Companies looking to experiment with generative artificial intelligence as an automation tool should consider starting with modest use cases, such as summarizing lengthy documents, according to Gartner analyst Arun Chandrasekaran.
While generative AI holds immense promise, some use cases present risks such as data privacy issues and may not be ready for deployment without steps such as careful “stress testing,” Chandrasekaran told CFO Dive Wednesday, after delivering a presentation at Gartner’s 2023 CFO & Finance Executive Conference near Washington, D.C.
“To me, whether generative AI is ready is really a function of what type of use cases you’re deploying it for,” he said.
In a recent Gartner survey, 70% of corporate executives said their companies were investigating or exploring generative AI, which refers to tools capable of producing text, images, or other content based on what they’ve learned.
Nearly 20% of respondents said their organizations had AI projects in “pilot or production mode.”
Corporate leaders are increasingly zeroing in on the technology following the launch of ChatGPT, a popular generative AI tool created by Microsoft-backed OpenAI. In Gartner’s survey, 45% of respondents reported that publicity around ChatGPT had prompted their companies to increase their AI investments.
“The generative AI frenzy shows no signs of abating,” Gartner analyst Frances Karamouzis said in a press release when the survey was announced. “Organizations are scrambling to determine how much cash to pour into generative AI solutions, which products are worth the investment, when to get started and how to mitigate the risks that come with this emerging technology.”
ChatGPT quickly became famous after its launch last November. It’s known for its ability to write reports, poems, and software code, among other capabilities. But it has also prompted data privacy concerns.
Apple, JPMorgan and Samsung are among companies that have reportedly restricted internal use of ChatGPT citing concerns over confidential corporate data.
In addition, ChatGPT has also been known to assert false information.
“I often say: everybody is entitled to their opinion, [but] nobody, including ChatGPT, is entitled to [their] own facts,” Chandrasekaran, a technology innovation analyst, said during his presentation at the Gartner conference.
One enterprise use case where generative AI can be “immediately productive” is document summarization, he told CFO Dive following his remarks.
“Organizations deal with very lengthy and complex customer documents, and these tools can go through these documents and summarize them into shorter, pithy sentences,” he said.
But Chandrasekaran advised caution when it comes to using generative AI for certain functions, such as those where customer engagement is involved. “That’s where I would wait and do some stress testing and fine tuning of the model before deploying it,” he said.