform a few different tasks using generative AI across different work applications, ultimately saving me a few seconds or minutes in different parts of my day.
The Assistant resides in a separate chat window in your browser, and can draft emails, make calendar invites, take meeting notes, and search the web for you. The company says its Assistant can also navigate enterprise applications, such as finding an invoice in SAP, taking notes on a video call, or analyzing information from numerous Salesforce applications.
I asked my AI assistant to craft a friendly-looking email declining an invitation I received. In seconds, a draft of an email appeared in my Gmail with the correct recipient (even though I didn’t tell him the person’s email, he found the correct one), subject, and body, all filled in with my signature at the bottom. All I had to do was review it and hit submit.
At another point, I asked the AI assistant to find a highly rated Japanese restaurant in my neighborhood in San Francisco, and book a calendar invitation to have dinner with a friend at a time that fit my schedule. He found a restaurant, created a calendar invite, and emailed my friend with the information.
How does the agent do all this?
To use your email and calendar, the agent uses APIs in part to access these programs through a developer-facing backend. However, Park says their AI agent also clicks, scrolls, and types across the front end of websites (this is how open email drafts appear in Gmail, for example). This front-end agent, which they call Web Redemption, should allow Narada to use enterprise applications without APIs, such as HubSpot.
The agent works like a Roomba, creating an internal map to understand new websites or apps, says Gholami, the startup’s CTO. Once a user tells Narada they want to use a new app, the agent is supposed to map it out so they can understand how to use it. This is the idea that the founders brought to me.
But Narada isn’t the only startup trying to create an AI agent that can use websites through a front-end. It’s similar to the idea behind Anthropic’s use of the computer or Rabbit’s LAM. However, these are difficult to implement in practice, and require a lot of maintenance to keep them running. If web pages update their layout, this can break the proxy.
The main difference with Narada Proxy is that it focuses solely on enterprise applications, rather than a general-purpose proxy for any website. (When I tried to use Narada for LinkedIn or Facebook, I got an error message, although there is a demo on the company’s website where an engineer can use the tool with LinkedIn.)
As for the LLM compiler, it seems that people in the industry are already implementing an open source method. Gholami told TechCrunch that LangChain and LlamaIndex already have integrations with LLM Compiler. But Narada’s tool is unique among these tools because it’s enterprise-focused — the startup already has a Fortune 500 company using its agent, but hasn’t disclosed which one.
Is this a substitute for a real assistant? Not real. However, the tool sometimes seemed to use a shortcut for mundane tasks, which is more than I can say for a lot of AI tools today.
The only thing that made me feel a little uncomfortable was the amount of access I had to give this AI assistant. Narada can read all my emails, can see my entire calendar, and knows my entire contact lists.
Like any “smart assistant” or assistant app like this, you have to trust not just the technology, but the company itself — and that Narada won’t misuse your data or your company’s data. However, the company promises not to train its AI models on any customer data.
So far, Narada says it has raised a few million dollars from the few advisors it has brought on, but the CEO says they are now looking to raise more traditional venture capital.
[ad_2]