Most CIOs welcomed generative AI as more than just a visitor in the enterprise IT workshop. Like it or not, the technology rapidly spread across cloud platforms, into software bundles and onto PCs, popping up in chatbots, copilots and, most recently, newly minted agentic tools.
“We are using AI all over the place,” Amanda Luther, managing director and partner at Boston Consulting Group, said during a CIO Dive virtual event on March 12. “We’ve created these different tools, and now we’re trying to bring that together into the overall workflow for a typical person on one of our teams.”
KPMG is leveraging the technology, too. The professional services firm is keeping a close eye on where and how generative AI is yielding the biggest efficiency gains, Htike Htike Kyaw Soe, managing director, advisory digital at KPMG U.S., said during the event.
“We’re integrating it into the overall process of how we think about how we build and how we do the work, as well,” Kyaw Soe said. “We’re also taking a hard look internally at where could be the pockets of value.”
Generative AI remains deceptively difficult to scale safely and profitably. It’s a technology so approachable it speaks the same language as its users. Yet harnessing the expansive vocabulary of a large language model for enterprise use cases has been tricky.
Companies have struggled to implement strategic adoption plans even as the great majority plow forward with short-term, tactical deployments, according to Pluralsight. As vendors roll out generative AI-powered autonomous agents, data security and privacy concerns continue to fester among IT leaders, a recent SnapLogic survey found.
Despite ongoing upskilling efforts, AI talent woes persist, blocking the path to adoption.
IT consulting firms like BCG and KPMG are developing task-ready enterprise AI tools while they scale up internal usage.
Luther said roughly 80% of BCG’s employees are using generative AI weekly. The company used the technology to soup up its internal knowledge navigation tool and developed a content-creation application for internal presentations and reports, she said.
At BCG’s campus in Austin, Texas, there’s a process in place to encourage further adoption. “Every month at our all-staff meeting, we have a GenAI moment,” Luther said. “We have somebody come up and talk about something that they did that month that made their lives easier.”
Failing forward
False starts and unsuccessful pilots have been a painful part of the generative AI adoption process. The average organization abandoned nearly half of its AI pilots before they reached full production last year, according to a recent S&P Global Market Intelligence report.
While it may seem counterintuitive, CIOs should embrace the occasional misfire and focus their AI efforts on a few promising use cases.
“It’s a trite thing to say, but celebrating failures on some of this matters,” Luther said.
Kyaw Soe concurred. “Sometimes when people hear the word failure, they freak out about it,” she said. “It’s OK to try some things out … some things work and some things don’t.”
Companies that are meeting the scaling challenge have limited the number of active proof-of-concept trials, BCG found. Organizations focused on three or four problems they’re trying to solve with AI are two times more likely to succeed than companies with 10 or more pilots in process, Luther said.
“If 2023 was the year of POCs all over the place, I think 2024 was the year of trying to clean that up and actually focus on a few things that matter,” said Luther. “It’s what you don’t do that matters.”
When an AI chatbot or assistant is ready to roll out, workforce buy-in is crucial and not to be taken for granted. CIOs should be cognizant of the learning curve inherent in a new technology.
Even the small things can be difficult initially, Kyaw Soe said.
“What we’re finding is that it’s challenging for people to say, ‘OK, I see that you’re giving me this opportunity, but I’m struggling to figure out what the first thing that I want to ask is.’”
Broader stakeholder sentiment can be a key lever to speed or slow adoption.
If a CFO is enthusiastic about AI use cases for accounts receivables, that can be a good place to get the momentum rolling. “Push on the open doors,” said Luther. “You’re fighting an uphill battle if the end user at the most senior level isn’t excited about what you can do.”
