It’s Sunday afternoon in the heart of Ohio Amish country, and after driving their horse and buggy home from church, the Wengerd family is teasing Dad about the time Mom caught him using an AI chatbot to write her a Valentine.

We’re all sitting in their farmhouse’s living room, which contains zero TVs but lots of comfortable chairs. Valentine’s Day was weeks ago, yet Mary Ellen Wengerd is still tickled about the love note her husband, Daniel, co-authored with generative AI. “ It was so eloquently written,” she says, biting back a grin. “Reading this, I was like, This is so not my husband.”

Mary Ellen wears a white bonnetlike cap over her hair and the ankle-length dress she’d put on for church. She’s curled into an overstuffed armchair alongside the youngest of their six children, Jethro, 7, who’s parked his feet on her lap. “And I was like, ‘Did you write this?’” Mary Ellen says, looking at Daniel. “‘Or did you ask ChatGPT to put a little love letter together?’ And he was like, ‘Ummm —’”

“I actually did!” Daniel says as laughter erupts in the room. “But I made a couple changes!” We seem to be on well-trod comedic ground. Daniel’s blushing on the couch opposite Mary Ellen, stroking his wavy gray beard. Daniel points at me, the writer visiting from New York, in the corner. “Read it to him,” he says, “and then tell him what doesn’t sound like me.”

As Mary Ellen reads aloud and ribs Daniel about the fancy phrasing, she still sounds touched by the note. “I know the intent was right,” she tells me. “He’s on a tight schedule, and I imagine he got into a pinch.” The new AI wingman offering to help those in a pinch — whether writing love notes or code — is finding a toehold in Amish country. Holmes County, Ohio, has the highest concentration of Amish people of any county in the U.S. Visitors expecting to see traditional horses and buggies, bonnets and Abe Lincoln beards, won’t be disappointed. Still, they’ll find Amish entrepreneurs plugging into the digital economy and one clan of early adopters weaving generative AI into their knowledge work without much hesitation.

Of course, none of this sounds like the tech-shy Amish life in the popular imagination. However, there’s no such thing as a single Amish approach to technology. There are some 2,600 Amish churches across the country, and each makes its own, separate decisions about what sorts of new hardware and software church members can use. The Wengerd’s church is Old Order Amish. Its married members dress plainly, don’t drive cars or own TVs, and don’t connect their homes to the electrical grid. They speak a dialect of German at home and at church — which Daniel’s eldest son kindly translated for me during the service I attended.

Daniel is a minister in his church and has played a role in the congregation’s collective decisions to interdict smartphones and social media but to allow e-bikes, flip phones, solar-generated electricity, and religiously curated internet access. “I don’t want to paint a picture that we’re pushing for new technology and we don’t have respect for our traditions and our values,” he tells me. “We’re not just opening the door to anything.”

Among the 400,000 Amish people nationwide, there’s no one tracking how many are using AI. Historian Marcus Yoder estimates that less than half of the Amish people in Holmes County have internet access and, among them, fewer than 10 percent have tried generative AI. But Yoder, who runs the Amish & Mennonite Heritage Center in Millersburg, Ohio, adds, “My Amish employees do use it here.”

The sociologist Steven Nolt, a leading scholar of Amish life, says an unusually high percentage of Amish men in Holmes County work in manufacturing — almost half — making the place a more likely hot spot for tech experimentation. “These things are governed by tradition,” Nolt tells me. “And there’s no 95-year-old grandmother who can tell you the Amish way to use computers in your manufacturing establishment.”

Daniel Wengerd runs a manufacturing company in the heart of Holmes County. His father founded the business in the 1970s, welding together horse-drawn plows for local farmers, and today, its 55 employees do millions in yearly sales with customers ranging from small wagonmakers to huge data centers. “With the Wengerds, they are typical in that they’re illustrating this process of experimentation,” Nolt says. “They’re atypical in that they have a large and sophisticated company.”

Daniel, 45, tells me he sees AI as just another labor-saving tool that might seem a little uncanny at first — until it becomes essential. Take the mechanical reaper, he says, invented in the 19th century to do the work of ten men, cutting wheat stalks, bundling them together, then knotting them with twine. “ When that machine first came out, there was Amish people who were convinced that the devil is in that knotter,” Daniel says. “A lot of what I see in that is there’s a fear of the unknown.”

As members of a community that has already erected elaborate guardrails around the internet and computers, the early adopters trying out ChatGPT and CoPilot don’t exhibit much fear. As far as I can tell, they see generative AI as just another thing computers do. “A computer’s a machine that you tell to do the right thing,” Daniel tells me. “We just have a better way to tell it to do the right thing than in the past.”

“I use it every day,” says Daniel’s son Brian, 21, who this year became VP of operations for the family business. Around the time ChatGPT first came out, Brian had just worked his way off the factory floor into a customer-facing design role. Weeks into the new job, he had to write a tough email, updating a client on an order while firmly demanding detailed input to keep work on schedule. New to high-stakes client communications, Brian asked himself, “Hey, I wonder if Chat can help me with this?”

