top of page
  • Writer's pictureJohn San Filippo

Q2 Leverages GenAI Internally and in Products

Updated: Apr 25

By John San Filippo

 

Digital banking and financial software provider Q2 is developing a generative artificial intelligence (GenAI) strategy that will enhance both its internal operations and the products it makes available to financial institutions. Finopotamus spoke with Q2 Managing Director Corey Gross to uncover more details about this important initiative.

 

It Starts with Governance 


Corey Gross

“The way we approached GenAI is, we built a governance structure for using AI ethically, responsibly and practically,” said Gross. “We're not going to run a bunch of shareholder demos for the sake of saying that we're a GenAI shop. We'll do things that deliberately improve the usability and value proposition of our products and can actually have an impact on employee efficiency and productivity versus just implementing the tools because everyone tells us that we should.” He added that the Austin, Texas-based company is currently experimenting with Microsoft Copilot for internal use.

 

“Once we were able to set that up, we, we built the AI COE (Center of Excellence) with three main pillars,” he noted. “The first is inform, teaching the company what is and isn't acceptable use of AI, teaching people how to use open-source tools and how that might benefit them. Teaching you the difference between what is a licensed and an open-source tool and how that changes the dynamic in terms of ownership of data and intellectual property,” he continued.” It’s also building blueprints to help developers not repeat the same mistakes as it relates to implementing AI.” 



Gross told Finopotamus that the second pillar is accelerate. “This is the AI COE acting as somewhat of an incubator for ideas across the company,” he said. “We know that the AI COE is not going to be the sole hub for innovation in this area. What it should do is enable the organization to take ideas they have and test viability. Whether that's code ideation, whether that's creating proofs of concept. See if it's the right tool for the job, and if not, disqualify it quickly and move on.”

 

The third and final pillar of the COE, he shared, is innovate. “This uses a group of subject matter experts in AI, bringing them together across the company and advising on where we should be making bets. That's both internally – who we should partner with to use AI – and also where our product bet should be. Should it be improving existing products, or should we be building discrete products that have GenAI?”

 

Meet the New Andi

 

Andi is an AI-driven Q2 product launched in 2017 as a component of the company’s PrecisionLender commercial relationship pricing and profitability platform. In October of 2023, the company announced that all future iterations of Andi Copilot would leverage generative AI and also that the product would be able to function and provide value outside the Q2 PrecisionLender platform.

 

“[Andi] used an older language model that allowed users to ask it questions and it would respond with results,” explained Gross. “A lot of its power was derived from its knowledge of how banking works and how it would learn from every single user action, every deal that was priced, and be able to offer more and more tailored advice for that type of relationship manager, for that type of deal, for that type of bank and that type of macroeconomic environment.”

 

GenAI, however, opened new possibilities for Andi. “We challenged ourselves with, what if Andi didn't have to just live within Q2 PrecisionLender – live within a single application – but we can bring it out? And what if instead of using older language models that weren't as conducive to two-way chat and context-related conversation, we gave it an LLM (large language model)? And what if we didn't just give it access to one application, but allowed it to see what the relationship manager or user is doing across all the various software tools that they use?”

 

The answers to those questions form the basis for the new GenAI-driven Andi Copilot. The product is being beta-tested in the commercial lending areas of several regional banks, Gross said. However, the technology will find its way into products more suitable for credit unions and other community financial institutions, he added with a level of certainty.

 

A Use Case

 

Finopotamus asked Gross to describe the Andi Copilot user experience. “Let's say you're in (loan origination software) nCino and you're working on a deal,” he said. “Andi Copilot might have already gathered all the documentation as it pertains to that deal that the customer sent you. They may have sent their articles of incorporation, profiles for all of their shareholders, their business plan – all the things that you as a relationship manager need to evaluate the viability of that deal.”

 

However, Gross explained that Andi Copilot also has knowledge of the institution’s lending policies. “So you might ask Andi Copilot, is this deal a good fit for our institution based on our policy and the types of customers that we work with? Andi Copilot would review all the documentation, crosscheck it against the policies like debt service-to-equity ratios that were, and then advise the relationship whether they should or should not move forward with this particular client.”

 

Among the many attributes Andi Copilot considers is the type of business. For example, Gross said that buried in a client’s documents may be the fact that the borrower has a small stake in a cannabis business. If that violates institution lending policy, Andi Copilot will alert the relationship manager.

 

No Integration Required

 

Finopotamus asked Gross whether Andi Copilot achieves its integration via API or some other method. “There's no integration that needs to happen between Andi Copilot and any of the products that it interacts with,” he stated. “It gets installed as a browser plugin and then it’s able to recognize and understand through its own program.”

 

Andi Copilot provides value “out of the box,” with minimum setup, he said. “The data that Andi Copilot needs to be effective on day one is access to the policies, procedures, and unique attributes of that financial institution for the use cases that it’s covering, which is right now centered around commercial relationship management. Once it has access to that knowledge base, it's able to start adding value immediately.”

 

Summing up Q2’s approach to AI, Gross said, “One of our core tenets is to build Iron Man suits, not Terminators.”

Comentários


bottom of page