Unlike B2C purchases which are often low-cost and for leisure, B2B transactions are expensive and complex. This makes them ideal for AI agents designed to save time and streamline multi-step processes, rather than for "killing time" like consumer shopping.
Alibaba.com defines a successful AI agent as the product of three essential components: the core intelligence (Model), the proprietary logic and tools (Harness), and real-world business data (Context). The multiplicative relationship means all three are critical; if one is weak, the entire agent fails.
To evaluate AI for commerce, Alibaba developed a custom benchmark using 107 real business tasks. It goes beyond simple accuracy, measuring agents on pass rate, completion time, and crucially, cost. This ensures the solution is not just effective but also affordable and useful for small businesses.
To overcome the slow pace of its 26-year-old legacy platform, Alibaba.com ran its AI agent, Axio, as an internal startup. This model allowed the Axio team to ship new versions daily, fostering rapid innovation without disrupting the core business, which released updates bi-weekly or monthly.
Alibaba.com initially launched its Axio product as an AI search engine. However, observing users employing it for unintended tasks like market research and creating professional design packs was the key signal to pivot and expand Axio into a comprehensive e-commerce agent.
Kuo Zhang, president of Alibaba.com, states that the biggest challenge in the AI era is the sheer speed of change. Traditional three-year business plans are no longer viable because "three years means forever." This uncertainty requires a shift towards more adaptive and shorter-term strategic planning.