Every day, employees around the world spend countless hours performing tedious, repetitive tasks, such as converting documents to PDFs, uploading them to Drive, and then importing them into a database and emailing them to their team. UiPath was a pioneer in this kind of “robotic” process, but has pivoted to an AI-driven mode of operation, even as a number of startups (Signavio, Servicetrace, etc.) are catching up. Now, after closing a new round of funding, startup Bardeen is launching a service to automate these tasks for businesses.
Bardeen's platform uses a natural language interface to automate repetitive knowledge tasks. The company secured $3 million in this new round, bringing its total funding raised to $22 million. That might be just a vague curiosity if not for the fact that investors in this round will provide significant distribution to the platform. Dropbox and HubSpot are strategic investors in the startup through their venture investment arms (Dropbox Ventures and HubSpot Ventures), and the two companies will also help distribute Bardeen's technology, which is released on Thursday.
If you think of it as “Zapier with more AI,” that's closer to Bardeen's product, but it's a bit more sophisticated than that sounds. It's not built for IT departments, but as a platform for average people in businesses to perform repetitive tasks, and a demo of the Bardeen interface for TechCrunch showed how easy it is to automate complex workflows.
The platform appears to be able to do quite a lot: copy and paste text from one document to another, search the web for related information, and send it all in an email. The startup claims to have more than 300,000 users and over 1,000 paying clients, including Deal, Miro, Kearney, WPP, and 10Web.
Founded in 2020 by Artem Harutyunyan and Pascal Weinberger, Bardeen's agent platform runs as a browser extension and is context-aware, allowing the agent to take prompts from the user and perform “plan” steps that Bardeen says help streamline repetitive tasks. The company also says the assistant continuously learns from usage patterns.
It also integrates with 100's of tools, including Microsoft 365, most CRMs, and sales platforms.
This repeatability is important because you can't easily get an AI platform to give the same answer twice, and the lack of this predictability in a business environment can completely kill any product.
CEO Pascal Weinberger told TechCrunch: “The problem with other AI solutions is that they don't provide repeatability — if you repeat the same task, it will do two different things. This is essentially how these language models work, but it makes them pretty hard to use in real-world business applications.”
So what is Bardeen's approach?
“For example, you can type in a prompt, like take notes for a meeting, convert it to a PDF, extract the email addresses and send the PDF to each person,” Weinberger says.
“The platform does that through a language model, and this is where the differentiation starts. There's a planning stage. The model understands that it needs to go into the calendar, extract the calendar events, extract the email addresses, and create a PDF. Once it can do that, you can simply type, for example, 'also send PDF to Pascal in Slack,'” he added.
Once the model has made a plan, follow it. “The next time you ask them to do the same thing, [the process] It's become a skill that can be learned, just like teaching an assistant or a junior colleague. So I can just write everything in natural language and execute it. Anyone can build automation like this.”
Of course, as always, the question arises: what kind of LLM will the platform utilize?
Weinberger says he uses Gemini to translate questions and OpenAI GPT models for “certain automation exercises,” adding, “But new models come out every week, and we have benchmarks to determine which tasks models excel at.”
As incumbents like Zapier, UIPath and others race to catch up, Bardeen appears to be trying to stay ahead, at least for now.