AI can help startups to accelerate, but business leaders should be mindful that gaining efficiencies isn’t always as direct as the hype suggests.Getty Images
Diana Birsan had to learn some AI lessons the hard way.
In 2023, she co-founded Ottawa start-up Downpay with Denis Zgonjanin. The app allows Shopify merchants to offer customers the option to make deposits and partial payments for bigger purchases. When Downpay had only two full-time and two part-time staff, Ms. Birsan, the firm’s CEO, launched an AI chatbot to help with customer support.
Instead of delivering efficiency, she says the chatbot frustrated many customers by giving irrelevant information or incorrect answers, such as claiming support for currencies the app didn’t offer. “It was just embarrassing. It did more harm than good.”
Ms. Birsan worried about Downpay’s reputation. Support requests often popped up when merchants had just installed the app or something had gone wrong in their payments system. “If you mess that up, they’re just going to uninstall. It’s a critical moment.”
Downpay deactivated the chatbot within a month. AI now plays a far narrower role in the company’s customer support: drafting suggested replies that a support representative checks for accuracy. “She’s able to parse it for truth before it goes out to the customer,” Ms. Birsan says.
In the early days of Downpay, Ms. Birsan used AI to try to improve the company’s marketing, but she says the text felt mediocre and machine-written.
Businesses need to pause the AI rush and ask a basic question: “Does it make it any easier?” says Dr. Majid Komeili, an associate professor of computer science at Carleton University and director of its Intelligent Machines Lab.
He cites the example of a hotel that replaces a straightforward booking page with a chatbot. A process that once took seconds can turn into a winding back-and-forth conversation. To Dr. Komeili, some businesses might be feeling AI FOMO (fear of missing out). But he says you need to examine whether AI improves an experience enough to justify its cost and complexity.
Downpay, which Ms. Birsan says has been used by customers of 1,600 brands on Shopify to help pay for $500-million in purchases, found better AI use cases for some internal tasks. She has to wear many hats at a small start-up. Ms. Birsan describes herself as a product manager “pretending to be a designer.”
AI has dramatically sped up her work on user interface (UI) design. In the past, she says building even a basic design required time-consuming manual work in HTML and CSS that she calls “slow and painful.”
Now she can use tools such as Figma’s AI features and Claude to generate UI mock-ups and prototypes in less than an hour, then hand them off to developers. She says the AI models tend to default to uniform styles, but even an AI-generated version of a design has been saving her a lot of time, freeing her up to focus on her company’s growth.
As someone who worked in IT security for Shopify earlier in her career, Ms. Birsan worries that many companies are experimenting with AI without understanding the privacy and data-leak risks. “If you’re on a free plan, don’t share data that’s sensitive,” she says. Even with paid AI tools, she adds that teams need to carefully think through the security implications.
Dr. Komeili says businesses should be proactive with AI adoption, but they should also set strong internal policies, or they risk employees using “whatever random AI tools are out there,” increasing the chance of data leaks.
For many entrepreneurs, it can be hard to proceed cautiously when you’re surrounded by AI hype, Ms. Birsan says. “The worst thing is this feeling that you’re going to be left behind.”
She says not to assume everyone else is on top of AI innovation and implementation and you’re alone. “All of us are just figuring it out as we go.”