Guide · 5 min read
How to Make Your Marketing Department AI-Ready
By Fredrika Frenkiel, Head of Studio & AI Creative Operations at Lunar, founder of Code of Alfred. About my work
Short answer
A marketing department becomes AI-ready by fixing its processes before it adopts any tool, most AI initiatives fail not because the technology is weak, but because they're layered on top of workflows, approval chains, and decision rights that were never designed to support automation. Being AI-ready means your team can say clearly, for any given task, what a system should own and what a person must own, before you buy or build anything.
1. What "AI-ready" means in practice: an AI operating model, not a tool list
Being AI-ready means having an AI operating model: a defined way work moves from brief to delivery with AI built into named stages, with decision rights and governance written down. It is the thing a marketing operations AI specialist builds before any tool is bought. A tool list is a shopping cart; an operating model is the system that decides what each tool is allowed to do, who reviews the output, and what happens when it is wrong. If you cannot describe that in one page, you are not AI-ready yet, no matter how many licences the team has.
2. Start with an honest audit of where work actually breaks down
Most marketing teams don't have an AI problem, they have a process problem that AI adoption exposes. Before evaluating a single tool, map where projects actually go wrong: Is planning happening in silos, so production teams get compressed deadlines for work that should have taken months? Is brand consistency dependent on one or two people remembering the rules? Is the same manual work, briefing, adaptation, localisation, QA, repeated every time with no shared structure? Write these down as specific, named failure points, not a general sense that “things are slow.”
3. Separate what AI should own from what a person must own
AI-readiness isn't a maturity score, it's a clear map of decision rights. As a working framework: AI should own generation (drafts, variations, first passes), structuring (briefs, tags, summaries), scaling (adapting one asset into many formats or markets), and pattern analysis (spotting what's working in performance data). People must keep ownership of strategic prioritisation, creative judgment, stakeholder relationships, and final sign-off. Write this split down explicitly for your team's actual workflows, vague enthusiasm about “using AI more” doesn't survive contact with a real production calendar.
4. Fix the workflow before you introduce the tool
If a process is broken manually, adding AI to it usually just produces the same mistakes faster. Redesign the sequence of steps first, who briefs what, in what format, reviewed by whom, before anything is automated. Only once the sequence is correct does it make sense to decide which steps AI should execute.
5. Build one small, real pilot before rolling out broadly
Pick one recurring, well-understood workflow, not the most complex one, and redesign it end-to-end with AI in a defined role. A brief-intake process or a copy-adaptation workflow are good starting points: contained, repeatable, and low-risk if something goes wrong. Measure the actual before/after, time saved, quality maintained, errors introduced, before expanding.
6. Treat governance as infrastructure, not a document
Brand guidelines and quality standards that live only in a PDF get ignored under deadline pressure. AI-readiness means encoding those standards into the tools people actually use day to day, so consistency doesn't rely on someone remembering to check a document.
7. Expect resistance, and address it directly
Teams often fear AI adoption because it feels like it erases judgment or job security. The teams that adopt fastest are the ones where the human role is made explicit and sharper, not smaller, where people can see exactly what they still own and why it matters.
How to do it, step by step
- 01
Audit where work actually breaks down
Map specific, named process failures before evaluating any tool.
- 02
Separate what AI owns from what a person owns
Define decision rights explicitly for your team's real workflows.
- 03
Fix the workflow before introducing the tool
Redesign the process sequence before automating any step of it.
- 04
Build one small real pilot
Redesign one recurring workflow end-to-end and measure the result.
- 05
Treat governance as infrastructure
Encode standards into daily tools rather than a static document.
- 06
Address resistance directly
Make the human role explicit and sharper, not smaller.
This is the exact sequence I use when redesigning marketing operations for AI, business problem first, workflow second, technical layer last. See how I approach this in practice.
Common questions
- How do I know if my marketing department is AI-ready?
- You are AI-ready when you can state, for each recurring workflow, what AI owns, what a person owns, who signs off, and how quality is checked. If that is undocumented, adopting tools will only speed up the existing confusion.
- What does a marketing operations AI specialist do first?
- Audit how work actually flows today and name the specific bottlenecks, before evaluating a single tool. The operating model is designed from real friction, not from a vendor roadmap.
- Do we need an AI operating model before buying tools?
- Yes. The operating model decides which steps are worth automating. Tools bought before that are usually retired within a year because nothing in the workflow depends on them.
Last updated: