"I don't think this is a finite game, I don't think there is a winner. I think this is the classic Simon Sinek infinite game. I think the players are going to play until they run out of resources and they can no longer play the game." This assertion from Chris Hay, a Distinguished Engineer, set a provocative tone for the "Mixture of Experts" podcast's Thanksgiving special, which gathered industry leaders to dissect the hype and reality of AI agents. Host Tim Hwang, alongside Hay, Lauren McHugh Olende (Program Director, AI Open Innovation), and Volkmar Uhlig (VP Core AI and WatsonX.ai), debated whether 2025 truly marks the "year of agents," particularly in the realm of agentic commerce. Their collective insights painted a picture of a rapidly evolving landscape, where breakthrough AI capabilities are emerging, yet significant hurdles remain for widespread, consumer-facing applications.
The conversation kicked off with a deep dive into Anthropic’s new Claude 4.5 Opus model, a recent entrant that has quickly impressed experts. Mihai Criveti, another Distinguished Engineer specializing in Agentic AI, lauded its efficiency. "It's 50% more efficient than Claude Opus 4.1... it’s one of the most efficient models out there," Criveti stated, highlighting its superior performance for coding tasks, even surpassing recent releases from competitors like Google and OpenAI. This rapid iteration and specialization of models underscore the intense competition and breakneck pace of development within the AI space. The immediate impact of such models, particularly in code generation and testing, signals a significant shift in developer workflows.
However, the enthusiasm tempered when the discussion shifted to consumer-facing agentic commerce, especially the prospect of it dominating shopping events like Black Friday 2025. Chris Hay was unequivocal in his skepticism. "No, I don't think it will be," he declared, suggesting that the industry is "probably another year away" from such a breakthrough. While companies like OpenAI are integrating shopping capabilities and partnering with platforms like Shopify, the sheer number of retailers yet to be onboarded, coupled with geographical limitations (currently US-only), indicates a nascent stage of development. Google's agentic commerce protocol faces similar early-stage challenges.
The true impact of agents, according to Volkmar Uhlig, lies not in direct consumer interaction, but in the backend. He argued, "I think the agents are not the consumer-facing agents, but the agents are actually the backend." Uhlig pointed to existing optimizations in areas like returns processing—a significant pain point for retailers, especially after major shopping holidays. These agentic workflows, though invisible to the end-user, are already streamlining operations and reducing labor costs for major retailers. This distinction is critical: while consumers may not yet interact with sophisticated shopping agents, businesses are increasingly leveraging AI behind the scenes.
Lauren McHugh Olende echoed this sentiment, questioning the perceived value of simply automating the checkout process. "I don't see a big revolution coming from the simple act of being able to check it out once you're in the AI application," she noted. She emphasized that the real challenge lies in the quality of product research and matching capabilities within AI models. Current general-purpose chatbots often fall short in understanding nuanced customer intentions or specific product attributes like size, dimensions, or style. This highlights a need for AI models trained specifically on vast datasets of e-commerce interactions and product information to truly enhance the shopping experience beyond basic automation.
The developer ecosystem for building these agents is vibrant, with numerous tools and frameworks emerging. However, the path from prototype to production remains arduous. Uhlig articulated this challenge, stating, "The Shopify moment for agents hasn't happened." This refers to the lack of an easily accessible, low-code platform that democratizes agent development and deployment for small to medium-sized businesses. Current agent development still requires a high degree of technical expertise, necessitating complex orchestration of multiple models and external tools.
Related Reading
- AI Agents Transform Holiday Shopping into a Transactional Battleground
- AI to Drive Market Higher, Deflate Costs Through 2026
- AI Shifts from Build-Out to Utilization: A New Investment Landscape
The complexity of managing agents, ensuring they stay "on track" and don't "go off the rails," is another significant hurdle. Agents often need guardrails, deterministic flows, and robust planning modules to execute tasks reliably, especially in real-world consumer scenarios where unpredictable inputs are common. This demands sophisticated frameworks and careful integration, moving beyond simple API calls to a more structured and controlled environment. The current state is akin to a "baby programmer interface," as Uhlig described it, suitable for those who can code but not yet for mass adoption by non-developers.
Despite these challenges, the panelists agreed that agents are indeed here, albeit in a less glamorous, more infrastructural role than often portrayed. They are woven into existing digital experiences, often unnoticed by the end-user. The future of agentic commerce, therefore, hinges on making these powerful tools more accessible, more specialized for retail, and more capable of reliably executing complex, multi-step tasks without constant human oversight. The market is ripe for innovation that bridges this gap between advanced AI capabilities and seamless, dependable commercial application.
