Building a Marketer's AI System: A Layered Framework

Marketing Against the Grain2026-08-18go watch the original →

the gist

Kieran Flanagan outlines a modular AI stack for marketers, emphasizing that effective automation requires an 'intelligence layer' of structured context files (goals, ICP, competitors, etc.) that specialized AI skills query to scale productivity and decision-making.

The Intelligence Layer: The Foundation of AI Systems

The core premise is that AI systems fail when they lack context. Instead of relying on generic prompts, marketers should build an 'intelligence layer'—a set of structured Markdown files (.md) that act as the single source of truth. This layer contains six critical files: Goals, Ideal Customer Profile (ICP), Competitors, Positioning, Team, and Execs. By anchoring all AI skills to these specific files, the system avoids hallucination and ensures outputs align with the user's actual business objectives.

Scaling Through Modular Skills

Once the intelligence layer is established, the system is built out in functional layers. Each skill is designed to query only the specific context files it needs, preventing token bloat and maintaining accuracy.

People & Leadership: The 'People Brief' automates one-on-ones by tracking KPIs, blockers, and wins, while the 'Voice of You' skill mimics the user's communication style to answer team questions autonomously.
Strategic Thinking: An 'Exec Panel' simulation allows users to pressure-test pitches against the likely critiques of their leadership team before presenting, while 'Challenge Me' skills use frameworks like inversion and pre-mortems to identify blind spots.
Product Storytelling: This layer focuses on differentiation. By maintaining a 'Voice of the Customer' (linked to ICP) and a 'Voice of the Competitor' (linked to competitor data), the system can grade positioning on a 1-10 scale to ensure the brand isn't drifting into generic messaging.

Operational Efficiency

The final layer focuses on daily execution. Skills like 'Priority List' and 'Blocker Watcher' ingest data from Slack, email, and project management tools to synthesize a daily to-do list based on urgency and strategic goals. The 'What Did I Miss' skill tracks outstanding commitments, surfacing the oldest pending tasks first with drafted responses ready for editing.

Continuous Improvement

The system includes a self-correcting mechanism: an AI Coach. This skill monitors the user's prompt history and file updates, suggesting improvements to prompt habits, identifying outdated context files, and recommending new skills to build based on repetitive manual tasks. The 'Record Skill' feature (available in desktop AI apps) allows users to screen-record their workflow, which the AI then translates into a reusable, automated skill.