Taktile's $110M: the winning AI pattern is agents plus rules

A $110M check just landed on a company that doesn't sell "an AI agent." Taktile raised a Series C led by Growth Equity at Goldman Sachs Alternatives to sell an Agentic Decision Platform — and the design choice that got it funded is the one every operator should copy: AI agents run inside a cage of hard rules and human checkpoints, not loose on a chat prompt. The dollar figure is the headline. The architecture is the lesson.

What actually happened

Per Taktile's announcement and the BusinessWire release, the round included Balderton Capital, Index Ventures, Tiger Global, Y Combinator, and Dig Ventures. Taktile's pitch to banks and insurers is a modular platform that combines AI agents, deterministic rules, business context, and human oversight to automate high-stakes decisions — loan underwriting, claims, fraud, onboarding. The results it cites are specific: customers including Mercury, Monzo, Faire, and Pleo have hit 95% automation in B2B underwriting and 75% fewer anti-money-laundering false positives, and one large insurer projects $90M+ in claims-processing efficiencies on the platform.

Why it matters for your business

You are not underwriting business loans. But you make repeatable decisions all day — approve this refund, flag this order, extend this customer terms, escalate this ticket — and the temptation is to hand them to a chatbot and hope. Taktile's whole valuation rests on why that's wrong: when a decision has consequences, you don't want a fluent guess. You want the agent to propose, a rule to enforce the non-negotiables, and a human on the exceptions.

That's the pattern to steal, at any size. Write down the decision. Encode the hard limits as rules that can't be argued with — refund ceilings, credit thresholds, the cases that always go to a person. Let the AI handle the judgment in the middle, where it's genuinely faster than you. And keep an audit trail, because "the AI decided" is not an answer you can give a customer, an auditor, or yourself at 2am. The 95% that automates cleanly pays for the 5% you still watch.

We build decision automation this way on purpose — agents for speed, rules for the lines you won't cross, and a log for everything.

Sources: Taktile, BusinessWire.