Artificial intelligence had been on Shraysi Tandon’s mind since the day she co-launched Kidsy, an online reseller of discounted baby and children's clothing, toys and products.The Chicago-based company launched in September 2023, and its prices are determined by AI tools, which scour the web to ensure Kidsy’s prices remain competitive. The technology helps the company source open-box items, products that were opened and returned but often in near-new condition. AI tracks the retailer's warehouse inventory and flags any transactions that may be fraud or high-risk — advising warehouse workers to cancel the order.“As soon as we launched Kidsy ... we saw that there were many gaps within the business that would be far more efficiently fulfilled by the use of artificial intelligence rather than any human labor,” Tandon said.A year after launching, AI has allowed Kidsy to grow sustainably, reduce labor costs around repetitive tasks and fulfill its niche of providing parents open box and returned items from major brands at a discount, diverting the products from landfills.She said the investment into AI — paying software engineers to build custom solutions and developing the data and infrastructure — has more than paid off.“AI has actually reduced costs for us in the long run because it's increased operational efficiencies,” Tandon said. “[It's increased] the speed at which we can move; the speed at which we can respond to customers; the speed at which we can fulfill transactions and has enhanced our end-to-end customer experience.”When AI exploded in popularity in 2022 with the introduction of platforms like ChatGPT, businesses and employees anticipated the technology would revolutionize the workplace — even replace some workers. Despite the fanfare, businesses have been slow to adopt the technology because of its high cost and a lack of ready-to-go solutions. In Chicago, companies like Kidsy have found AI to be worth the investment and time but experts said companies should proceed with caution in a fledgling space lacking widespread regulation.A March report by the U.S. Census Bureau found the number of firms nationwide using AI was "relatively low but rising." It surveyed an estimated 1.2 million businesses from September 2023 to February and said in that time, the use of AI increased from 3.7% to about 5.4%. Most companies used AI for chatbots, automating marketing tasks and analyzing data, according to the bureau, and it didn't impact employment.Restaurateur TJ Callahan, who co-owns Farm Bar in Ravenswood and Lake View, deployed an AI phone answering service for his restaurants this summer. He added the tool after fearing the farm-to-table restaurant might be losing out on business.“I would go into the restaurant and look at the call logs and there'd be several phone calls during the day,” when the restaurant was normally closed, Callahan said. “And when you called the guests back, they don't pick up.”After about a week of setup and a few hundred dollars in monthly fees, the AI service was answering the phones at Farm Bar and answering questions about restaurant hours, menus and reservations.But the tool has its limits, Callahan said. For example, it can't tell a customer if the rooftop is open based on the weather, a variable it's not yet able to account for. But with rising labor costs, he said AI could play a larger role in the future.“Once the guest is in the restaurant, sitting at the bar, AI has no rule,” Callahan said. “But you wonder, down the road, will AI play a bigger role?” Customers chat outside of Farm Bar’s Lake View restaurant at 1300 W. Wellington Ave.Zubaer Khan/Sun-Times Balancing risk and rewardDeming Chen, professor of engineering at the University of Illinois Urbana-Champaign, said businesses adopting AI can face a few hurdles.It can be pricey to employ software engineers to develop AI models. And using third-party AI tools can take extra time to plug into a business's database and software systems. Questions about ethics and a lack of federal regulation can further complicate things.“I really advocate for openness,” Chen said, who occasionally uses ChatGPT. For example, businesses should disclose if any user-generated data will be used for training AI models. “It's a balance between innovation and risk."While the promise of what AI can do is great, Chen said users shouldn't "blindly trust" that all AI models are accurate. He said businesses need to rely on self-imposed standards to mitigate potential conflicts and retain consumer trust.“As long as we have a clear policy and standardized procedures with strong regulations, I feel the generative AI, large language models and machine learning models are going to help us instead of hurt us.”Creating a custom fitHeather Denniston, president and co-founder of Ludex, an app that scans and identifies sports trading cards, encountered two hurdles in building out AI for the startup: time and money.Ludex first developed basic AI for image and text recognition, which compares a scanned card to a database of collector's items and provides details. But it was easily stumped by slight variations in card designs or reflective finishes and required more work, Denniston said.