Everyone is using AI now. And almost everyone is using it the same way.
You log into Google Ads, export a report, paste the CSV into ChatGPT or Claude, get an analysis, then repeat the whole process for Meta, Google Analytics 4, and whatever else is on your plate that week. Same painful process, every platform, every week.
Image from author, June 2026 That is not AI-powered marketing. It is AI-assisted copy-pasting.
The AI in that workflow is working on a static snapshot. Not live. Not connected to your actual account. Not aware of what happened yesterday or what your cost-per-acquisition (CPA) target is. It is a powerful engine running on stale fuel, and it explains why the output feels inconsistent: great one day, generic the next, always requiring more editing than it should.
The problem is not the model. The problem is the setup. There is a three-layer stack that changes this fundamentally: MCP for live data access, Skills for behavioral consistency, and Claude Projects to package everything into a reusable team environment. Each layer solves a distinct failure mode. Together, they are the difference between AI as a novelty and AI as infrastructure.
Layer 1: MCP Gives AI Eyes Into Your Actual Business
Model Context Protocol (MCP) is an open standard designed to connect AI models to external tools and data sources. Think of it as the Zapier layer for AI, except instead of moving data between apps, it gives the AI the ability to read, query, and in some cases act on that data directly.
Without MCP, your AI is working blind. It knows a lot in general, but it knows nothing specific about your business, your campaigns, your customers, or your performance. You copy-paste numbers into a chat window and ask it to analyze them. That is not intelligence at scale. That is a very expensive clipboard.
With MCP connected, the AI can pull live data directly from your tools. Google Ads has an official MCP server, which means you can ask Claude to check which campaigns are underperforming against your target CPA right now, pull search term reports , surface budget pacing issues , or compare performance across campaigns, and it queries the actual account rather than waiting for you to paste in a report. No export, no copy-paste, no manual formatting step.
Build A Live Data Stack
Image from author, June 2026 The same principle applies to GA4, your CRM, or any other data source with an MCP server available. But Google Ads is the clearest starting point for PPC teams because the data is live, the decisions are time-sensitive, and the performance gap between acting on Monday data versus Friday data is real and measurable.
For marketing teams specifically, this matters because performance data is always moving . The analysis you do on Monday is stale by Wednesday. An AI that can see live data is categorically different from one that cannot.
Layer 2: Skills Tell AI How To Behave In Your Context
MCP handles the data problem. Skills handle the consistency problem.
A Skill is a set of persistent instructions that tells Claude how to approach a specific type of task. Not what to do once, but how to behave every time. You define the rules once, and every conversation that uses that Skill inherits them automatically.
For agencies, this is the single biggest operational unlock available right now.
Think about how much implicit knowledge lives inside your agency that never gets documented. Your senior analyst knows your reporting format, your preferred attribution model, how to frame recommendations for conservative clients versus growth-stage ones, which metrics your most common client types actually care about. A junior hire takes six months to absorb that knowledge through osmosis.
A Skill captures it in a few hundred words. You write your agency’s best practices once: how to structure a campaign audit , how to frame budget recommendations, what tone to use in client-facing summaries, which key performance indicators (KPIs) to flag automatically. Every team member who uses Claude with that Skill active gets the senior analyst’s judgment baked in from day one.
Image from author, June 2026 A concrete example: your agency has a standard approach to Google Ads account audits. You check Quality Score distribution , search impression share by campaign type, conversion lag windows before touching return on ad spend (ROAS) targets, and you always frame recommendations against the client’s stated growth goal rather than platform benchmarks. That entire checklist, framed as a Skill, means Claude runs that audit consistently every time via the Google Ads MCP connection, pulling live account data and applying your framework automatically, not just when your most experienced person is doing it manually.
Layer 3: Claude Projects Package Everything For Teams
Projects are Claude’s way of creating persistent, context-rich environments. Each Project has its own instructions, its own knowledge base, and its own memory that carries across conversations. It is the operational container that makes the MCP plus Skills combination actually usable at a team level.
For agencies, the setup is straightforward: one Project per client.
Each client Project gets the client’s context loaded in: their business model, their target audience, their historical performance benchmarks , their seasonal patterns, any brand guidelines relevant to copy or messaging. You also connect the agency-level Skills, so they apply automatically. Now, every conversation about that client starts from a fully briefed position.
The result is that whoever on your team opens the client Project, whether it is the account lead, a strategist covering while someone is out, or a junior pulling a quick report, starts from the same informed baseline.
For in-house marketing teams, Projects work differently but just as powerfully.
Instead of one Project per client, an in-house team typically builds one Project per function or workflow. A paid search Project holds the brand’s campaign structure, naming conventions, bidding philosophy, and target metrics. When that Project is connected to Google Ads via MCP, a question like “which campaigns are pacing over budget this week and which are under-delivering against impression share targets” becomes a two-second query rather than a 20-minute reporting exercise. A content Project holds the brand voice guide, approved messaging frameworks, and the current content calendar. A reporting Project knows the stakeholder who receives the report, what they care about, and what format they expect.
Skills in an in-house setup carry the company’s own institutional knowledge rather than agency best practices. If your brand always measures success by new customer acquisition cost rather than blended ROAS , that lives in the Skill. If your growth team uses a specific attribution model for budget allocation, that lives in the Skill. If the chief marketing officer prefers a one-page summary over a data dump, the Skill handles that too.
