Gradient Labs has successfully applied Artificial Intelligence (AI) further and faster into customer service than most fintechs. The key to its success, co-founder and CEO Dimitri Masin said, is to provide a better and faster result that leaves customers more satisfied than they’d be with the best human support available.
That ethos has allowed Gradient Labs to move well beyond applying AI to the most basic of customer inquiries. Masin said the founding team knew they had to reach higher from Day One, as those interactions only represent, at most, 25% of customer operations. Mission accomplished, as Gradient Labs automates specialist support and back-office processes covering 75% of workflows while earning a customer satisfaction score higher than most human teams.
Masin and co-founders Danai Antoniou and Neal Lathia worked at Monzo, with Masin serving as vice president in Data, Data Science, Financial Crime, and Fraud. He said that as regulations and technology improved, they empowered neobanks to innovate.
But that innovation can be pursued in different ways. Some founders push the regulatory envelope, believing that either regulation will catch up or they’ll force it to change.
Better to begin from the perspective that one can launch innovative products that both push boundaries and keep regulators happy. Regulation isn’t bad; organizations like the Financial Conduct Authority care about protecting customers. Build a truly customer-first business, and you can safely forge new paths.
Why AI is the logical next customer service stage
Masin sees AI as the next stage in the evolution of financial services and customer service. Branch banking ceded ground to digital-first fintechs, who often created slick experiences but hit the same wall.
Massive human teams were still needed to complete a long list of repetitive behaviors like account inquiries, KYC, AML, dispute resolution, and more.
“And that second part of the problem is what’s typically still causing bad experiences,” Masin said.
Moving beyond answering those generic few questions is what has allowed Gradient Labs to reach that 75%. Masin said answers to those Level One queries are comparatively easy to map out. Investment questions? Claims? Dispute resolution? That’s a different level.
Common disputes will see more than a dozen regular queries that can be asked in different ways. More steps are required, including considering related customer transactions and overall history.
But many of those steps are well-documented. Customer service staff follow standard operating procedures; AI agents can, too. Financial institutions define those processes, while Gradient Labs automates them.
While some think that they can simply input data into ChatGPT, sit back, and wait for a solution, Masin said it’s nowhere near that simple. The foundation of that 75% is based on successfully addressing the “ambiguation stage,” where intent is clearly defined.
“That is key because this is where customer senses whether or not you truly understand their question,” Masin said.
While older demographics are less likely to become comfortable with agentic customer service, Masin believes the rest will as those processes improve. Even now, 50% of companies don’t disclose when a customer is talking to AI.
“At almost every single one of those companies, the AI agent has a higher customer satisfaction score than the best human interviewer,” Masin concluded. “They get their answers way quicker, and they’re more accurate.”
Gradient Labs in action: Helping a European bank go from 10% to 75%
Gradient Labs works with a European bank with 10 million customers and a digital-first customer experience. The bank’s product list includes savings, investments, pensions, current, and business accounts, as well as subscription tiers.
Scaling customer support was challenging, but before it committed to AI, the bank wanted a platform that it could build on and encode its own rules.
The inbound support requests numbered in the thousands across those many products and tiers. To achieve precision, the bank has to share its internal knowledge base of more than 1,200 articles outlining the many different service aspects. Banking staff added additional internal notes not outlined in the articles.
Gradient Labs next added thousands of human agents who recorded customer interactions. This added context and another 700 reference points.
“The result of this work was an agent that could handle the breadth of the bank’s product portfolio from day one, without months of training or a lengthy ramp-up period,” a case study states. “This set the stage for the bank’s first successful use cases and rapid scale-up.”
To ensure strict compliance and quality standards, the bank added the full suite of Gradient Labs’ finance-specific guardrails. They cover everything from prompt injection detection and vulnerability identification to financial advice detection, sensitive information handling, and complaint flagging.
Each conversation is real-time screened against the guardrails, pre- and post-response, so every reply meets compliance standards. The AI agent needed to meet a 95% standard; it scored 98%. The bank expanded the use cases it allows the AI agent to handle.