Behind AI's slick demos and sky-high valuations lurks a growing wave of tech that doesn't or can't work as advertised. Companies in the market for new solutions need to cut through the flash, scrutinize bold claims, and sift through the buzzwords to find the tools that deliver real value.

Ethan DeWaal, Strategy & Operations Lead of Asana's AI Studio, joined us to talk about avoiding vendors who are "heavy on the buzzwords, light on substance"—and why execution, not potential, is what really matters.

Can vs. will: "So many people are selling snake oil right now," says DeWaal, adding that false promises abound because "it's easier than ever to build a really slick marketing website."

While there are two evaluation factors for agents, most people only focus on one. "One is capability—what can the agent do?" explains DeWaal. "The other is reliability—what can the agent do consistently, every time?" Reliability, says DeWaal, is where many agents fall short. The difference between what an AI can do and what it does reliably is where many businesses get tripped up.

Red flags: "The second you're on a call and [a vendor] can't clearly explain the use case or value they offer today, you're probably talking to the wrong vendor," warns DeWaal. Too many are selling partnerships to co-develop a vision, not a product. And worse, some are designing for models that don’t even exist yet. "The value is not available for you today," he says. "They’re designing for capabilities that current models simply don’t have."

Raising the bar: The widespread uncertainty fuels a hunger for concrete, present-day results that shapes how enterprises vet AI tools. They demand clear success metrics, transparency, and provable ROI. "[For my team at Asana], we need 3x ROI on any AI software purchase, and we need to see that ROI before we purchase it," says DeWaal.