Adrian Martinez runs a digital marketing agency in Toronto focused on Website Design, SEO, and answer engine optimization (AEO). He and his wife deliver for about twelve clients today.
The constraint is time. Each account takes 10 to fifteen hours a month in hands-on work: research, drafts, technical SEO, and reporting. That math caps growth before it caps ambition. Adrian's bet isn't to hire a dozen account managers first. It is to build a delivery engine that maintains high quality while the team stays small.
One place to run the agency
Adrian had already been using Zapier for years on classic automations for client SEO work. When he started building in Claude, he connected Zapier MCP so the agent could access the apps his clients actually use: WordPress, reporting sources, image generation, and dozens of tools that do not offer a native Claude connection.
His day-to-day cockpit is Claude. One project holds the artifacts, skills, and knowledge files for the whole system. When he needs something executed in production, he describes the outcome and the parameters. Zapier MCP carries it into the connected apps.
I probably will never make another automation by hand. Why would I go a step backwards instead of just telling Claude, make me this, this, and this. These are the parameters. Go.
He still opens Zapier directly when he has to. Most of the time, he does not. The work is done on the agent side.
The SEO/AEO engine
Adrian is building what he calls an SEO/AEO engine: a repeatable path from client intake to published content to monthly performance reporting.
It starts with a seed form. Location, services, positioning, target areas. The fields you would normally collect in a kickoff call are captured once.
From there, the system produces a master research document: keyword direction, article targets, and content strategy. A Claude artifact serves as the control surface, almost like a lightweight project board for each client. Adrian chooses which pieces move into draft, which stop for his review, and which can run automatically.
Approved articles flow into WordPress with technical SEO handled in the pipeline. At month close, the engine pulls together site performance, traffic, Google Business Profile, and the rest of the campaign picture into a client-facing report.
He also routes image generation through Zapier. For one moving company client, generated visuals place the client's logo on trucks and boxes. Each account can carry its own visual brief inside the same hub.
The target he is building toward: roughly 30 minutes of hands-on time per client per month, down from 10 to 15 hours today, without adding headcount. He expects the finished system to produce sharper output than manual delivery because decisions are run on live data rather than static spreadsheets.
Adrian's AEO Engine portfolio dashboard: active clients, pending review, posts published, and LLM citations across his agency roster.
Claude thinks. Zapier MCP executes.
The split matters for anyone running a services business with AI.
Claude is where Adrian plans, iterates, and holds context across clients. Zapier MCP is how those plans become updates in WordPress, rows in a report, or assets in a client's stack. The language model is not the integration layer. Zapier is.
When a platform already exposes a solid native MCP, Adrian uses it. When it does not, he reaches for Zapier because the app catalog is already there and he trusts the connections.
Overall, it's been very, very great. Zapier is a solid bridge with a lot of other APIs and MCPs that aren't native to Claude.
The next layer: voice and scale
The SEO/AEO engine is not the only build on his list.
Many of Adrian's clients are trades businesses running Google Ads. Calls go unanswered while crews are on site. He is wiring ElevenLabs voice agents through the same Zapier MCP path so leads get answered when the owner cannot pick up the phone.
Same pattern: describe the workflow in Claude, let Zapier MCP handle the app layer, keep the team in one harness instead of five tabs.
For a two-person agency, that is the point. More clients, same team, delivery that does not depend on who had bandwidth that week.
Try the SEO/AEO agency engine pattern
Adrian's workflow generalizes into a repeatable MCP pattern for small agencies: one intake surface, one agent cockpit, and Zapier as the execution layer across client stacks that will never share a single native MCP. Get the skill now on the Zapier GitHub.
The pattern:
Input: Client intake (services, geography, positioning), live SEO/AEO signals, and the apps each account already runs.
AI/MCP workflow: Research and draft in Claude; route publishing, reporting, and asset generation through Zapier MCP.
Output: Published content, monthly client reports, and optional voice-agent follow-up — without adding account managers.
At a glance: Claude holds context and decisions. Zapier MCP connects WordPress, reporting, images, and voice tools. The agency team reviews exceptions; the engine handles the repeatable path.
