Presented by KPMG
AI is more hype, less reality, say three-quarters of the executives surveyed in a 2021 study by KPMG, “Thriving in an AI World: Unlocking the Value of AI across 7 Industries.” And half think AI is moving too fast in their industry, even as they wish their company was moving faster. The hurdles to implementing AI are high, from a fundamental misunderstanding of what AI actually is and what it can do, to the significant lack of expertise available to help AI-curious organizations get their footing, says Dr. Ellen Campana, head of enterprise AI at KPMG LLP.
“The overhype is a real concern,” Campana says. “There are a lot of people trying to get into the game. Many have had a bad experience along the way, because they have put their trust in a group or person that didn’t have a lot of experience with AI, or they didn’t have a clear understanding of what to expect when launching an AI program.”
At least some of the hype is warranted, says Campana. According to KPMG research, AI is becoming ubiquitous — 92% percent of the execs that have implemented the technology in their organization believe in its ability to deliver value and make their organization run more efficiently. There is clear confidence that AI has the potential to solve some of their industry’s biggest challenges. Yet there are still long-standing hurdles to overcome.
The long-term impact of the hype cycle
This isn’t the first time people felt that AI was moving too fast and too much was being promised, Campana says. The term ‘AI Winter’ was coined as a way to describe those periods of boom and bust, specifically when interest in artificial intelligence faded during the technology’s long history — the last was in the 90s, during the dot-com boom.
Despite the public discourse, work has never stopped on advancing the technology. In 2012, breakthroughs in machine learning began to renew interest in the tremendous potential of AI. But those earlier hype cycles have had a long-term impact, Campana says, leaving a painful dearth of experienced and knowledgeable AI experts.
“Because the training programs were shut down due to that AI Winter, the people who are trained to do this work are scarce,” she explains. “It feeds into the current sense of things being overhyped, because there’s not a lot of people with deep training, but there’s a big market demand for it. This leads to a lot of variability in the advice people are hearing.”
Many of those companies that are having unfortunate experiences with AI now have unrealistic expectations about the technology — they haven’t been educated about what to expect. Campana points to the surprisingly common idea that AI is something that you install, load up, and run.
“People expect things to happen quickly, and seem to believe and expect that AI will just know things,” she says. “Improving literacy of these AI systems and improving people’s knowledge about what can and can’t be done is key.”
The real promise of AI, and the risks
Due to major breakthroughs in business platforms and tools, AI is prevailing across industries, Campana says.
“AI excels when you find a way for the human and the computer to collaborate efficiently,” she says. “If you divide the tasks based on the characteristics that each participant is good at, then there’s a lot of promise that together, people and machines can do a lot more.”
Campana and her colleagues have been engaged in a broad variety of AI projects in their work at KPMG. To address the impact that COVID-19 has had on the supply chain, they’ve been using AI to reorganize and reconfigure the food distribution system and optimize it despite disrupted supply chains and changing markets.
In technology, they’ve been working on automating identification of commercial leakage in contracts, and for financial institutions, they’re applied AI to determine whether companies are in compliance with regulations.
Their AI solutions also deal with earnings calls in order to understand how companies are talking about their finances, the implications for stock market valuation, and whether companies are communicating accurately. In health care, they are helping to optimize customer experience at payer organizations which are being flooded with calls at an unprecedented level. And in education, conversational AI is helping to distribute the technology children need in order to keep learning, particularly during the pandemic.
The risk in implementing solutions like these, or the thousands of others available, is underestimating both the need to participate actively in developing the systems, and the need to find expertise, Campana says.
As well, people will sometimes listen to marketing from technology vendors rather than seeking out third-party intelligence. That can lead to spending money on the wrong technology, or investing in a system or vendor that doesn’t understand how to maintain and keep the system current. Or companies can run into a problem with their AI implementation because their vendor has told them they don’t need a particular component, but that may be because the vendor doesn’t provide it.
“Companies need to be aware of the difference between marketing and implementation,” Campana says. “They’re sometimes turning to a technology vendor for advice about strategy, which is not something to do. They should either develop their own strategy considering multiple perspectives or come and ask for help to develop strategies. But they definitely need to make sure that they have a strategy that’s independent of a particular vendor view.”
Implementing a hype-proof AI strategy
Nearly eight in 10 executives in the KPMG survey reported that AI is functional in their organization, and a majority who are using it say it’s delivering value beyond what was promised. But how does a CTO get company buy-in without stirring up the fear that that they’re overpromising, or buying into the hype?
Campana says that you can start small, with a proof-of-concept project if necessary. It’s important to know that it will need to scale, and to have a plan for how that can realistically happen.
For C-suite stakeholders, it’s critical to continuously monitor progress and provide reporting while offering concrete, practical evidence as the project evolves. That includes documenting performance improvements as well as outlining opportunities for improvement, all the while keeping executives informed of the iterative process.
Of course, that also includes impact on the bottom line: it’s vital to identify KPIs that are about money. For instance, documenting call deflection in conversational AI, or identifying commercial leakage in contracts, so the efforts can be tied back to the business value, even in the early stages.
“We’re not waiting until three months down the line when [execs] say, ‘We invested a million dollars in your AI initiative, what have you got to show for it?’” Campana says. “You have to show them the iterative improvement. They have to understand that it’s not something that gets installed and just works. It gets better because you teach it.”
In order for that to happen — for the AI to learn and get better –it’s imperative to get buy-in from internal teams and have them actively participate in the shift to AI. For instance, if you’re implementing an AI solution to help contracts for commercial leakage, the group that handles that issue needs to be actively involved in helping the AI learn what to look for — not just the vendor, and not just a consulting group.
“A consulting group can help accelerate and make things more efficient, and so can a technology team, but they will need the help of the people on the ground,” Campana explains. “They need to spend time making sure that those people are bought in and understand that the goal is to make their lives better, not to replace them.”
For all levels of the company, implementing AI successfully takes patience as well — or in other words, understanding that AI isn’t like a lot of other IT domains where you install the software, configure it, and then it just pretty much works. It is, instead, a process.
That requires the most important part of a successful AI implementation: AI literacy, or combating the long-term damage caused by the last AI Winter. That’s the most important piece of advice Campana has for her clients about realizing the promise of AI.
“You need to pay attention to data literacy and AI literacy from the bottom to the top of the organization as you begin your AI journey,” she says. “If employees know what to look for, they’ll know how they could delegate their least favorite tasks to a computer. That means innovation opportunities that drive a lot of efficiency gains, coming from the people who are doing the work.”
Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. Content produced by our editorial team is never influenced by advertisers or sponsors in any way. For more information, contact [email protected]