Although financial institutions see clear benefits for using AI, measuring return on investment is much murkier. 

Of roughly 1,000 financial services firms surveyed, 77% said their AI investments are not delivering measurable ROI, according to an Aug. 3 report by professional services firm PwC.  

While 53% are measuring productivity improvements, just 38% are calculating outcome metrics and 36% are determining payback period times for AI investments.  

Because measuring ROI is complicated, banks should think about AI investments as if they were a portfolio with a healthy mix of short- and long-term paybacks, Jay Budzik, senior vice president and director of AI/ML model development at Fifth Third Bank, told FinAi News.  

“You want to have a nice blend of hard financial returns with hard financial commitments to the soft returns that you might expect from an efficiency-oriented program,” he said. 

For example, software engineering productivity is bound to increase by using AI, but engineering expenses probably won’t decrease unless an FI deliberately cuts labor costs, Budzik said. 

Quality over quantity  

Measuring ROI starts with upfront discussions about exact metrics and realistic payback timelines for each use case, Sahil Sagar, chief AI and information officer at commercial real estate lender Greystone, told FinAi News.  

A small number of use cases to start helps ensure alignment with these objectives, he said. 

“You can’t go into it 70% and then say I don’t see an ROI,” Sagar said. “I would rather do a few [use cases] and do them really, really well wherein there is real ROI versus running a series of them.” 

Sagar, who started as chief AI officer in June, said Greystone is reorganizing use cases into a smaller list. Despite urges to ramp up AI adoption, focusing on quantity over quality leads to disjointed AI objectives and muddied performance metrics, he added. 

Not just productivity gains  

Most FIs can measure productivity gains resulting from AI tools that automate repetitive tasks.  

While productivity gains are effective for short-term return metrics, FIs also must evaluate the higher-value work being freed up for more precise ROI projections over time, Sagar said.  

Vinay Mehta, chief technology officer at lending SaaS provider Solifi, said banks should break ROI metrics into three levels:

1. Foundational infrastructure expenses;
2. Application-specific metrics; and
3. Workflow transformation.  

Fixed or sliding costs? 

Banks also must account for how AI-related investments will evolve, whether that’s “embedding it into an existing legacy infrastructure, building a new infrastructure, tokens or vendor costs,” Annie DeStefano, chief executive of financial services-focused consultancy Ann DeStefano Advisory, told FinAi News.  

Grasshopper Bank Chief Technology Officer Peter Chapman told FinAi News that the digital bank is making its technology stack “as flexible as possible” to minimize volatile cost swings.  

That way, “we don’t get locked into a vendor who then cranks up the cost and forces you down a token path that isn’t going to make economical sense,” he said.  

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.