As the chief product officer for AI customer service startup TalkDesk, Charayna “CK” Kannan said that enterprises often say they want to automate different workflows but that it’s really hard to implement AI. Enterprises are dealing with clunky, legacy software, that often doesn’t have APIs, creating a daunting task their IT departments weren’t prioritizing.

“Every company that we talked to had anywhere from 50 to 1,000 automation requests from different teams in their backlog that they just never got to,” Kannan (pictured above on the right) told TechCrunch. “This just doesn’t make sense. In this day and age, you shouldn’t have a 1,000-line item automation backlog. You should be able to do it really fast.”

This realization became the impetus behind Kannan’s new startup Pinkfish, which helps enterprise customers build AI agents and other AI-driven workflows through natural language prompts. The software has more than 200 integrations, like Salesforce and Zendesk, and is focused on deterministic execution, which means the same user prompt produces the same result each time.

Kannan said that Pinkfish has tried a different approach than competitors when selling to enterprises. Instead of pitching its platform as a golden ticket to automate every workflow, Pinkfish tells the companies to try the software just to automate one or two different workflows at first. “So that’s where they start, and then they go from two to four, from four to 10, from 10 to 20, and hopefully 1,000 [automations through Pinkfish],” she said.

So far, that strategy has paid off. Pinkfish launched in stealth in January 2024 with Kannan as CEO and co-founder Ben Rigby as chief product and technology officer (CPTO). The company focuses on a few areas, including retail and services, and has landed hundreds of users and enterprise customers including Ipsy, Elevate, and TalkDesk, among others.

Kannan said that while many workflow automation startups are looking to help companies cut out some of the more “extra” aspects of a job like automating market research, or pulling potential sales leads, Pinkfish is focused on mission-critical workflows.

She gave the example of Ipsy, a makeup subscription service. One of the first workflows Ipsy used Pinkfish to automate was its price request feature, which was previously taken care of by a three-person team. This team would have to attend to each request manually regardless of whether it came in overnight or on the weekend. Kannan said now, that whole process runs through Pinkfish.

“It’s so mission critical,” Kannan said. “If Pinkfish screws up somewhere, guess what, your prices are not on your website. You leave money on the table. “

Now Pinkfish told TechCrunch exclusively that it is emerging from stealth and has raised a $7.6 million pre-seed round led by Norwest Venture Partners with participation from Storm Ventures and angel investors.

Scott Beechuk, a partner at Norwest who will be taking a board seat at Pinkfish, told TechCrunch that he has known Kannan since her time at TalkDesk and would tap Kannan to be an adviser for various Norwest portfolio companies.

Beechuk told TechCrunch that he was excited to back the company because he thinks Kannan and Rigby have the right balance of understanding the underlying technology and understanding the customer base to stand out in a crowded AI agent landscape.

“They are launching with a bunch of significant logos and paying customers who are finding real ROI, you back these seed-stage companies, they could take years to deliver real ROI,” Beechuk said.

Kannan also thinks Pinkfish stands out from competitors because it lets customers use natural language to prompt the system while using full code in the background to build these AI workflows. She said that while low code was popular for years, and still is for some of their competitors, she thinks in today’s environment its become too limiting and is effectively “dead.”

She added that companies don’t want to pick from a set of pre-coded building blocks, but rather would have a solution that gives them access to a full-code back end but with a simpler-to-use interface. As the AI agent market gets increasing crowded, she hopes that message resonates.

“How can we go bring tangible value to the mission critical, complex use cases? By grounding it with the agent and determinism, and bringing in one platform with the right level of guardrails for all of these connections,” Kannan said. “I think these are the two areas we are thinking differently.”