This is not a client's result, submitted on the agency's behalf. It's Dialect's own operation, the one that runs a live multi-client MCC through Claude every day, put forward as the proof itself.
A client result can be true and still be a story someone else tells. Anthropic asked for a case study; the obvious move is to hand over a client's ROAS chart and let the number do the arguing.
Dialect is putting forward its own operation instead. Not because there's nothing to show on the client side, there's a full roster of it further down, but because the thing worth showing here isn't one campaign's outcome. It's how the work actually gets built, inside a live multi-client account structure, with a person still making every real decision.
A client can describe what happened. The agency can show the process that produced it, in the accounts, on the day, with the tools that run the business. That's a harder thing to fake, and a more useful thing for Anthropic to evaluate.
Replacing a role with AI is where most agencies get stuck. The account owner still needs to own the account. What changes is how much of the surrounding work they carry alone.
Dialect runs on a small team of contractors, each covering one discipline: paid media, finance, commerce operations. Claude sits inside their actual tools through MCP connectors, not in a separate chat window they have to remember to open.
Not a chatbot bolted onto the workflow. A connector into the actual systems: the ad platforms, the ledger, the storefront.
Every example below is a live account, not a demo. The split shows what stays with the person and what moved to Claude.
Sets PMax strategy, decides budget shifts, owns the client conversation when performance moves.
Runs the campaign audits, manages negative keyword lists, and checks conversion data against the account's actual conversion actions.
Decides what to do about a pacing issue once it's found, and owns the creative and targeting calls.
Runs the pacing review itself, catching spend and delivery issues before they turn into a budget problem.
Resolves the actual discrepancies and owns the final ledger.
Matches transactions across entities and currencies, and surfaces the mismatches worth a human look.
Handles the customer relationship and anything that needs a judgment call.
Builds draft orders and handles routine fulfillment questions directly in the store.
Dialect runs embedded marketing leadership across e-commerce, membership organizations, healthcare services, and its own consumer brand.
The same Fractional CMO practice, applied to a direct-to-consumer subscription brand over eighteen months.
The partner network is built for agencies that extend Claude's reach into other businesses, not individuals who happen to be good at prompting. Dialect already does this: other agencies come to Dialect for AI enablement, specifically on media buying, and get taught the same thing shown on this page, where Claude belongs in the workflow and where a person has to stay.
That distinction is the actual product. Agencies that try to hand a whole role to AI end up redoing work a person should have owned in the first place. Dialect's pitch to those agencies isn't "replace your media buyer." It's "make your media buyer's hour worth ten." Partner network status compounds that reach past Dialect's own accounts.