- one session
- the MCP fleet
- the platforms
the seven
- GAGoogle Ads
- MEMeta
- G4GA4
- SCSearch Console
- TMTag Manager
- MCMerchant Center
- DMData Manager
The story of seven dashboards
Chapter 1. an hour of your life that is not billable
Seven dashboards, one notebook and a spreadsheet opened.
Each platform's dashboard has advice and no way to tell one from the other:
- a budget to raise, a campaign to pause
- suggestions to accept or dismiss across every tab opened
Feeling confused whether to defer a decision or at least try a change, or somewhat better — whether to export all data by hand, then try to get an answer in one Claude or ChatGPT session with everything attached.
The reasoning stays in your head on a Friday. It doesn't survive the weekend, then it repeats on Monday.
Chapter 2. the other version of that same week
The dashboards are still there. You do not have to open them.
Someone else combined data, guides, suggestions and asked each platform your questions.
The comparison isn't in your head any more. It's a record, and will be still there on Monday.
The advice may come but does not belong to the platform's own vendor, it becomes your agent's skill.
Accept or dismiss is now a decision you can defend, rather than a guess you are forced to take.
And this hour of your life goes back to the part that actually matters:
- deciding what action to take on the problem instead of working out whether one exists.
Chapter 3. where I enter the picture
I'm not a marketer. I am a software engineer.
I build marketing automation.
I gradually automated my own workflow one layer at a time.
I'm currently automating digital marketing operations for my clients:
- same problem, similar context, different tools.
You know your market — which months lie, why the obvious number is usually wrong. I know how to make systems answer questions without a person in the middle.
Chapter 4. what changed in my workflow
what the agent reads now, instead of me
- Jira to read the ticket breakdown and find the BRDs
- Confluence to read the design decisions and cross-check against the code
- Bitbucket / Github / Jenkins for obvious reasons
then
Five dashboards with half a dozen tabs each, a notepad with about as many, a word document, a spreadsheet, and logs.
now
One agent session. One terminal window.
The specialist knowledge moved out of my head and into the agent, and through this process of correcting and instructing the agent, those corrections became skills it carries, written out of my working day.
Every layer I added to the skills was one that came off my day, until I wasn't doing the assembly at all. And each layer coming off moved me further out:
- operating the tools, then describing what I wanted, then deciding what was worth wanting.
I don't think I work less, but it feels like it. I work on things I couldn't get to before, because the repetitive part had already spent me. I moved from specialist to generalist on my way to mastery.
Chapter 5. why the fleet exists at all
Each server and the skill that goes with it do one thing — one platform, one adapter, one set of operations it knows.
Even one is worth having. It takes a whole interface off your day: the tabs, the panels, the choices and the suggestions the platform puts in front of you. That is cognitive load that stops compounding.
Together they do more than that. The answers land in one session, then cross-checked against each other off the agent's reasoning and skill's speciality. Thus the comparison you were trying to do in your head is now presented to you ready for a decision.
See the fleet, server by server ›Chapter 6. the part that isn't just buying tools
Google and Meta APIs are rate limited based on token access, and others may need allowlisting.
That part is already done here.
Chapter 7. when we talk
I'll ask more than I answer. I don't know your business yet, and guessing is how people end up paying for something that doesn't fit their problem.
Chapter 8. if this sounds like your week
Tell me about it and we'll see if it's a fit.
let's see if we fit ›