Building AI Agents by Removing Features
Henry Hopcroft · Chief Product Officer · Gradial
At Gradial, Henry Hopcroft is rethinking a familiar product question: what should an AI team build next? The Chief Product Officer explains why his team is removing hardcoded agent features, learning directly from enterprise customers, and focusing on the core capabilities that let agents handle more kinds of work.
Henry describes why Gradial has been moving away from individual agent features toward a small set of flexible core tools. As agents become more capable, he argues, the range of tasks they can perform increasingly depends on customers' needs and imagination rather than a fixed list of product features.
Enterprise environments make AI agents difficult to test. Gradial builds adversarial internal evaluations based on actual customer failures, including unreliable systems and unexpected handoffs. Its forward deployed engineers work directly with customers, then turn what they learn into more generalizable improvements for the platform.
The conversation also explores the growing overlap between product and engineering roles. Henry explains why coding agents make product judgment and understanding customer experiences more valuable, and why long term roadmaps are harder to justify as AI capabilities and shipping speed evolve. At Gradial, clear product vision and rough interactive prototypes help keep teams moving in the same direction.