“The technology is frankly just incredibly expensive to create,” she said. “It becomes less and less expensive as you maintain it, but we're still in the creation mode. Soon, we’re going to be at a nice break even point and we can start enjoying the fruits of the labor. But financially, that was our first roadblock.” Heather Denniston, chief operating officer, president and co-founder of Ludex, inside the company’s River North headquarters at 223 W. Erie St.Ashlee Rezin/Sun-Times Denniston said the second roadblock was designing AI programs that could identify modern cards. Before Denniston and her team could start building identification algorithms, they had to find massive amounts of data to feed them — and that took time.“It took about a year and about a million dollars for us to solve base cards, which are just the simple cards, the vintage cards," Denniston said. "Those were kind of easy. It's the modern-day cards, which took us another year and $8 million to solve.”Now, Ludex has more than 1.5 million app downloads since its launch in November 2022.The company plans to build out its own custom AI that can identify a trading card and use sales data to estimate how much the card is worth.The investment into AI paid off but it wasn’t cheap. Denniston and co-founder Brian Ludden raised nearly $14 million to cover startup costs and a team of software engineers.“Chicago is such a great place to build technology and to be an entrepreneur,” Denniston said. “I would like to see more support of entrepreneurs in the city, especially female founders and minority founders. I wish there was a little bit more attention on what's being built here in the city.” Inside the River North office of Ludex.Ashlee Rezin/Sun-Times Karen Plesh, CEO of property management firm 29th Street, said potential tenants can chat or text with an AI-powered phone service for information on available units, apartment amenities and how to set up a tour. Similar technology guides prospective residents through self-guided building tours.“Our team members feel that they are really more focused on providing optimal customer service and ensuring that the property is well taken care of,” she said. “They have time to talk to their contractors, to walk the property, to work with their maintenance team to make sure everything is in tip-top shape.”Plesh hopes to glean more predictive insights from AI eventually. For example, getting a notification about a building with consistent air conditioning issues and recommending an overhaul.Ethical concernsA 2023 report from the Small Business and Entrepreneurship Council found the median annual investment in AI-powered tools by small businesses was $1,800, with nearly half of owners surveyed saying they plan to increase their investment in AI in the coming year.Steve Banke, vice president of transformation at IT consulting firm Ntiva, said the firm hasn't seen widespread adoption with most companies "just getting started." He expects that to change but warned the technology isn’t yet reliable.“AI can be confidently wrong,” Banke said. “Removing the human element from understanding the quality and the validity of the data is a serious concern. Business owners should make sure there is written policy in their organizations that AI content is not simply created and distributed or relied upon in any way.”Data privacy should also be a top concern, said Rita Garry, an AI and data privacy attorney at Howard & Howard.“Companies need to know the legal landscape ... and how they're going to document and prove that they're actually in line with those consent requirements,” Garry said.At least 25 states, Puerto Rico and the District of Columbia introduced AI bills last year, according to the National Conference of State Legislatures.“One of the main concerns of AI is that it's a machine learning model that's based on data that's been fed to it. And if the data isn't clean and the learning process of the model isn't supervised, it can produce discriminatory outcomes for decision-making,” Garry said.For example, using AI software to screen job applicants' resumes could leave a business open to accusations of employment discrimination if the AI was trained on biased information, Garry said.In 2019, Illinois was one of the first states to enact restrictions on the use of AI in hiring when it passed the AI Video Interview Act. And HB 3773 will restrict employers from using AI to inform decisions on hiring, promotions and other job-related functions starting in 2026.Garry advises businesses to think critically about how AI can fit into their operations, carefully evaluate any third-party AI tools and continuously test and refine the model.Banke, of Ntvia, said businesses should learn study the technology before making an investment.“I think people need to have the most basic understanding of AI before they dive into it for their business,” Banke said. “AI is just a software program. That's all it is. It's going to do what you tell it to do. And understanding how to tell it what to do is the real key to being successful with it across the board.”