Agency client remediation: from vague brief to renew-able coverage reporting

The inbound request was familiar: a mid-market B2B client asked their digital agency to “make the site’s images accessible” ahead of a procurement questionnaire. There was no shared sample of failing pages, no definition of in-scope content, and no agreement on what “done” meant.

The agency’s previous pattern — juniors clicking through Media Library one attachment at a time — burned hours without producing a renew-able artefact. Account managers could not show coverage. Production leads could not price the next similar client. Leadership wanted a SOP they could productise without overclaiming WCAG certification.

Week one was paperwork with teeth. The SOW listed informative images on posts, pages, and key templates; excluded PDF icon systems and theme chrome; and defined empty alt as correct for decorative images. That single page prevented scope creep when stakeholders later asked for “every attachment since 2014.”

The agency also agreed a coverage target — not a legal conformance claim. “≥85% of in-scope attachments with reviewed, useful alt by closeout” was specific enough to report and honest enough to defend.

Before the contract was signed, the delivery lead ran free OpptiAI public-page audits on eight URLs the client already cared about: homepage, three service pages, two case-study templates, and two high-traffic blog posts. Screenshots went into the proposal deck — not as fear tactics, but as shared evidence.

The pattern was consistent: hero images with filename alt or blanks, in-content photography described with generic stock phrases, and decorative section dividers incorrectly filled with keyword strings. Estimated in-scope useful-alt coverage sat around 32% — low enough to justify the sprint, specific enough to measure against later.

Execution followed Tier A → B → C. Tier A covered homepage heroes, service landing templates, and the top twelve conversion pages. Tier B swept high-traffic blog archives and category heroes. Tier C was parked as an optional follow-on — explicitly, not as silent scope creep.

Operators worked in wp-admin with OpptiAI Alt Text: scan for empty and weak alt, generate drafts that could see the image plus surrounding page context, then edit brand voice and product specifics before save. Batch sizes stayed around 200–300 attachments so fatigue did not turn into blind approval.

The agency’s risk was not “AI wrote something” — it was “AI wrote something wrong on a money page and nobody noticed.” Every batch got a 10–15% senior sample against a one-page rubric: no filenames, no keyword stuffing, no invented product names, empty alt only when decorative.

Juniors used the free alt text grader on flagged items before escalating. When a batch showed a pattern — wrong executive names on team photos, for example — the team paused generation, fixed the prompt context, and re-ran that slice rather than patching one attachment at a time forever.

Closeout coverage on in-scope attachments landed around 85% — short of perfection by design, because Phase 2 long-tail backlog was a separate line item. The client renewed a light monthly retainer for new uploads rather than waiting for the next questionnaire panic.

The agency reused the SOW language, QA rubric, and report template on two similar engagements the same quarter. No revenue multiple was invented for this composite. The useful lesson is operational: scoped remediation with coverage reporting sells cleaner than open-ended “accessibility hours.”

The client did not receive a WCAG conformance certificate. Image alt remediation on agreed templates is not the same as a full accessibility audit across forms, video, keyboard traps, and third-party embeds.

Organic traffic did not double because alt text improved. The procurement questionnaire got a factual coverage answer. Screen-reader users on remediated templates had a better experience. Those outcomes were enough to fund the retainer — without inventing SEO fairy tales.