ChatGPT nailed the message, and soon Brian found lots more work for it to do. Brian opens his work laptop to show me the software that tracks the company’s sales, finance, and production workflows. You can buy commercial tracking platforms like this off the shelf, but Brian says none of them have the interactive dashboard that he’s demonstrating to me — a new feature he says powered the company’s highest-production month ever in February. As he clicks, colorful pie charts bloom onscreen to display real-time production capacity for every department in the factory. I’m barely following, but Brian’s proud of these pie charts — because he and a couple other guys at work coded this dashboard themselves. Or, I should say, vibe-coded it. “Probably 80 percent was from ChatGPT,” Brian says.

Not all the AI power users in Holmes County have office jobs. Brian’s cousin John, 19, works at an animal-feed store and runs side-hustles selling chickens and managing property bought with the proceeds. When a potential buyer for one property requested a rent-to-own arrangement, John researched the details with ChatGPT and asked the chatbot to draft a contract. Just to be sure, John says, “I sent it to an attorney, and he said he had nothing to change.”

John’s brother, Lewis Wengerd, 21, works for a concrete contractor who tasked him with researching an excavating machine the company wanted to buy. “At the moment, I didn’t have good access to a computer,” Lewis tells me. “So I called the number from my flip phone.”

I must seem confused, so Lewis tries to explain. “I run a Sunbeam F1 Pro Spruce,” he says, indicating his device, marketed as a “rugged dumb phone,” because his church bans the smart variety. That isn’t news to me, I say, but I confess I’d never heard of 1-800-ChatGPT, let alone dialed the number. “Have you never done?” Lewis says, laughing. “You ought to try it. It’s pretty neat.” (I did try calling 1-800-ChatGPT later. The female voice that answered — millennially lifelike down to the touch of vocal fry — responded to my questions until I felt slightly spooked and hung up.)

The guiding principle on new technologies, Daniel says, “is always that it doesn’t take away from what we experienced today.” The day he invited me over, our experience above all prioritized worshiping together and spending as much face-to-face time as possible with a community of close-knit families.

Raising these families takes resources, and over the last generation or so, economic necessity has pushed most Amish breadwinners off the farm. Steven Nolt, the sociologist, has chronicled the problems faced by an agrarian people raising huge families in a country paving its farmland over for suburbs. “Even if development had been frozen,” Nolt writes in a book he co-authored, “the growth of the population itself would have created a crisis with too many babies and too few farms.” For nearly all the Amish people I meet in Holmes County, agriculture is a side job, and their actual living comes from working for or running small businesses. Staying competitive means adapting to the digital economy.

“Our very specific geographical area has quite a few complex businesses,” says Ian Wengerd, 39-year-old father of six who wears the traditional Amish beard and trousers but at work favors polo shirts featuring his company logo. Just across the road from his cousin Daniel’s factory, Ian runs a 30-employee metal-fabrication shop specializing in structural beams, stairways, and railings. The company has a slick website, and Ian and I speak via a Google video call. Like almost everyone else in the Wengerds’ church, Ian pays for a Christian-run service that installs an app on all company devices blocking access to social media, sexual content, and anything else the church deems inappropriate.

The shop’s technological evolution helped me understand why he and the rest of the Wengerd clan seem so nonchalantly all in on generative AI. “We get involved in some state work, federal work, private work,” Ian says. “For us to try to do business with just a fax machine and a voice-mail, I mean I’d have to shut my doors.”

Ian’s father started out hand-welding steel railings in a converted calf barn, he says, but as sales took off, they couldn’t keep doing their books by hand. So in 2015 the company bought its first computer to run some accounting software. Email and internet arrived at the office in 2017. That same year, a contract came in for a big commercial job that required submitting plans drawn with computer-aided-design software, and so, Ian says, “I taught myself at night how to do CAD when I should have been sleeping.”

After ChatGPT’s release, Ian began staying up nights learning about generative AI. “I started using it soon after it came out, more or less testing it,” Ian says. “The more I used it, the more I thought this could actually be a good thing.”

It’s good for shortening emails Ian sends to clients; good for finding weak clauses in 200-page subcontractor agreements; and even good for planning out meals for his family, which has been on a health kick recently. “We’re trying to eat whole foods, right?” he says. “It’ll spit out the recipes and …boom, we’re ready to go.”

Ian’s not too worried about LLMs hallucinating false answers. For his purposes, Ian says he knows enough about the material at hand to spot mistakes. On smaller contracts, for example, he double-checks the AI’s work himself. “Once it gets over a million dollars,” he says, “I bring in an attorney.”

For the people I spoke with, all the headlines about the job-killing, attention-sucking, corner-cutting downsides of generative AI aren’t sparking panic. It’s early days, so perhaps these problems are yet to arrive. But so far, the Wengerds have managed to fence those hazards out.

John Wengerd, the chicken-selling property manager, doesn’t see much risk in falling down rabbit holes texting with his chatbot. Like everyone else I spoke with, John has no smartphone and his internet access is both filtered and mostly available at work. “I can’t lay in bed for half an hour asking Chat stuff. So the times when I’m vulnerable it’s not at my fingertips,” he says. “When I go home, I’m riding a horse or feeding chickens.”