The AI gold rush has hit marketing, and every CMO is digging for an edge. But for every nugget of genuine value, there’s a mountain of fool's gold—overhyped tools and ROI mirages. For now, most teams are in a deliberate test-and-learn phase, betting today’s GenAI trials will set tomorrow’s standards.

Ankit Bansal, Forbes 30u30 and Co-Founder and CTO of Sortment, ShopFlo, and Flo Labs, has seen this phase up close. He offers a clear-eyed look at how marketers can separate signal from noise, and where AI is already delivering real returns.

Explore vs. exploit: "If you don't try AI, you’ll never know what works for you. We're in a classic 'explore versus exploit' phase right now; most of us are in that brief period of exploration," says Bansal. "But within the next three to six months, exploitation is going to happen—people will realize what AI actually works for, and they're going to reap the benefits." In other words: experiment now, or risk falling behind.

Mission control: "Marketers feel very uncomfortable with giving the entire control to an AI system," observes Bansal. While many are leveraging tools like custom GPTs, "they do not understand how the black box works, and they feel it can make a lot of mistakes," Bansal explains. "They want to supervise and control it."

Bells and whistles: The desire for supervised augmentation is further complicated by practical limitations. "The native platforms that marketers work on are slow in adopting AI," says Bansal, forcing users out of their typical workflows. "They have some bells and whistles, but these platforms are not using AI capabilities to augment in their existing flows." He points to the common workaround: marketers copy email text into ChatGPT for review because their primary tools lack these capabilities.