Agency economics reward exactly one kind of offering: recurring, high-margin, low-marginal-cost services that clients can't easily evaluate or do themselves. SEO retainers fit. Site maintenance fits. And agent-readiness — keeping client sites legible and operable for AI buyers — fits better than either, because the problem regenerates with every deploy and the client can't see it happening.
Why now, and why agencies
Your clients are already seeing AI referrers in their analytics and already asking what to do about it. They lack the tooling to measure agent behavior and the engineering context to fix what's broken. You have both relationships and repos. The agency that answers this question first for a client portfolio owns it — it's a land grab measured in quarters, not years.
Packaging that works
The clean structure is a three-part retainer. Baseline: an agent-readiness audit per site — score, category breakdown, competitor comparison — as the anchor deliverable. Remediation: the fixes, delivered as reviewed code changes, prioritized by severity and revenue impact. Monitoring: scheduled rescans and a CI gate so regressions surface as alerts to you, not surprises to the client. The monthly report writes itself from the scan data: score trend, issues fixed, regressions blocked.
Pricing logic
Anchor to the client's stakes, not your hours: a store doing meaningful volume with a growing AI-referral segment is protecting real revenue. A per-site monthly fee tiered by site complexity — bundled into existing maintenance retainers or sold standalone — keeps it simple. Your marginal cost per additional site is mostly tooling, which is the margin story.
The pitch to clients is one sentence: "A growing share of your buyers are machines — we make sure your site works for them, and prove it monthly."
Delivery without drowning
The trap is doing this artisanally — hand-auditing twenty client sites monthly doesn't scale. The leverage move is a multi-site dashboard with white-label reporting: automated scans, automated fix generation for the deterministic classes, your team reviewing and merging. You stay the trusted layer; the machinery does the repetition. That's what turns a service line into a product line.