Taktile raises $110M to let AI drive bank decisions

Goldman Sachs Alternatives has led a $110 million Series C investment in Taktile, a startup that wants banks and insurers to hand over their most consequential decisions to AI agents. The stakes are enormous: a single misstep in underwriting, fraud detection, or claims processing can cost millions.

Financial institutions already spend billions on human teams to screen risky transactions, verify new customers, and process insurance claims. Taktile’s pitch is straightforward: let AI take the wheel on these high-stakes calls, while keeping a human in the loop to verify the outcome. The Berlin-and-New-York based company has now raised $184 million in total, with participants including Balderton Capital, Index Ventures, Tiger Global, Y Combinator, and Dig Ventures. Taktile declined to disclose its valuation.

The company’s focus is deliberately narrow. It does not aim to build a general-purpose chatbot like ChatGPT. Instead, it positions itself as the trusted middle layer that turns frontier models from major AI labs into reliable agents for the most sensitive tasks in finance.

What Taktile actually does is sell an Agentic Decision Platform. The system blends AI agents with hard-coded rules, relevant data, and human oversight. It automates decisions while allowing a person to review them. The use cases are specific to banking and insurance: underwriting business loans, assessing insurance claims, onboarding customers, and detecting financial crime. These are jobs that once required hours of human expertise, with real financial consequences if they go wrong.

The customer list is credible. Taktile counts banking startups Mercury and Monzo, wholesale marketplace Faire, and spend platform Pleo among its users. The company says it powers millions of decisions every day. Founded in 2020 by machine-learning engineers Maik Taro Wehmeyer and Maximilian Eber, Taktile claims striking results: customers have achieved 95% automation in business-to-business underwriting and a 75% reduction in false alarms for anti-money-laundering checks. One of the world’s largest insurers projects savings of over $90 million in claims processing alone.

The timing of this funding round reflects a shift in what AI can handle. Wehmeyer argues that models have only recently become reliable enough for high-stakes work. “AI has been around for a couple of years, but 2026 is the year where AI comes to financial services,” he told Fortune. The agents, he says, can now beat humans at many complex tasks. Taktile’s own research arm, Taktile Labs, spotted the turning point in December 2025, when frontier models crossed a threshold and could suddenly handle the kind of judgment calls banks had long reserved for trained staff.

The costs at stake are enormous. Moody’s estimates that financial institutions spend an average of $72.9 million a year on know-your-customer and anti-money-laundering work alone. That is a vast pool of manual labor that Taktile aims to automate. The startup is not alone in chasing this opportunity. AI labs and tech giants are racing to automate white-collar work across the board. Salesforce agreed to buy the support-AI firm Fin for $3.6 billion this month, and Meta unveiled a business agent to handle customer chats. Software engineers were the first to feel the impact. Office workers are next.

But Taktile argues that finance is different. A chatbot that invents an answer is awkward. A loan or claims agent that invents one is a regulatory problem. The stakes change what counts as good enough. “General purpose AI tooling is fine for simple automations, but it isn’t sufficient for operating mission-critical financial decisions where errors can cost millions,” Wehmeyer said. His pitch is that business owners, not just engineers, need to understand and steer the system. A head of credit or a fraud officer must be able to see why an agent made a particular decision.

That framing sets Taktile apart from rivals that bolt a thin layer onto a frontier model. It also fits a wider lesson from finance. Other startups, including former Citadel quants behind Moment, are selling operating systems built for the messy reality of regulated money rather than generic tools. The control angle matters because the work is sensitive. Catching financial crime wrongly flags innocent customers and lets real fraud slip through. A system that cuts false positives by three quarters, as Taktile claims, is valuable precisely because the errors are so costly.

The harder question is what happens to the people. Taktile is candid that thousands of employees currently process these decisions by hand. The promise is to free them for higher-value work, not to cut them. That is the optimistic reading, and it is one many AI firms offer. The reality has often been blunter. Companies across tech and industry have used autonomous agents to trim headcount, not just redeploy it.

Wehmeyer prefers a concrete example. Picture a tornado tearing through a house in Minnesota. One agent reads the claim document. A second interprets the damage against the policy. A third decides whether to pay. What once took an examiner weeks could take minutes. For the homeowner waiting on a cheque, that is a clear win. For the examiner, it is less obvious. The same automation that speeds the payout also thins the desk that used to process it.

The fresh capital targets growth. Taktile will spend it on better tools for complex banking and insurance cases. It also plans to widen its reach across the United States, Europe, and Latin America, including a new office in São Paulo. The Goldman backing carries a signal. Christian Resch, a partner in the bank’s growth-equity arm, praised Taktile for pairing technical depth with an understanding of how regulated institutions really work. A bank putting its own money behind a tool that automates banking is its own kind of endorsement.

The wager is bold all the same. Taktile is betting that banks and insurers will trust agents with decisions that, until recently, only humans were allowed to make. The savings look real. So does the risk. Whether a regulated industry hands its hardest calls to software, at the scale of millions of decisions a day, is the question this round leaves open.