"The heartfelt email, that clever turn of phrase, those pithy bullet points in the slide deck -- did they come from the boss, or a bot?
Executives are using artificial-intelligence tools such as Google's Bard and Microsoft-backed ChatGPT to spiff up their communications, though the prevalence is hard to measure for one big reason: Some won't admit to having AI assistance, lest employees think less of them.
Leaders are often eloquent or think they ought to be. Needing high-tech help to find the right thing to say can feel like a cop-out. Never mind that immortal words uttered by the likes of Moses, Susan B. Anthony and Ronald Reagan were sometimes conceived by others.
Looming over managers is a recent episode at Vanderbilt University in which an associate dean and assistant dean drew criticism from students and colleagues -- not to mention a thrashing by late-night host John Oliver -- for relying on a text generator to compose their response to a mass shooting on another campus. They were reinstated after brief leaves.
For leaders observing the situation, it is unclear whether the real mistake was enlisting a computer co-author or forgetting to delete the disclosure -- "paraphrase from OpenAI's ChatGPT AI language model" -- before clicking "send." How many others have done something similar and escaped scrutiny by covering their digital tracks?
"I don't think people care if you're using tools to augment yourself," says Steve Chase, who heads KPMG's U.S. consulting practice. "I think they're bothered when you lose the human touch."
Nearly two-thirds of senior executives expect generative AI to have a major impact on their organizations in the next three to five years, according to a KPMG survey, and most say they're still figuring out how to implement the technology. Potential uses include helping supervisors write performance reviews of their subordinates and assisting human-resource managers with drafting job descriptions. IBM said this week it will pause hiring for back-office roles that could soon be automated.
ChatGPT has been adopted faster than many corporate rule makers can police, though JPMorgan Chase and Verizon are among the companies that have restricted usage. With few guidelines, a lot of bosses are on their own to determine when using artificial intelligence boosts efficiency and when it's a crutch.
Mr. Chase says he treats ChatGPT like an editor; it can take a long email he's drafted and make it concise. The substance is still his, but the prose is better.
AI is capable of much more, however. As a test, I typed into ChatGPT a prompt that might tempt executives looking for a shortcut this month: "Write a chief executive's email to employees for Memorial Day."
The resulting note struck an appropriately reverent tone, encouraging employees to enjoy a long weekend and reflect on "the brave men and women who have made the ultimate sacrifice." Though a bit cliche and repetitive -- it went on for seven paragraphs -- the email could pass for one written by a person in the C-suite.
Natural-language software works by drawing on what's already in the public domain, which might explain the one glaring flaw in my AI-generated email: It included last year's date of Memorial Day.
With a little proofreading, formulaic messages produced by ChatGPT perfectly capture the beige tone of most corporate communications. That makes them attractive to risk-averse executives, says Ryan Todd, CEO of Headversity, a digital mental-health company.
"Templates are a safe move, and ChatGPT helps executives make safe moves," he says.
Dr. Todd, a psychiatrist, says he uses ChatGPT to help prepare remarks at conferences and workshops, while still relying on his own expertise and peer-reviewed research for the material. He says it helps him overcome writer's block.
Once, he leaned on the software to churn out internal memos and received a compliment on his speed. He considered taking full credit but decided to be honest.
Leaders of large companies have long had help workshopping their communications with human aides. Kenneth Freeman, CEO of Quest Diagnostics from 1995 to 2004, says a trusted vice president often handled first drafts and was invaluable when crafting messages to employees and customers after the Sept. 11, 2001, terrorist attacks.
He can't imagine delegating something so sensitive to a bot. On more routine matters, however, he says AI can assist executives at small companies who don't have deputies.
Shanea Leven leads a team of 14 people as founder and chief executive of enterprise software startup CodeSee. She doesn't have a VP, chief of staff or executive assistant to draft her messages, so she sometimes uses AI to get started.
Before running a recent webinar for fellow entrepreneurs, for instance, she prompted ChatGPT to write a description of the event.
More than substance, the first-time CEO wanted an example of a webinar promo's format because she had never written one. Ms. Leven says inexperienced business leaders shouldn't be judged too harshly for seeking AI's help with things that others might expect them to know.
"I hope that people have enough grace to be, like, 'OK, this is the first time she has done this,' " she says.
ChatGPT also can smooth rough writing by executives who are non-native English speakers, says Dror Weiss, CEO of Tabnine, an Israeli company that makes software to automate lines of computer code. Even at an artificial-intelligence business like his, there is a risk of overpolishing to the point of inauthenticity.
Mr. Weiss, who has about 30 employees in Israel and the U.S., says he tries to keep his own voice in everything he writes because he doesn't want messages to sound computer-generated, even if they're computer-aided.
"If someone says, 'Oh, ChatGPT probably wrote that,' it's a criticism that you're not an independent thinker," he says." [1]
1. On the Clock: As Bosses Reach for AI Tools, They Need to Be Smart About It --- Artificial-intelligence software can lend executives a hand but risks losing that human touch. Borchers, Callum.
Wall Street Journal, Eastern edition; New York, N.Y. [New York, N.Y]. 04 May 2023: A.11.
Komentarų nėra:
Rašyti komentarą