When the buyer is a bot: Keeping your startup eligible in agent-run procurement

The last two articles assumed a human on the receiving end of AI-generated content. However, this assumption is changing. In ESBO Ltd, a leading agency, client conversations now focus on whether an AI could buy from them instead of just reading about the company. The distinction may seem academic at first, but observing an agent's behavior reveals its limitations.

Unlike humans, agents do not read company homepages nor forgive gaps that a human would overlook. For instance, Salesforce found that AI influenced 20% of global online sales during the 2025 holiday season, amounting to around US$262 billion. Retailers with their own shopper agents grew sales 59% faster than those without. AI shops for people in a literal sense, and protocols to facilitate this have been developed, including OpenAI’s Agentic Commerce Protocol with Stripe.

However, fully autonomous checkout has not become the norm yet. Instead, agents research and shortlist products, then hand the final decision to a human. Adobe’s data shows that AI-referred traffic converted 42% better than other traffic in March 2026, a significant improvement from the previous year. B2B purchases are also expected to move towards AI agents by 2028, handling trillions of dollars in machine-to-machine transactions.

The strictest buyers are AI agents, which evaluate vendors based on verifiable facts and reject any friction that a human would tolerate. Agents focus on public, structured sources, disregarding any discrepancies between websites, directory entries, and press coverage. To cater to AI buyers, startups should publish clear facts, such as pricing bands and specifications, using structured pages and schema markup.

This approach ensures that agents can easily verify the information and shortlist the most eligible vendors. Although agents do not replace human relationships in closing deals, they determine which startups get to the final handshake.