News

August 20, 2024

Monzo alumni raise £2.8m led by LocalGlobe for AI customer service agents that 'act like humans'

UK challenger bank Monzo continues to be fertile ground for producing Europe's next generation of founders

Tom Matsuda

4 min read

A trio of former Monzo employees have raised £2.8m in a seed round led by LocalGlobe for Gradient Labs, an AI customer service startup that supports companies working in “complex and higher-risk environments”. 

Other investors in the round include Berlin-based Puzzle Ventures and Monzo cofounders Tom Blomfield and Jonas Templestein. 

Gradient Labs’ funding round follows in the footsteps of other former Monzo employees who’ve since pivoted to entrepreneurship. Ex-employees from the digital bank have founded at least 16 other startups, Sifted reported in March. 

Advertisement

Cofounders Dimitri Masin, formerly Monzo’s VP of data science and financial crime, Neal Lathia and Danai Antoniou — both engineers at the neobank — have been building the startup in stealth since June last year. 

Masin was Monzo’s 30th employee, he tells Sifted in an interview, and was leading a team of around 150 people when he left last year. 

“Naturally, we spent quite a lot of time looking into operational and customer support problems and seeing how we can make the quality better, improve productivity and automate certain tasks,” he says. “That’s how I acquired deep domain knowledge not only within the operational space but also in AI and machine learning.” 

What are they working on?

Gradient Labs is building a suite of LLM-based autonomous agents that can “act like humans” when handling tasks for companies, says a job listing on its website. 

The company has so far partnered with 10 companies, four of which are fintechs. And, while the company is yet to generate revenue, it plans to monetise by charging a fee when one of its agents completes a successful outcome.

Masin notes that working with companies handling money means that Gradient’s agents may have to operate in “high-risk environments”. 

He says the startup’s agents have guardrails in place to detect whether they’re dealing with a vulnerable customer, or if they’re in danger of giving financial advice. 

For instance, if an agent detects that a customer might be vulnerable, it could assign the case to a more appropriate team to handle the inquiry. 

Gradient’s clients are supplied a “Notion-style interface” where subject matter experts can input processes for typical customer queries, such as a change of address request. 

The startup’s AI agents — built on top of Anthropic’s Claude 3.5 Sonnet model — are then fed that information so they can execute that request. Masin adds that the chance of hallucinations as agents can only generate replies based on these existing documents and processes, a technique known as Retrieval Augmentation-Generation (RAG).

Advertisement

“We believe that customer support will be completely redefined over the next three to five years,” he says. “That’s because with today’s technology you can automate 70 to 80% of work.” 

Fintechs and AI

Gradient’s fundraise comes as Europe’s fintechs increasingly begin to dip their toes into the world of AI. Nvidia’s head of financial services Kevin Levitt told Sifted at Money 20/20 in June that financial services companies are uniquely positioned to capitalize on AI. 

“We’re used to as an industry operating with new technologies in the construct of regulatory and compliance frameworks that are designed to protect the ecosystem,” Levitt says. 

As a longtime fintech employee, Masin says the Gradient team has built its experience of reassuring third parties — such as a regulator or a prospective client — into developing the product. 

“Before you switch it on, you essentially have all the controls in place and you have the confidence that nothing will go wrong,” he says. “And that mindset we took from fintech.” 

In the past six months, neobanks like Amsterdam-based Bunq and Nordic challenger bank Lunar have begun to roll out chatbots advising on spending limits or outstanding payment schedules. Other fintechs like Paris HQ’d Defacto are using AI to determine the creditworthiness of users. 

Currently a team of 11, Gradient is looking to increase its headcount to around 15 individuals so that it can serve larger enterprises before raising a Series A round. 

Tom Matsuda

Tom Matsuda is a fintech reporter at Sifted. Find him on X and LinkedIn