In May 2026, Google launched an always-on AI agent with a payments framework attached. Users can set budgets, approve merchants, and let software shop within those rules. OpenAI, Anthropic, Microsoft and Apple are all building toward the same capability. Whatever the rollout timeline, the direction is settled: a growing share of purchasing decisions, consumer and B2B, will be shortlisted or executed by software acting on human instructions.
Boards do not need to understand the technology in depth. They need to understand that this is a channel shift, and channel shifts are squarely a governance concern, because they redistribute revenue between the prepared and the unprepared. Here are five questions worth putting to management, and what a good answer sounds like.
1. Can a machine buy from us today?
Not “do we have a digital strategy.” Literally: if an autonomous agent attempted to evaluate our offer and complete a purchase or enquiry right now, would it succeed? Where exactly would it fail?
A good answer names the failure points: the pricing that is not machine-readable, the checkout step that blocks automation, the product data gaps. A bad answer is a general assurance about digital maturity. If management has never tested this, that itself is the finding.
2. Who owns machine legibility?
Agent readiness cuts across ecommerce, IT, marketing and data teams, which in most organisations means it is owned by none of them. Ask for a name. Ask what that person is measured on. Ask what budget follows the accountability.
Shared ownership of an emerging channel is how incumbents lose emerging channels. The board minutes from retailers who were slow to ecommerce in 2005 make instructive reading here, mostly because the topic barely appears in them.
3. What is our exposure if agents cannot reach us, or misread us?
There are two distinct risks and management should be able to speak to both. Availability risk: many businesses currently block automated traffic indiscriminately, which was sensible when bots were scrapers and is self-harming when bots are buyers. Accuracy risk: an agent that misparses your pricing, terms or stock does not shrug like a confused human. It records the wrong answer and decides accordingly, at scale, invisibly.
Neither risk shows up in any dashboard the company currently runs, which is precisely why it belongs on the board's radar.
4. Where do we stand relative to competitors?
Agent selection is comparative. The agent is not asking whether you are good. It is asking whether you are better than the four alternatives it can also parse, on the criteria in its mandate. A business can improve steadily in absolute terms and still lose every automated evaluation because a competitor is more legible.
Management should be able to show a benchmark, not an opinion. If the honest answer is “we have no idea how we compare under machine evaluation,” that is a gap worth closing quickly and cheaply, and it is closable: this is now measurable in a way it simply was not two years ago.
5. What happens to our brand investment when the buyer cannot see the brand?
This is the strategic question underneath the operational ones. Decades of moat-building through brand preference assume a human is choosing. When an agent shortlists on verifiable criteria, brand still matters, because humans write the mandates and set the allowlists, but it matters differently and possibly less at the point of transaction.
The board should hear management's view on which parts of the company's advantage survive machine-mediated buying, which parts erode, and what replaces them. Technical excellence, data quality and verifiable reputation are the assets that compound in this environment. It would be useful to know whether the capital allocation reflects that.
The pattern behind the questions
None of these questions require a board to bet on any particular vendor, timeline or technology. They require something more basic: that the organisation has looked at itself through the eyes of the buyer that is coming, measured what it saw, and put a name against fixing it.
Most have not looked. That is the whole opportunity, and the whole risk.