Course

AI Systems for Marketing Agencies: Agents, MCP and Evaluation

Build the agency's own agents, registry, evaluation and ledger, and own every line.

Seriously hardDigital marketing

0/1010 stops · about 3 hours · final quiz · certificate

Start course Sign in to save progress

Start with

• AI at Work: FoundationsStart from zero and finish with three things you can actually use.
• AI for Digital Marketing (Beginner)Your first year running social, reviews and a small ad budget alone.
• AI for Digital Marketing (Hard)Twelve accounts, two juniors, every grind automated and every decision still yours.

Who this is for

Agency owners, heads of performance and marketing engineers who want their agency to run on systems they own. Every example follows Nour, who runs a six-person paid media agency with twenty accounts and a developer on contract.

You will be able to

• Build a scheduled reporting agent that pulls three ad platforms and writes twenty client reports
• Build a read-only MCP server over the account registry and connect it to every Project
• Implement the approval workflow: proposal, review, signature, application, log
• Build an evaluation harness for the audit assistant with a fifty-case test set
• Run a creative-testing pipeline that briefs, generates, launches under approval and reads results
• Track cost, time saved and errors per system and decide what stays

Stops

1. 1 Tokens, scopes and the apply ruleThree platform credentials at the smallest scope that works, kept outside the repository, and the one rule every system in this course obeys. Set once
2. 2 The account registry and its MCP serverOne file that knows what every account is, wrapped in a read-only MCP server so every Project and every agent reads the same truth. Weekly
3. 3 The reporting agentPull three platforms into one row shape, write twenty reports with the Claude API from those rows only, and hold every one in a review queue until a person reads it. Weekly
4. 4 The audit assistant and its evaluationFifty account snapshots with the findings written down first, a runner that scores the assistant against them, and a gate that decides whether this version ships. Weekly
5. 5 Proposals to applied changesOne proposal format, a signature over the exact bytes, an apply step that refuses anything unsigned, and a rollback that was written before the change went out. Weekly
6. 6 Creative testing pipelineA brief built from the registry and the winner history, variants generated one variable at a time, launched through the signature path, and read only when the pre-registered numbers arrive. Weekly
7. 7 Guardrails as codeThe rules stop being a document and become one imported module with a table-driven test suite, a fail-closed default and a red-team run against itself. Weekly
8. 8 Client-facing surfacesThe delivered report, the portal page and the monthly summary, all built from an allowlist, with a test that proves what a client can never see. Weekly
9. 9 Cost, time and errorsOne ledger row per system per week covering what it cost, what it saved and what it got wrong, read every Friday, with a written rule for killing a system. Weekly
10. 10 Nour's capstoneThe scheduled runs and the routines that read them, then one system built end to end with tests, guardrails, an evaluation score and a cost line, graded out of a hundred. Weekly

Final quiz