What Downpay learned from a support AI chatbot that lasted one month
Downpay co-founder Diana Birsan was featured in The Globe and Mail on the AI experiments that worked, the ones that didn't, and why the difference matters more than the hype.
The honest version of an AI story
Downpay co-founder and CEO Diana Birsan was featured in The Globe and Mail this week, in a piece by Nickie Shobeiry on how AI shortcuts can become detours for small businesses.
The article opens with Downpay's own story. Early on, with four staff total, the team launched an AI chatbot to handle merchant support. It gave wrong answers, including claiming the app supported currencies it didn't. Diana pulled the chatbot within a month.
"It was just embarrassing. It did more harm than good," she says.
Merchants reaching out have often just installed the app or hit a payment issue. A bad answer at that point is costly in a way a bad answer in a lower-stakes context isn't.
Today, Downpay uses AI to draft suggested replies that a support team member reviews before they go out. One human step makes the difference.
Diana also talks about where AI has genuinely helped: building UI mock-ups in Figma and with Claude, collapsing what used to take days of manual HTML and CSS work into under an hour. That time goes back toward building.
The piece also covers data privacy. Diana's view is that teams need to think through what they're putting into AI tools, especially on free tiers, before they start. Security implications deserve more attention than most companies give them.
On the broader pressure to adopt AI quickly, Diana is direct: "All of us are just figuring it out as we go."
Read the full piece at The Globe and Mail.

The humans at Downpay celebrating 3 years in business.