Ovetta Sampson is a design researcher, AI leader, and founder of Right AI. She previously served as VP of ML and AI Platform Design at Capital One and worked at Google and IDEO. Ovetta brought that experience to the 2026 ITX Product + Design Conference, focusing her keynote on a question that often goes overlooked: what happens when people interact with increasingly powerful machines? To help us answer that question, Ovetta offers a pair of frameworks that center on human engagement risk, responsible AI, and rethinking how product teams design AI products. Building AI responsibly requires more than better models or more sophisticated tools. Product builders must also understand the cognitive, social, cultural, and physical risks that can emerge when humans interact with technology. AI inherits many of the biases embedded in the data used to build it, Ovetta says. So organizations need to rethink how disciplines collaborate; as the lines between traditional product, design, engineering, and security silos blur, responsible AI requires organizations to redesign not only their products, but also the processes used to create them. Here’s what else we learned: Human Engagement Risk Belongs in AI Product Design Ovetta’s human engagement risk (H-E-R) framework is based on a fundamental premise: technology can harm people when designers fail to account for how humans behave around machines. The H-E-R framework identifies cognitive, social, cultural, physiological, and community risks. These risks become especially important with generative AI, where people can easily attribute human qualities to systems that do not actually possess them. As a result, AI product teams must consider psychological and cognitive outcomes alongside traditional usability concerns. Ovetta cautions: “There are real dangerous risks when we engage with machines and don’t mindfully think about the outcomes that can happen when we don’t protect humans psychologically, cognitively, physically, and physiologically.” AI Strategy Starts With Executive Leadership Responsible AI also requires leadership decisions that extend beyond individual tools or experiments, Ovetta says. Many mid-sized organizations are hesitant to adopt AI because executives are concerned about intellectual property, trust, and data leaks. Meanwhile, employees may already be integrating AI solutions without an overarching organizational strategy. It’s a disconnect that creates opportunity for leadership to establish clear priorities before adoption becomes fragmented. AI strategy starts at the top, Ovetta adds, because executives have the authority to establish the conditions under which technology gets developed and used. “Once the C-suite understands the risk to their shareholders, to their products, to their employees, to their customers, it is much easier for me to bring in the implementation of how to mitigate those risks.” Dismantling Silos Is Essential for Effective AI Development AI challenges the traditional handoff model in which designers, engineers, security, legal, compliance, and other teams work separately before passing projects along. Ovetta says AI development requires continuous cross-functional input instead. Her D-C-R framework – draft, critique, revise – organizes teams around development stages, bringing the right expertise into each phase. “Instead of saying, ‘I’m a designer’ or ‘I’m a researcher,’ or ‘I’m a product manager,’ or ‘I’m an engineer,’ we say, ‘I’m in the draft mode,’” Ovetta adds. “Each skill set in that move brings what they need to get that draft ready for critiquing, right? And so it’s something that I give to organizations and teams to try to reimagine how they actually do their jobs.” Ovetta Sampson is not arguing for less innovation with AI; instead, she’s arguing for a different definition of responsible innovation – one that embeds human consequences, executive accountability, and cross-functional collaboration into the product development process itself. [03:10] Protecting the fragility of humanity. There are a lot of things us humans engage in, especially what I call the cognitive biases, that make engaging with machines and other automated systems that make it risky for us. [05:57] The H-E-R Framework. But what it really is, is there are five dimensions. There’s the cognitive, there’s the social, there’s cultural, there the physiological. And then the overall community risks that when humans engage with machines, that can happen. [10:08] LLMs built on ‘traumatized data sets.’ Generative AI has no moral code. It does not know truth or fact. And accuracy is not in its wheelhouse. In fact, it’s not in this training and it’s in its goals. So why when we type something into chat GPT, we expect truth back? I don’t know. [13:24] Protecting my values as a creator. That’s where I really want to start, because I don’t want to be a part of that. I don’t want to part of designing something that harms people. [14:00] I observed one reoccurring truth. Whatever is happening in the basement of a company starts in the C-suite. So if there is sexism, if there’s homophobia, if there racism, if there is bad culture, if it starts at the C-suite. Because the C-suite is the person who has the ultimate authority about what occurs in every floor of an organization. [18:26] The D-C-R framework — draft, critique, revise. The D-C-R framework is something I created because I was trying to explain to designers and product and engineers how their processes would change when they’re designing for and with AI. Each one of these disciplines should go through that makes the handoffs more like a circular iterative. The post 194 / Ovetta Sampson: Designing AI Products Around Human Needs – Not Just Technology appeared first on ITX Corp..