by Video interview on April 22, 2026
In this episode, A Coffee With… Gemma Spence, CEO of VML Marketplaces (WPP Group), joins ExchangeWire COO Lindsay Rowntree for a coffee chat about AI and agent commerce.
Gemma explains how the focus is moving from keyword-based strategies to prompt-based strategies and metrics such as model share and share of mentions in AI-generated answers. Their conversation delves into advertisers’ concerns about agent commerce, its impact on the consumer decision-making process, and how brands can optimize for LLM and measure results.
To succeed, you need to augment traditional marketing with a data-first approach. This means prioritizing high-quality structured data and committing to large-scale testing and rapid engineering to decipher AI behavior. The field is still maturing, but early, structured adoption focused on clear use cases is key.
What is agent commerce?
Agenttic commerce describes a shift in online shopping in which AI-powered agents act as “personalized shoppers” and interact with retailers and brands on behalf of consumers. This creates machine-to-machine interactions that fundamentally change the dynamics of traditional product discovery and purchase.
Gemma highlights the three ‘C’s’ of advertisers’ concerns when it comes to agent commerce.
- control: How to influence agent shopping decisions
- Clarity: Limited transparency into how AI models like Rufus and Gemini prioritize and rank products
- commoditization: Concerns that brand value will be undermined as functional, data-driven factors dominate purchasing decisions.
The purchase funnel still exists, but it has been enhanced. The core stages remain, but influence tactics are adapting to AI-driven discovery. Optimizing agent commerce is complex and testing-intensive because AI behavior is shaped by many opaque factors. There is no silver bullet. Brands need to conduct continuous testing and engineer quickly to figure out what works.
Strategies for success with agent commerce
The first step is to ensure complete, high-quality structured data, including consistent product naming, feature-driven bullet points, and a clear price pack architecture. If this basic “digital shelf” data is weak, subsequent optimization efforts will perform poorly.
Brands also need a systematic testing approach to analyze category dynamics, competitor behavior, and key levers such as pricing and promotions. Through extensive instant engineering, brands can see how their products stack up against the competition and identify the signals that drive AI recommendations.
Finally, campaigns must engage human emotions and satisfy machine logic. Brand creativity and salience are still essential to the end user, but the direct “customer” is the AI agent. Brands need to blend functional brand building with a focus on structured data, authority-building content, and media overlays to ensure visibility within the AI decision-making path.
The future of measurement and deployment
While traditional metrics like ROAS and impression share are still important, measurement has evolved and become more important. model share (Brand penetration in certain AI models like Amazon’s Rufus) share of mentions (Recommended frequency in AI-generated answers). The goal is to be the preferred choice. Lack of mention means lack of consideration.
Adoption is two-pronged. Fast-moving companies in fashion, beauty, consumer electronics, and more are leading the way by aggressively testing and setting standards. Meanwhile, many consumer goods brands are still in “wait and see” mode due to complex B2B structures and slow ROI validation.
Either way, agent commerce is expected to be a disruptive force in all sectors. According to Gemma, building AI readiness will not be one person’s job, it will be everyone’s job.
Agentic AIAgentic Commerce
