Most lawyers experimenting with AI are making one of two mistakes, according to Samuel Puchala, President and General Manager of LexisNexis Canada. One group briefly tries a tool, gets a disappointing result, and shelves it. The other dives in, gets impressive early results, and over-relies on the output. "There's a happy middle ground," he says, "but you have to keep your professional judgment." In this episode of CL Talk, Puchala offers a framework for thinking about AI that any lawyer can apply immediately. He compares working with AI to managing a young lawyer: the leverage is real, but the instructions matter and the output always needs review. That means knowing what the product is actually designed to do — using it outside those parameters is one of the most common sources of frustration — and being specific with prompts. Vague instructions, he says, produce the same result as sending an articling student out with ambiguous directions. On accuracy, Puchala draws a distinction that often gets lost. A verified citation is not the same as a correct analysis. Within a grounded legal AI platform, hallucinations are rare, he explains — but whether the cited authority actually fits the facts at hand is still a judgment call that belongs to the lawyer. The conversation also gets into agentic AI — what it means beyond the buzzword. "With agentic AI, you can give it more of an objective, and it can work through the multiple steps towards that objective rather than have you direct its every move," Puchala says. And in procurement, law firms that are still evaluating legal AI tools sequentially are already behind. "By the time you're looking at that second, third competitor, the first one's already moved several generations ahead." Puchala's path to LexisNexis ran through pharmacology, strategy consulting at Bain & Company, and 16 years as a partner at Satov Consultants — an unusual background for a legal tech executive, and one that shapes how he thinks about what lawyers actually need from software. Puchala will take part in a panel on getting reliable results from legal AI — covering prompts, workflows, and emerging agents — at the Canadian Legal Summit on October 14 at The Carlu in Toronto. Register now: https://events.canadianlawyermag.com/canadian-legal-summit?utm_source=newsletter&utm_medium=editorial_article&utm_campaign=CLS2026&utm_content=awareness&utm_term=asset01