\n<\/p>\n
On April 22, the\u00a0<\/span>AI Horizons<\/span><\/a>\u00a0webinar\u00a0series from\u00a0<\/span>Wharton Human-AI Research (WHAIR)<\/span><\/a>\u00a0examined why AI agent adoption keeps stalling,\u00a0and what the research says leaders should do differently. In this episode,\u00a0<\/span>Stefano Puntoni<\/span><\/a>,\u00a0faculty co-director of\u00a0WHAIR\u00a0and\u00a0the\u00a0Sebastian S. Kresge Professor of Marketing at the Wharton School,\u00a0joined\u00a0<\/span>Thomas McKinlay<\/span><\/a>, founder of\u00a0<\/span>Science Says<\/span><\/a>, to present the\u00a0<\/span>Wharton Blueprint for AI Agent Adoption<\/span><\/a>. Drawing on academic research and expert interviews with leaders at Google, ServiceNow, Zapier, and\u00a0Workato, the two unpacked the psychological barriers slowing adoption \u2014 and the design strategies that can overcome them.<\/span>\u00a0<\/span><\/p>\n AI agents have advanced rapidly, but adoption in organizations continues to lag. The bottleneck, according to\u00a0Puntoni\u00a0and McKinlay, has shifted from what the technology can do to whether people are willing to use it. Employees wrestle with three core psychological frictions:\u00a0whether the agent is perceived as\u00a0competent, whether they can trust it, and whether\u00a0they\u2019re\u00a0comfortable delegating control to it. Until those frictions are addressed, even the most capable agents will sit unused.<\/span>\u00a0<\/span><\/p>\n \u201cThe problem in the context of AI agent adoption is becoming, in a way, less and less the technology, and more and more the psychology.\u201d \u2014 Stefano\u00a0Puntoni<\/span>\u00a0<\/span><\/p>\n Counterintuitively, making an AI agent overly conversational or personable undermines users\u2019\u00a0perception\u00a0of its\u00a0competence,\u00a0especially when\u00a0it\u2019s\u00a0being asked to\u00a0take action, not just respond. Research supports designing agents to\u00a0communicate in ways that signal capability: explaining the criteria and processes behind their\u00a0outputs, and\u00a0providing reasoning traces rather than just conclusions. The goal is not a friendly assistant, but a credible one.<\/span>\u00a0<\/span><\/p>\n One of the most effective,\u00a0and underused,\u00a0strategies for building user trust is telling people upfront what the agent cannot do. Doubt about an agent\u2019s data access, accuracy, or scope is what erodes trust quickly. Being explicit about limitations removes that uncertainty before it calculates into skepticism. Research also found that using precise numbers rather than rounded figures increased perceived trustworthiness by a meaningful margin,\u00a0signaling care and rigor in how the agent reached its output.<\/span>\u00a0<\/span><\/p>\nThe adoption gap is a psychology problem, not a technology problem<\/span><\/b>\u00a0<\/span><\/h3>\n
Designing an agent to seem warm and friendly can backfire<\/span><\/b>\u00a0<\/span><\/h3>\n
Transparency about limitations builds more trust than projecting confidence<\/span><\/b>\u00a0<\/span><\/h3>\n
AI agents should be co-deciders, not autonomous actors<\/span><\/b>\u00a0<\/span><\/h3>\n