Every podcast I listen to and parent I chat with at a soccer game right now seems to be grappling with the same problem: how not to become overly dependent on AI to help with everything from daily work tasks to planning next summer’s vacation. It’s a topic I explored in my last article for Forbes and one that dozens of founders wrote in to share their fears and frustrations.

But amidst this growing AI dependence we are all struggling with, I also heard from many founders who are getting better at identifying the best ways to use AI, basically out of necessity. Many of these founders seem to be coming from the agriculture industry where one founder explained, they’re still trying to explain the difference to clients who reflexively think AI means avian influenza. Once they’ve cleared up that confusion, they face even more skepticism. So they have to be extra careful before introducing it.

For Wiggs Civitillo , founder of Starfish, a startup which uses AI to monitor food recalls, regulatory changes, fraud and contamination alerts that then can all be connected to a company’s own supplier data, that means his work starts with accepting that people don’t want him to use AI. He appreciates that his clients don’t trust AI entirely and find it too expensive.

“There are certain parts of our platform that will never be well suited to generative AI, especially where we are looking at how we handle and protect customer data,” he said. “Leaving any of that open to exposure is not something that we’re able to do.”

In one case, Civitillo did not pursue one AI project using client data because he thought it would have hurt his relationship with his clients more than helped.

“It’s more of a deal breaker from a client respect and client protection perspective than it is an efficiency gain,” he said. “Making sure that they know exactly where their data goes and they know that their data doesn’t go anywhere at the discretion of some AI agent is very important.”

But Civitillo has found plenty of places where AI can be incredibly helpful, specifically in four areas that he identifies as “capturing, connecting, checking, and answering.”

“AI, unlike any technology before, is able to read messy, real-world data and normalize and standardize it into a usable format. PDFs, paper documents, handwritten receiving slips, emails, texts in English, Spanish, French, every language of the world with messy handwriting can now be processed in a meaningful way,” he said.

So he’s working with clients to process these documents and put them all in standard formats. He is also using AI to connect clients’ various programs and make sure that their work is correct.

“We only use AI where it removes real hours of manual work,” he added.

But Civitillo has found that what he has used AI for has produced real results. He noted that one specialty produce supplier client in California had staff enter about 100 orders a day at roughly 10 minutes each.

“Now their team takes a picture or forwards the document, Starfish’s AI reads it, and the order flows straight into their system,” he said. “That gives them back more than 3,300 hours a year, or roughly 65 hours a week of retyping (about the workload of one and a half full-time people), with fewer keying errors on top of it.”

Agriculture startup founders are also finding that AI can help produce solid first drafts for ideas and projects. Eddy Mejia, co-founder of biotech startup Earnest Agriculture, said that he has used AI to write draft grant proposals, startup applications, legal work and software applications for clients.

“I wrote a whole grant in five hours, at least the first draft,” he said before he then worked on edits. “Obviously we have to go back and spend another maybe three, five hours. But it would’ve taken us way longer if we just did it ourselves.”

Other founders like Michael Rhys , co-founder of Barnwell Bio, which monitors microbiome data in barns to preserve animal health, has found it useful for his engineers to use it for code development.

“That just makes our engineer even more efficient than he otherwise would’ve been without using AI,” said Rhys.

But as AI technology evolves, Rhys of Barnwell Bio has big hopes for how he can use it in the future. He would love to take all of this data he has to help predict how barns, particularly poultry farmers, may respond to things like unusual storms.

“Our view isn’t to just throw AI into it for the sake of doing it," said Rhys. "We really want to make sure that any use or any features we build using AI are going to help our customers solve their problems more efficiently, quicker, and make sure that we’re able to deliver even more insight and not just say that we’re using it.”