Banking voice AI replacing human agents is the most compelling enterprise thesis in this space right now - and the metrics you're quoting are serious. For European bank clients, worth flagging that Article 50(1) of the EU AI Act goes live on August 2, 2026 for voice agents that interact with customers. The spoken disclosure - "This is an AI assistant from [bank name]" - must come at the start of every call, in the customer's language. For collections specifically, the tone and phrasing of that disclosure also has to meet EBA's emerging guidance on fair treatment in AI-driven debt recovery. The UX of that disclosure is not trivial in banking - it affects how many customers hang up immediately, how compliant the interaction log is, and how the transcript reads in a dispute context. If you're scoping EU pilots, the disclosure design is worth getting right upfront. Happy to share what compliant implementations look like from a UX/accessibility angle given the specific WCAG requirements for financial AI disclosures.
The outbound call thesis is interesting - handling the messy phone-mediated workflows is genuinely the hard part of agent execution. One compliance layer worth building in now: Article 50(1) of the EU AI Act requires voice AI agents interacting with people in the EU to identify themselves as AI at the start of the call. For an agent calling a dentist's receptionist on behalf of a user, the opening line needs to be something like: "Hi, I'm an AI assistant calling on behalf of [user name] to [purpose]." That's also good caller experience - businesses responding to calls appreciate knowing what they're talking to. This obligation sits with whoever deploys the agent (Asmi), not just the underlying model provider. The August 2 deadline is 48 days out. If EU expansion is on the roadmap, building the disclosure into the call-start template now is much easier than retrofitting it later.
