What an Agent \"Cares About\" When Buying

The next wave of buyers won’t have feelings about your brand. They’ll have decision criteria. AI agents are already making purchasing decisions in software, and most products aren’t built to be evaluated by one. What agents weigh when they choose, and what they ignore completely, determines which products survive the shift.

Agents don’t browse the way humans do. They evaluate. The criteria they apply are fundamentally different from any buyer you’ve worked with before. Humans can forgive friction if they love the brand. Agents can’t.

Constraint satisfaction first, everything else later. Agents filter before they reason. A product must clear the hard rules, budget, timing, technical compatibility, or it’s eliminated from consideration. No amount of cleverness changes that. A human might stretch their budget for a brand they love. An agent stops looking the moment a price exceeds the ceiling.

Predictability rewrites the buying decision. Clear pricing and clear outcomes matter far more to an agent than to a human evaluator. Vagueness gets penalized as uncertainty, and uncertainty is risk. Even if your product is objectively superior but harder to reason about, an agent will often choose the simpler alternative. YC Lightcone and Stripe’s research both point to the same pattern: agents treat complexity as a hidden cost. When a human can call a sales rep and get reassurance, an agent has no such option.

Reversibility determines which products agents touch at all. Agents are conservative by design because they’re delegating on behalf of someone else. A decision that can’t be cheaply undone looks like a liability. This is why Level 3 automation (recommendation) and Level 4 (autonomous execution) sit at a hard threshold: once you cross into autonomous action, reversibility becomes the deciding factor. If your product commits the agent to an expensive, irreversible path, it gets screened out early.

Trust surfaces replace gut feel. This is where many founders misunderstand agent buying. Humans trust brands through reputation and social proof. Agents verify quality through APIs, audit logs, documentation, and SLA guarantees. If your product can’t be audited, if there’s no machine-readable proof of what it does, you’re invisible to agent networks. APIs and transparent infrastructure become your brand.

Cognitive cost is the hidden filter. Agents default to simplicity when reasoning about options, even when complexity would technically yield a better outcome. Overly complicated onboarding, login walls, CAPTCHA challenges, things that slow a human briefly, cause agents to abandon you. Clarity determines whether an agent bothers evaluating you at all.

Notice what’s missing from this list: brand vibes, clever copy, emotional storytelling. Those still matter, to the human setting the policy and choosing which agent to delegate to. Once the decision is delegated, the criteria shift. The human buys on trust and narrative. The agent buys on clarity and risk.

Pick the criterion your product fails most. Build for that gap. Your current customers might not care. Your future ones, the ones using agents to do this shopping for them, will.