Customer engagement tech provider Glia is rolling out 25 AI agents to handle frontline customer service tasks for banks and credit unions. 

Glia CoPilot has a library of agents for tasks such as freezing a lost debit card, making a payment and initiating a dispute, co-founder and Chief Strategy Officer Justin DiPietro told FinAi News. 

“Our goal is to get to hundreds by the end of the year,” DiPietro said.

The launch pairs two capabilities Glia has spent recent months building: a knowledge layer, delivered through its Copilot product, and an agentic workflow builder that lets the AI execute multistep processes, such as issuing a new card or logging a Zelle dispute, inside a live conversation rather than simply answering questions about it, he said. 

“Knowledge, in my mind, is only 20% of the value,” DiPietro said. Now it is about how AI executes tasks easily without making mistakes. 

The agents are designed to blend deterministic steps — required actions like identity verification — with probabilistic ones that adapt based on what a customer has said, DiPietro said. 

If a customer has shared their credit card or account number on the phone call, the AI will capture the information and fill it in required fields along with providing suggestions to the customer representative for next steps, he said. 

The agents can also reduce errors in the FI’s operations, he said. 

“The number of cases for ‘fat-finger’ mistakes in FIs is actually pretty high,” DiPietro said. “The agent will closely listen to if a customer wants to transfer $5,000 or $50,000.” 

The development and use cases 

FIs can create their own agents as well, DiPietro said, adding that customer reps and engineers can give commands to Glia CoPilot in natural language, and it will do all the coding and provide an agentic workflow that employees can test and approve.  

The company uses Anthropic Opus5 for Glia CoPilot, he said. 

Glia said each workflow can be validated against unusual cases so that the agent understands all parameters, such as a customer requesting a travel notification for a sanctioned country, before it goes live, DiPietro said. 

Pricing for the new agent product is based on outcomes rather than usage, he said, adding that institutions purchase a bundle sized to their asset base.  

“If you go over your bundle, we’re not charging you and we will absorb the cost” this year, he said, citing budgeting concerns tied to consumption-based or token-based pricing. 

AI agents gain ground 

Nearly one-third of AI use cases reported by major banks in the first quarter were agentic applications, up from 15% in Q4 2025, according to data analytics company Evident AI’s March report. 

Anthropic was the most-referenced vendor in the quarter, and specialized vendors beyond the hyperscalers now account for 68% of all deployments, concentrated in workflow-specific use cases across credit, anti-money laundering and treasury, the report stated.  

As agents become integral parts of financial services, keeping governance in check is essential, DiPietro said. 

“At the end of the day, a human has to approve the code, the agents and its decisions,” he said. “Our CoPilot is not supposed to replace humans but rather augment them to be quicker and more efficient in their tasks.” 

Agentic systems don’t behave like prior enterprise technology. They can act autonomously, at machine speed and in ways that are difficult to anticipate, David Pan, director and AI industry practice lead at financial services giant Moody’s, told FinAi News.  

“The organizations most likely to be best positioned are those that meet that reality with transparent data foundations, auditable outputs and governance that keeps pace,” he said. 

The company did not disclose efficiency gains or number of FIs using the tool in its early deployment, saying it is too new to provide such numbers. 

Register here for the FinAi Lending Summit, set for Oct. 7-8 in Las Vegas. This inaugural event will include speakers from Fifth Third and Capital One as well as a fireside chat with Piermont Bank founder and Chief Executive Wendy Cai-Lee.