For the last 6 months, we've talked a lot about how personal agents will change the nature of customer support.

Individuals with access to relentless, 24/7 agents are going to unleash a tidal wave of questions onto organisations. The natural instinct of most organisations will be to ban or deflect agents. We think that is a mistake.

Organisations and brands that set up systems to help and resolve queries from personal agents will attract even more agents and, simply, get more business.

Last week, we saw the first breakthrough of personal agents into the mainstream.

On September 6, 2026, JC Bahr-de Stefano posted a screenshot of an email from Resy. The reservation platform had deactivated his account and canceled every future reservation on his account.

Bahr-de Stefano had asked Instinct, an invite-only personal AI agent, to get him a table at 4 Charles Prime Rib in New York. The agent's activity log showed it sent Resy about 200 API requests an hour. It checked availability every 10 minutes through the day, and every 0.4 seconds around the time the restaurant released tables each morning.

From Resy's side, an agent that polls every 0.4 seconds looks like the scalper tools people use to resell tables. Bahr-de Stefano saw it the same way once he read the log: "Of course, this account got flagged for spam, because it was acting like a bot." Resy reinstated the account on September 8.

Great experiences remain the north star - agents or not

Resy is an early sign of the new agentic economy.

Reservations got there first because the prize is obvious (exclusive bookings!) and the target is one API. Personal agents are already here for consumer businesses but this is coming for everyone.

The best businesses will start updating their terms of service and creating the guardrails for personal agents to engage constructively. As Bahr-de Stefano says "most of these sellers or providers need to get to a point where they can understand the difference between a verified agent acting on someone's behalf versus a bot farm."

We think the answer is an AI concierge equipped to handle personal agents.

If Bahr-de Stefano's agent could ask and be told the rules, it wouldn't need to poll constantly. If it was a verified personal agent, Resy may have helped it achieve its goals instead of instituting a ban.

When an agent asks a clear question, your own AI concierge should give it an accurate answer within seconds, under the same policies and identity checks a human would get. Your concierge can set request limits, tell agents when to check back, and explain the rules. The agent can still achieve its goals but within appropriate constraints.

We're building for this at Lorikeet now. If you run a consumer business and more of your traffic or contacts come from AI agents, or you expect them to within the next year, we'd love to hear what you're seeing.