Mid-agency transition is one of the more reliable ways to find out what actually owns a company's public presence.
The answer is usually: nobody.
The agency had access to everything. When the agency exits, the access inventory — CMS, DNS, analytics, search console, ad accounts, social profiles, CDN — either transfers cleanly to an internal owner or fragments across whoever happened to be on the contract. In most cases, it fragments.
What we do first
We treat the website the same way we treat any other system of record when we arrive: inventory the access layer before touching anything else.
Six systems, one document: site host and deploy pipeline, CMS and content schema, forms and CRM connection, performance analytics, social and paid management, and accessibility layer. For each one: who has access, under which credential, and whether that credential is a personal account or a service account.
Three of the six typically require recovery steps. A previous employee who needs to transfer ownership of a social business account. A Google property that requires domain reverification. An accessibility platform that requires a support ticket.
None of this is technically complex. All of it is time-consuming in proportion to how long it has been unreviewed.
Lock brand direction before design work
Parallel to the access inventory, we lock brand and homepage direction with the internal marketing lead before any further design work is done on the site.
This sounds obvious. It is frequently skipped. The result of skipping it is design work that gets redone when the internal owner has opinions that differ from what the agency was building toward.
The brand review covers four things: logo files and usage rules, color and typography, photography and imagery standards, and tone of voice. All four need to be in one internal document, owned by the marketing lead, before the site is updated.
The homepage direction requires two working sessions. What the company does, who it is for, and what a visitor should do next. The existing site answers these questions in agency language. The internal owner rewrites the answers in operator language. The site is updated from that document.
The same gate
The pattern used for operations — read, draft, named person approves before anything posts — applies to the website with no modification.
Content drafts for new pages or updated copy land in the marketing lead's channel as drafts, with sources attached. SEO checks run against the draft before it is staged. Changes are staged in preview before going to production. The marketing lead reviews the preview and approves the deploy.
Nothing publishes without that step. The model does not have deploy access. The CMS content is drafted, not published. The publish step is the marketing lead's.
This is not a technical constraint — it is an operational one. The same logic that keeps an AI agent from posting to job cost or AP applies to a website. The site is a system of record for how the company presents itself publicly. A draft that goes live without review is an unsupported write. The same gate applies.
What it means
The site moves under internal control on the existing stack. No platform migration. No new CMS. No redesign project.
The marketing lead has a working process — an access document, a brand reference, a content workflow, and a preview step — instead of a pile of open access requests and a site she cannot change without calling the agency.
The same approval pattern that the operations team runs on invoice drafts and timesheet exceptions now covers website changes. One pattern. Multiple systems. The gate does not change based on what the system does.
Access recovery is not glamorous work. Neither is locking brand direction before the first design call. Neither is writing a preview step into a content workflow. These are the unglamorous prerequisites that make a website something the company actually controls instead of something the last vendor managed.
The site does not finish itself. Neither does the transition.