Editor’s note: This article draws on insights from an Aug. 14 CIO Dive live virtual event. You can watch the sessions on-demand.
Generative AI showed the enterprise its capacity to transform basic business functions, injecting seamless chatbot interactions and rapid summarization into daily processes. The novel, compute-hungry technology has also revealed several less savory qualities, including an ability to chew through IT budgets.
As an initial wave of C-suite excitement recedes, serious questions about ROI are surfacing. CIOs are getting down to brass tacks with the technology, asking how long is too long to wait for the payoffs to show up in the corporate balance sheet.
“We’re still plucking low-hanging fruit on a lot of AI initiatives,” PwC US Chief AI Officer Dan Priest said during a CIO Dive live event earlier this month. “The costs are not insignificant, but the payback and the pace of returns has been pretty good so far.”
PwC raced to the front of the professional services pack in the adoption race. The firm committed $1 billion to a three-year generative AI development push over a year ago and spun up a ChatGPT-based assistant called ChatPwC internally last summer.
After rolling out a generative AI tax solution built using Google’s BigQuery cloud data warehouse and Vertex AI developer platform in April, the company gave 100,000 employees access to ChatGPT Enterprise in May and solidified its commitment to the technology with Priest’s appointment.
“We took a look at the opportunities that AI represented for our business, the complexity that it was driving into our operations and, frankly, the importance of navigating this journey responsibly, and we thought those were all important enough issues that we need C-suite level leadership driving the agenda,” Priest said.
The executive guided the firm through an intensive 45-day impact analysis and coordinated implementation of a responsible adoption training initiative called My AI.
The findings were nuanced and reinforced conventional wisdom around the technology, according to Priest. The most promising use cases from an ROI perspective are industry-specific solutions that drive efficiency gains, he said. Generative AI coding assistants and software development lifecycle tools rank high on Priest’s effectiveness list.
“It’s not just about the technology,” Priest said. “It's about the business problem you're solving. Different sectors have different imperatives and having people who understand those sector dynamics and are able to design that understanding into AI solutions is really powerful.”
Risk encouragement
Emerging technologies come with inherent uncertainty and there are two sides to that coin. Early adopters are the first to reap the rewards and, potentially, stumble into uncharted hazards.
“There’s risk that needs to be managed around data,” Priest said. “The other risk is the risk of falling behind relative to your competition and you have to balance those risks.”
To mitigate data safety and security risk, Priest acknowledged the need for a responsible AI framework. Don’t use proprietary data to train a large language model without a firewall, he said. But don’t overcompensate with inflexible guidelines either.
“Make sure there's a sandbox in place for teams to experiment safely,” Priest recommended. “What’s exciting about AI is you can really reimagine your function and you don’t want to overengineer the governance on that — you want just the right amount.”
Another key aspect of risk tolerance is trusting developers to factor business value into use cases as they evolve without discouraging progress.
“We’re not micromanaging,” Priest said. “We’re really trying to see these innovations in a way that supports experimentation and a little risk taking.”
While ROI doesn’t take care of itself, Priest takes an intuitive approach to measuring value.
“We have this philosophy that we'll let the market tell us where we've innovated and where we've built the right solution,” he said. “If we're winning with our clients in the market, that's a pretty good signal that we've invested in the right solution and that will drive the ROI that we were looking for.”
So far, the market is telling PwC that it's worth fighting through the experimentation stage, Priest said. Generative AI’s ability to understand context, interface through natural language and perform cognitive processing tasks has applications that are just now coming into focus.
“I’m pretty realistic about the promise of GenAI, but those three features suggest there’s going to be a lot of use cases and it’s going to be with us for a while,” Priest said. “We’re just scratching the surface of what’s possible.”
