Safely manage your Zendesk from the AI assistant you already use, via the Deltastring MCP. Beacon configuration platform
← Back to news

LHV Bank to explore AI customer support with Gradient Labs

LHV Bank's proof of concept with Gradient Labs represents a deliberate pivot toward agentic AI in financial services customer support, specifically targeting email-based enquiries where controlled reasoning and workflow automation can operate within regulatory guardrails. Rather than deploying broad generative AI, the bank is testing systems that can plan, reason, and execute actions within defined parameters—a meaningful distinction for teams evaluating AI tooling. The collaboration emphasises explainability and compliance as design principles rather than afterthoughts, suggesting that financial services organisations are moving beyond chatbot-style implementations toward systems that can genuinely augment support teams whilst maintaining audit trails and human oversight.

For CX leaders already operating in regulated industries, this signals a maturation in how agentic AI can be positioned: not as a replacement layer, but as a force multiplier for response quality and consistency. The focus on email-based workflows is instructive—asynchronous channels allow for more deliberate AI reasoning without the latency pressures of live chat, reducing the risk surface. However, the critical question for teams considering similar deployments is whether your existing platform stack (Zendesk, Freshdesk, Salesforce Service Cloud) can integrate agentic systems without creating data silos or compliance friction. Gradient Labs' emphasis on "safe deployment" in sensitive interactions suggests that off-the-shelf AI add-ons may prove insufficient for regulated verticals, potentially fragmenting your tech estate.

The broader implication is that financial services—historically cautious on AI adoption—are now treating agentic systems as a legitimate operational lever, provided they're built with explainability and human judgment baked in. This validates the market for specialised AI vendors in regulated spaces, but it also raises questions about whether generalist platforms will develop comparable capabilities or whether CX teams will need to orchestrate multiple point solutions. For support leaders, the takeaway is clear: agentic AI in customer support is no longer theoretical, but the implementation path in regulated industries requires deliberate vendor selection and architectural planning.