For most of the last twenty years, a marketing team was drawn the same way. A head of marketing at the top, a few generalists underneath, agencies and freelancers around the edges, and work flowing down as tasks.
That drawing is coming apart. Forrester's 2026 analysis of the Fortune 500 found the Chief Marketing Officer title held by 36% of companies, down from 49% the year before, and marketing's seat on the executive team down from 58% to 52%. The easy reading is that marketing is being cut. Our read is that it is splitting. The work is being carved into narrower disciplines, growth, brand and product marketing among them, each with someone who owns the calls in that area.
The question founders ask us is whether the next marketing hire should be a person or an AI tool. That is the wrong question now. The right one is where the lines go: between disciplines, between people, and between people and AI.
When work moved through people, the org chart was what held a team together. It said who did what and who reported to whom.
Increasingly, what holds a team together is the workspace it works through. For us that is Claude. The skills that do the repeatable work, the context those skills read from, the routines that run on a schedule, and the master rules every workflow follows all live in one shared place. Call it an AI marketing operating system, an AMOS. Everyone on the team works through it, improves it, and inherits what others have built into it.
That changes what good team design looks like. You are no longer only arranging people. You are designing the system they work inside, and deciding how they collaborate through it.
In this kind of team, the manager's job moves up two levels.
The people closest to the work sit hands-on-keyboard inside the AMOS. Each owns one narrow discipline and runs the skills for it: the paid creative operator, the lifecycle operator, the content operator. They produce, test, fix and improve the skills as they go.
The manager sits above that and does four things no skill can do. They are the architect, designing which skills exist, what context they read and how work flows between them. They are the review gate, deciding what ships. They are the strategy and vision lead, setting what the system is for. And they are the prioritizer, deciding what the team works on this week and what it ignores.
Done well, this is where the extra output comes from. In our experience, a well-architected AMOS gets something like five to ten times more output from each hire, because every operator is backed by the system instead of starting from a blank page. It also changes how you hire. Instead of one more generalist, you stack narrow specialists, each one more focused than the last, because the system carries the breadth.
The failure mode is not that AI does too little. It is that it does too much, and nobody notices until the work stops sounding like you.
Last week we spent a day in Parker Simon's workshop on his Value Activation Model, a set of AI tools that assess designs against human values. It was a small-scale version of the whole argument. Each analysis traced a design back to the values it served and the principles that could close the gap, then stopped short of the answer. Simon built it that way on purpose: "Design judgment is not being automated; it is being made visible." The gap is where the most human, most authentic ideas of the day came from. Once a person made the call, AI picked up the execution and saved weeks.
That is the line we want inside every skill we deploy. Each skill should know exactly what it does, where it stops, and who takes over. Stripe's Katie Dill calls the alternative "zombie UI": work that functions and has no point of view. It is what you get when nobody drew the line.
1. MAP the disciplines. List the distinct areas of judgment in your marketing: growth, brand, product marketing, lifecycle, creative. Each becomes a seat with one owner, not a shared responsibility.
2. DRAW the line in every skill. For each skill in your workspace, state what it does, where it stops, and which person takes it from there. Build skills that hand back a diagnosis and options when the decision needs judgment, the way Simon's tools do.
3. SEAT the operators. Put one person hands-on-keyboard per discipline, running and improving that discipline's skills. Hire narrow and stack specialists rather than adding generalists.
4. ARCHITECT from two levels up. Give the manager the four jobs that stay human: architect the AMOS, run the review gate, set strategy, and set priorities. Protect their time from execution so the system keeps improving.
Plan on about 45 minutes. Have your team list and your current AI tools and automations in front of you.
You are a marketing org designer helping a founder
restructure their team around a shared AI workspace.
Company: [WHAT YOU SELL, STAGE]
Current marketing team: [NAMES AND ROLES, PLUS
FREELANCERS AND AGENCIES]
AI tools, skills and automations in use: [LIST]
Main marketing outputs each month: [ADS, EMAIL, SITE,
SOCIAL, CONTENT, ETC.]
Work through these steps in order:
1. MAP: Group the work into disciplines (for example
growth, brand, product marketing, lifecycle,
creative). Name one owner per discipline, or flag
it as unowned.
2. DRAW: For each skill or automation, state what it
does, where it must stop, and which person takes
over. Flag any that make judgment calls with no
human after them.
3. SEAT: Recommend who operates each discipline's
skills day to day, and where a narrow specialist
hire would add the most.
4. ARCHITECT: Describe the manager's role across four
jobs: system architect, review gate, strategy
lead and prioritizer. List what they should stop
doing hands-on.
Output an org map table (discipline, owner, operator,
skills used, where each skill stops) and a list of
the 5 changes to make first, each with an owner and
a deadline.
Ask me one clarifying question at a time when you
need more info.
When we start a fractional engagement, we run a version of this already loaded with the client's team, tools and brand rules, so the first session ends with a draft org map instead of a blank page.
MKT1 is the clearest picture of where this is heading. Three marketers, Emily Kramer, Halley Johnson and Katie Mitchell, run what used to take a much larger team: two newsletters, more than 20 social posts, more than 10 graphics, five or more videos and about 200 vetted job listings every month.
Their AMOS is built in Claude Code: two shared GitHub repos installed as plugins, more than 100 skills, 30 scheduled routines, a master CLAUDE.md file that sets voice, tone, design and data rules for everything, and a skill that points each workflow at context kept in more than a dozen connected tools like Airtable, Asana and Attio. Each of the three owns a clear area, and Kramer is the system's main architect. Her own rule of thumb: "If writing is the main way your team uses AI, I think you're doing it wrong."
They are honest about the limits. More than fifteen connected tools is probably too many, plugins need manual refreshing, and a bigger team would need clearer ownership boundaries.
Scale that shape up and a larger marketing org looks like a few discipline leads acting as architects, plus a small team of operators running the skills and agents underneath them.
Every marketing team is being redrawn right now, whether anyone draws it on purpose or not. The teams that come out ahead will be the ones that decided where the lines go: which disciplines exist, who sits in each seat, where every skill stops, and who architects the system that holds it together.
That architecture is leadership work, and it is the part of marketing that did not get cheaper. If you want help designing it for your team, book a call.
Product and operations are handled. If marketing is the missing piece, let's talk.