LHV Bank has gone live with agentic AI for customer email support in the UK, powered by Gradient Labs. The launch follows a proof of concept announced earlier this year and means LHV’s retail customers can now get immediate answers to everyday banking questions by email, around the clock.
The AI agent handles common enquiries from start to finish, such as what to do about a lost card, an explanation of a payment, transfer queries, or what happens next with an existing request. Cases that are sensitive, complex or outside the agent’s approved procedures are passed to LHV’s customer support team.
Faster answers, more time for complex cases
Like other banks, LHV Bank's traditional queue-based approach to customer queries didn't distinguish between simple questions and complex ones. With this launch, simple queries get immediate answers, freeing up LHV's team to focus on the cases where human judgement matters most.
How it works
- AI agent handles scoped journeys: Using Gradient Labs, LHV Bank's AI agent resolves clearly defined customer journeys from start to finish.
- Works within set limits: The agent operates only within approved procedures, tools and controls.
- Humans set the rules: LHV's teams define each process, its boundaries and what the agent is allowed to do.
- Humans approve changes: Any updates are tested and signed off by LHV’s team before going live.
- Ongoing oversight: LHV monitors quality and handles any exceptions the agent escalates.
Guardrails built for regulated banking
As LHV Bank operates in a regulated environment, its AI agent pairs large language models with deterministic guardrails. These control what data and tools the agent can access, when actions are blocked or escalated, and how complaints and vulnerable customers are handled. Customer data is never used to train the models, or build the agent, and is accessed only during live interactions.
Improving the agent like software
LHV Bank is also using Collaborate, part of the Gradient Labs platform, which applies software engineering discipline to customer-facing AI. LHV’s operations, product, engineering, risk and compliance teams can draft changes to the agent’s procedures in a controlled branch, test them against representative customer cases, review and approve them before they reach customers, and roll them back if needed. Every change carries a clear record of what changed, who approved it and when, so the service can improve continuously without the bank losing control of it.
What’s next?
LHV Bank’s immediate focus is to extend the agent to more customer servicing procedures within UK email support. Longer term, the two companies will explore further channels such as live chat and voice, back-office and financial crime workflows including onboarding, KYC and AML, and potential expansion into LHV’s Estonian business.
Kris Brewster, Interim CEO of LHV Bank, comments:
"As a digital-only bank, we don't have to accept the way customer service has always been done. Others in the industry are using AI co-pilots that simply draft replies for staff to approve. This might speed up individual steps, but customer queries still sit in the same queue, and when the AI is usually right, people can start approving suggestions without checking them closely. We wanted to go further and give customers an answer straight away, while keeping our people firmly in control of how the service works and stepping in whenever a case needs them. Getting that balance right is what responsible innovation looks like to us."
Dimitri Masin, CEO of Gradient Labs, added:
“The future of AI in financial services is not a faster human approving AI suggestions at every step. It is trusted agents taking responsibility for defined work, supported by deterministic controls, clear escalation points and human judgement where it matters. LHV is showing what that can look like in practice.”
