Enterprise adopters should avert the temptation to buy into the hype blanketing artificial intelligence when deploying AI solutions and need to be ruthless when AI projects fail.
That’s some of the advice in a recently released report by Forrester outlining some best practices to avoid common pitfalls when deploying AI in the enterprise.
At the top of the list is “avoid marquee AI use cases.” If your AI use case feels like a sci-fi movie, the report noted, it’s likely to fail or rely heavily on a person hiding behind a curtain, or both.
In general, it continued, good applications of AI will take an existing process and do it better, more efficiently, and cheaper.
The Forrester report maintained that such applications should augment complex human jobs, such as the AI tools that help nurses monitor and identify at-risk patients. Such applications can deliver extraordinary returns, it noted, but will never be featured in a sci-fi movie.
Real-world AI projects should look like that — highly functional with an expected ROI, it asserted.
New generative AI projects that feel too futuristic should give organizations pause, it added. The technology is in its infancy, so be careful not to rush projects, particularly customer-facing applications, into production.
Clear Use Case Needed
“We have this shiny, new technology that, in some ways, seems quite magical. We’ve never been able to talk to machines like we can today thanks to large language models,” said Forrester Vice President and Principal Analyst Brandon Purcell, an author of the report, along with Jeremy Vale and Rowan Curran.
“At the end of the day, you don’t want to adopt technology for technology’s sake,” Purcell told TechNewsWorld. “You need to have a clear use case in place. It needs to have real ROI attached to it. It needs to be technically feasible at scale, and there needs to be significant guardrails around it as well.”
It’s important for enterprises to be aware of what can and can’t be provided by the current state of AI, explained Kevin Butler, a professor at the University of Florida’s Department of Computer and Information Science and Engineering in Gainesville, Fla.
“The reality of what AI can do compared to what some may think AI is capable of can create a mismatch of expectations,” he told TechNewsWorld.
“You can use some of these tools as a starting point for thinking about how to approach a problem, but thinking of them as answers in and of themselves will often lead to very problematic situations,” he added.
Inhibitor and Catalyst
The hype around AI can deter some organizations from embracing the technology while having the opposite effect on others.
“The hype around AI is certainly impacting how organizations are assessing it,” said Erich Kron, a security awareness advocate at KnowBe4, a security awareness training provider in Clearwater, Fla.
“It is no surprise, given the complexity of AI and the inability to explain everything it does in the same way we can a standard decision tree, that organizations, especially leadership within organizations, may be hesitant to assess or deploy these tools,” he told TechNewsWorld.
The hype is pushing companies to implement AI before they understand the technology, leading to avoidable failures, added Rob Enderle, president and principal analyst at the Enderle Group, an advisory services firm in Bend, Ore.
“Because so many of these tools are being poorly implemented, care must be taken not to be over eager but also not to doubt the AI because of your own lack of capability and understanding,” he told TechNewsWorld.
“If you aren’t ready, it isn’t the AI’s fault,” he observed. “If you are overly eager, the failure is yours, as well.”
Mark N. Vena, president and principal analyst with SmartTech Research in San Jose, Calif., agreed that the relentless buzz and lofty promises about AI have created unrealistic expectations, pushing some companies to rush into AI adoption without a clear understanding of its limitations or strategic alignment.
“This can lead to misguided investments and disappointment,” he told TechNewsWorld.
“On the other hand,” Vena added, “the hype has also spurred innovation and investments in AI research, which can benefit organizations in the long run.”
“Striking the right balance between enthusiasm and informed decision-making is crucial for organizations to harness AI’s true potential,” he said.
For most organizations, AI won’t be replacing employees or providing an infinite boost in productivity, added Aron Rafferty, co-founder and CEO of StandardDAO, a decentralized autonomous organization and its subsidiary, BattlePACs, a political discourse platform.
“Images and chat through natural language is the focus of most startups in this cycle,” he told TechNewsWorld. “For most businesses, this does not make an impact, and if it does, it will take a lot of time and monetary investment to ensure a meaningful difference specific to the business.”
What kind of investment? He noted that Netflix recently hired a director of generative AI at a salary of US$900,000 a year.
Forrester’s best practices to avoid AI hazards also include:
- Prioritize projects in the sweet spot of business value and technical feasibility. If you start purely with the business value, you’ll choose use cases that play to AI’s weaknesses and miss its strengths.
- Improve your data iteratively. When it comes to AI projects, data is an ongoing process, not a static resource you can check off a list.
- Improve your AI capabilities iteratively. Just like with data, most successful AI initiatives take the capabilities that are available or can be rapidly acquired, deliver value quickly, measure and communicate that value, and use that success to justify investment in better skills, platforms, and processes as part of an ongoing virtuous cycle.
- Actively counter your human biases and then worry about biased AI. Actively seek out and counter biases in the data you want to use to train your models and source multiple technical and subject-matter-expert perspectives on your projects.
- Kill zombie AI projects. Despite the desire to cut dead weight, AI projects can persist in limbo either because powerful executive sponsors have set ill-conceived goals for them or because too few people in the organization understand AI well enough to spot the lack of progress.
Forrester also recommends that organizations plan with the entire AI lifecycle in mind. Your insights won’t drive value unless they drive action — that is, end users adopt them, the report noted.
“Companies have a unique opportunity to advance AI innovation and adoption in the workplace by building upon trust in the employer-employee relationship,” observed Hodan Omaar, a senior AI policy analyst with the Center for Data Innovation, a think tank studying the intersection of data, technology, and public policy in Washington, D.C.
“One thing they can do is start building on employee trust today,” she told TechNewsWorld. “They should focus on AI innovations that benefit workers and improve employee well-being.”
“If AI technologies offer clear employee benefits or employee value, then workers are more likely to embrace them in spite of concerns they may have,” she said.
Executives that adopt best practices and take the time to learn at a high level about AI will lead their firms to success, maintained Purcell.
“AI is an incredibly hyped technology, but there’s a good reason for it,” he declared. “It’s going to be transformational. It’s going to transform the way that humans interface with machines.”
“Up to now, we’ve interacted with them on their terms — through Windows or MS-DOS — but now we can communicate with them on our terms, through natural language,” he said.