Developers Max Brodeur-Urbas and Rahul Behal think that AI has the potential to automate lots of business-relevant tasks, but that many of the AI-powered automation tools on the market today are unreliable and costly. Part of the problem is that users expect too much of AI, Brodeur-Urbas told TechCrunch — for instance, they assume that it can handle highly specialized, niche workloads where precision matters.

“If users ever want to use AI for enterprise purposes, the technology really has to have no margin for error,” Brodeur-Urbas said. “Leaving specific workflows completely up to AI is not realistic. Users would be paying for [an AI] to spin its wheels performing the same Google search over and over again.”

Still, Brodeur-Urbas, an ex-Microsoft software engineer, and Behal, previously a software developer at Amazon Web Services, thought today’s AI had promising narrower applications. So they started thinking about ways they could squeeze what Brodeur-Urbas called “real value” out of AI tech.

These ideas became a wrapper for the open source app Auto-GPT, then a proof-of-concept, and eventually a startup: Gumloop. Gumloop automates repetitive workflows with AI, aiming to streamline basic tasks.

“We started the company in a bedroom in Vancouver as a side project,” Brodeur-Urbas said. “We were trying to solve a very simple problem for a group of nontechnical people in a Discord server, and it spiralled into something larger than we could have ever imagined.”

Gumloop provides a workflow builder that integrates with third-party apps and tools including GitHub, Gmail, Outlook, and X. Users can drag modular components onto a canvas to build automations, or choose from prebuilt pipelines for tasks like generating daily stock reports and summarizing documents.

Brodeur-Urbas claims that teams at Instacart and Rippling are using Gumloop for various use cases.

“Today, thousands of users rely on Gumloop as a core tool for their business,” he said. “Giving nontechnical people the tools to solve their own problems without relying on engineers is where we found market pull.”

There’s no shortage of workflow automation tools out there. Parabola, Tines, Induced AI, and Nanonets come to mind. And on the horizon are “agentic” tools from OpenAI and others, which promise to automate more complex tasks end-to-end.

To remain nimble, Gumloop plans to keep its team quite small. The company is hiring, but Brodeur-Urbas said that the plan is to cap headcount at 10 people.

“Using AI to code let us have the throughput of a 20-person team and outpace competitors,” he claimed. “Our plan is to be a 10-person, billion-dollar company.”

As it prepares to relocate from Vancouver to San Francisco, Gumloop has closed a $17 million Series A round led by Nexus Venture Partners with participation from First Round Capital, Y Combinator, and angel investors including Instacart co-founder Max Mullen and Databricks co-founder and chief architect Reynold Xin. To date, Gumloop has raised $20 million in capital.

“We didn’t need the money at all,” Brodeur-Urbas said. “Raising money isn’t the goal — building a product people love is. This new venture capital will help us build and scale that product even faster.”