Peapack Private Bank & Trust considered spending more than $375,000 a year on software to replace a process it was managing in spreadsheets.
Then its engineers built their own platform using AI. Peapack Private Bank & Trust’s Chief Technology Officer, John Kowal, says each account review now takes 50% less time.
The work is the wealth team's recurring process for reviewing client accounts and handling the associated approvals. Peapack has about 700 employees and serves individuals, families, and businesses in the New York tri-state area. As the bank grew, that process needed to grow with it.
Kowal shares the case in Episode 238 of The Artificial Intelligence Show, part of AI Transformations presented by Google Cloud. It is one example of a larger change at the bank: AI-assisted development has made it possible to build more software around its own business needs.
This all started with a deep pain point within the bank:
The wealth team was managing account reviews in spreadsheets. Kowal says the approach would not scale, so the bank evaluated third-party software.
The proposed alternatives would have cost Peapack upwards of $375,000 per year. The size of that commitment made the decision difficult.
“That's a significant investment, to replace something that, you know, frankly was working reasonably well in spreadsheets,” Kowal says.
There was no question the bank needed a more scalable way to organize the work. But the cost of the solution made it more palatable to just keep doing business as usual.
Not to mention, historically, buying had been the practical answer for a bank of Peapack's size. Kowal explains why:
“We simply didn't have the development resources to build those customized solutions for every business need,” Kowal says.
Naturally, that constraint had helped determine the bank's purchasing decisions.
But when AI changed what its engineers could build, it also changed which options were worth considering.
So, Peapack's AI engineers used Codex, an AI coding tool, to build a platform for the account review process.
The platform organizes reviews and approvals and pulls information from the bank's data warehouse. It gives the recurring workflow its own software, built around the work the wealth business needs to perform.
The data connection was an essential part of the build. Peapack had spent years developing its warehouse, a shared store of business information, before it began this stage of its AI work. It also had employees with development experience.
That meant the engineers could use AI to build against data the bank had already organized. They had both a defined business process to improve and existing information to put to work inside it.
As a result, a process managed through spreadsheets moved into an application that organizes its reviews, connects them to approvals, and brings in bank data. And AI played a starring role in helping Peapack’s engineers create that application.
With the platform in use, Kowal reports that each account review takes half as much time as before.
That gave the bank two reasons to value the build. It gained a faster review process and developed its own alternative to software priced above $375,000 a year.
For the wealth team, the practical value is time returned to employees. Kowal connects improvements like this to the bank's goal of getting people back to clients and reducing the work that keeps them at their computers.
That makes the review itself the useful unit of measurement. The bank can look at a recurring piece of business work and ask whether it now takes less time to complete.
It also keeps the reason for building clear. The application earns its place by improving a process the wealth team already has to perform.
Despite the gains made with AI, the human also plays a critical role remaining in the loop at Peapack.
That’s because account reviews and approvals carry responsibilities that a new application does not remove. So Peapack applies a bank-wide rule to AI's role in business processes:
“AI cannot replace a control. It can supplement controls, it can double check our work. It can identify something a human reviewer may have missed,” Kowal says.
In practice, that means keeping required safeguards in place when AI helps with the work. Leaders can use AI to build better software around a review process while preserving responsibility for its checks and decisions.
For companies reconsidering a spreadsheet-based process, Peapack's experience suggests a practical starting point:
The original choice at Peapack looked like continuing with spreadsheets or accepting a large annual software bill. AI-assisted development made a custom application feasible.
The lesson: An old purchasing decision can deserve a fresh look when the reason for ruling out a build has changed.
Did you enjoy this transformation story? Go deeper on how AI is reshaping work and business with The Artificial Intelligence Show. Each week, we break down what matters in AI and what leaders should do about it.