Glia has won the Banking and Financial Services category at the 2026 Artificial Intelligence Excellence Awards for a platform designed around the constraints of customer service in regulated finance.
The company says its Banking AI can automate up to 80% of interactions for banks and credit unions. It has also announced a contractual commitment to resist hallucinations and prompt injection for clients using the platform.
The recognition highlights a shift in enterprise AI: buyers increasingly judge products on accountable operation, not the novelty of generation.
Banking-specific design narrows the risk
General-purpose models can answer broad questions, but financial customer service contains specialised workflows, sensitive information and regulatory obligations. A system needs to understand what it may say, which data it may access and when a person must take over.
Glia trains its platform specifically for banking work. Narrower context can improve relevance and make controls easier to define than an assistant expected to handle every domain.
Automation at the claimed 80% level could remove a large volume of repetitive inquiries. Employees could then spend more time on complex service, client relationships and activities connected to lending and deposits.
That percentage should still be evaluated within each institution. Interaction types, escalation rates and customer expectations vary. A high automation rate is valuable only when resolution quality and trust remain strong.
Safety becomes a commercial commitment
Hallucination and prompt injection create distinct problems. A hallucinating assistant may invent a policy or account explanation. A malicious instruction can attempt to override controls and expose information or trigger unauthorised behaviour.
Glia’s decision to make contractual promises around these risks is notable because it moves safety from product marketing into the vendor relationship. The precise terms and remedies determine how strong that promise is, but the direction aligns incentives more clearly.
Banks need evidence through testing, monitoring and incident reporting. A contract cannot prevent every failure. It can establish expectations for how the provider designs safeguards, responds to incidents and supports the institution’s obligations.
Customers also need a clear route to a human. Automated service should recognise uncertainty and transfer context rather than trapping users in a conversation the system cannot resolve.
Human connection remains the differentiator
Glia CEO Dan Michaeli argues that institutions face pressure to provide instant, intelligent service while protecting the human relationships that define their brands. That balance is central to banking, where a routine balance question and a distressed customer’s financial problem require different treatment.
AI can offer speed and availability for standard interactions. People provide empathy, discretion and judgment when the situation carries emotional or financial complexity.
The award panel described 2026 as a year of execution and results. For banks, those results should include more than deflection from call centres. They should measure successful resolution, customer satisfaction, escalation quality and whether employees gain capacity for higher-value conversations.
Glia’s momentum suggests that sector-specific safety can become a competitive advantage. Financial institutions are more likely to deploy AI deeply when the platform reflects their risk environment from the beginning.
The strongest banking assistant will not try to imitate a person in every interaction. It will handle bounded work reliably, explain its limits and bring in a professional at the right moment.
That is how automation can increase efficiency without erasing the relationship on which a bank’s service ultimately depends.
Institutions evaluating Glia or a similar platform should test ordinary questions, adversarial prompts and emotionally complex cases before rollout. They should also examine what happens after escalation: whether the employee receives the full context, whether the customer must repeat information and whether the original AI response remains available for audit.
Safety becomes credible when it survives these operational details. An award can identify promising direction, but production evidence determines whether secure automation genuinely improves service.
