From silos to pods: how to rebuild a marketing team for the AI era

TL;DR: AI has made marketing production cheap, so the teams organised around production (email, paid, content, social) are organised around the wrong thing. The replacement is small pods built around a buyer or a lifecycle stage: two or three senior generalists, a set of agents doing the execution, and one junior learning the craft. Below is the model, the shape trap most teams fall into when they cut juniors, the five capabilities to build before you hire anyone new, a 30/60/90 to start, and a list of courses mapped to each capability.

Every CMO I speak to this autumn is holding the same two documents: an AI budget request and a headcount plan. They were written by different people, and they contradict each other. The AI request assumes production gets faster. The headcount plan still buys production.

The honest version of the shift is this. Marketing is moving from production to orchestration. AI handles execution. People handle strategy, taste, brand and ethics. Once you accept that split, the org chart, the role descriptions and the hiring profile all have to be rewired around it.

“AI amplifies whatever you feed it. Fuzzy positioning, weak briefs, and shallow audience understanding all show up in the output, just faster and at greater scale.”
Harris Beber, CMO, monday.com

Why channel teams stop working

Marketing teams organised around channels because production was the bottleneck. An email campaign took a specialist a week, so you needed someone who owned email. The same logic gave you a paid person, a social person, a content person, each with their own brief, tone and view of the customer. The customer experienced that as fragmentation, and every handoff between those people cost days.

AI collapses production time. A nurture sequence that took days now takes hours. Variants, segmentation, send-time optimisation and performance analysis all run faster than a person can check them. That moves the constraint upstream. The question stops being “can we produce this?” and becomes “do we understand this buyer well enough to know what to produce?”

Three things follow:

1. The bottleneck is now thinking. Pods are organised around thinking about one buyer, so they sit where the constraint is.
2. Handoffs cost more than the work. At AI speed, waiting two days for the right specialist is the most expensive step in the process. Pods remove the handoff because the coordination is inside the team.
3. Customers see one journey. Every touchpoint is owned by the same people with the same understanding of the buyer, so the message holds together across channels.

Channels are a way of organising production. Pods are a way of organising thinking. When AI does the production, organising around thinking wins.

The model

Four parts, and the proportions matter more than the boxes.

A strategic core. The CMO and the directors. They set the outcomes, own the positioning and the brand rules as written context, and coach the pod leads. They do not approve every asset.

Pods, one per buyer or lifecycle stage. Default lean: two or three expert generalists, with agents filling the specialist depth. Each generalist owns a workflow end to end, for example mid-market activation, and never a channel. Give each pod one to three outcome KPIs (pipeline from the segment, activation rate, retention) and freedom on how to hit them. Be relaxed about how the work gets done and strict about the result.

A shared studio. Design, social and events serve every pod rather than sitting inside one.

A platform layer. Data, automation build, the prompt and context library, brand and claims rules, measurement, and the hard limits that live outside the AI. Two rules keep it honest. The platform enables and never gates: if a pod has to raise a ticket to get work done, the silo is back. And whoever writes the prompts and automations owns their performance, with no handover to a separate ops team.

One scale rule: one to three pods can run without a platform. Beyond that, duplicated prompts, stacks, tracking and brand rules cost more than building one.

Inside a pod

The pod lead owns the buyer’s journey and pipeline across every channel. They write the brief, set the KPIs and decide what AI accelerates and what needs human judgment.

The generalists produce across every channel for that buyer, with agents doing first drafts, variants, personalisation, research synthesis and reporting. Their craft shifts to briefing, reviewing and verifying.

The agents are the lean middle: research, drafting, brand and editorial review, evidence checks, campaign build, weekly reporting. Every agent has a named human owner in the pod.

The junior is on an apprenticeship track: reviewing AI output against the brief and the approved claims, running experiments under the pod lead, listening to customer calls, and rotating pods to build breadth early. More on why this role matters in a moment.

Budget for the verification tax. AI produces output far faster than people can check it. Plan review time for claims, brand, accuracy and compliance into pod capacity, or review becomes the new bottleneck and quality slips without anyone noticing until a customer does.

The shape trap

Most marketing teams today are pyramids: lots of juniors at the base, a few seniors directing. The common overreaction to AI turns that into a diamond. The juniors go, because their work is the work AI absorbs first (landing page builds, email execution, first drafts), and the middle bulks up with people whose job is to oversee the AI. You end up with plenty of managers and nobody learning the craft underneath them.

The right shape for a pod is an inverted pyramid: a few senior generalists with AI doing the execution. But an inverted pyramid has no learning path. The right shape for the organisation is an hourglass: senior execution at the top, a lean middle, and juniors learning the craft at the base.

“How’s that going to work when 10 years in the future you have no one that has learned anything? My view is you absolutely want to keep hiring kids out of college.”
Matt Garman, CEO, AWS, as quoted in Steven Brovich’s talk on building teams for the agentic era

So redesign junior roles rather than removing them. Reviewer: checking AI output against the brief, brand and approved claims. Experimenter: running tests inside a pod under a senior owner. Customer contact: interviews, sales-call listening, community, the context the agents don’t have. And rotations across pods every six months or so. Watch for the deskilling trap while you do it: juniors using AI ship more but understand less of what they ship, so make “explain why this works” part of every review. When a manager role opens, ask whether it is better filled as a junior plus agents before you backfill like for like.

Capabilities first, headcount second

The articles about the AI-era marketing team tend to list five new job titles: prompt engineer, AI ethics officer, AI integration manager, human-AI creative director, predictive journey architect. Hiring all five would rebuild the specialist silos under new names and push you straight into the diamond. Read them as capabilities instead.

Capability	What it covers	How to cover it
Prompt and context engineering	Instructions for AI aligned to your brand voice, personas and claims	A skill every marketer builds. The shared library sits in the platform layer.
AI ethics and governance	Responsible-AI rules, bias audits, regulatory compliance	Written into the rules and context every agent runs on. A named exec maintains them alongside their current role.
AI integration	Tool stack, integrations, automation pipelines	Upskill your MarTech or marketing ops lead. A dedicated hire only once you run more than two or three pods.
Human-AI creative direction	Deciding what AI produces and what humans craft	Your existing creative lead, reskilled.
Predictive journey design	AI-predicted journeys, personalised sequences, decision trees	A skill inside pods and your analytics function.

Underneath those sit three capabilities every marketer in the pod needs, and four behaviours worth rewarding. The capabilities, from Customer.io: AI orchestration (delegate execution while owning strategy), cross-functional fluency (acquisition, activation, retention and expansion as one system) and audience insight (read behavioural data, turn patterns into strategy). The behaviours, from Harris Beber: sharper briefs, stronger review standards, strategy before execution, and taste, meaning the willingness to say no to fast but mediocre work.

Set the bar for “AI fluent” high when you hire. Casual chatbot use doesn’t count. The test is a proven ability to build repeatable workflows that save ten or more hours a week, paired with domain knowledge of your customer, category and regulation. AI fills the production gap. It cannot fill the context gap.

Governance lives in the system

Governance works best as rules the systems enforce rather than a policy document people are asked to follow. Before any agent or automation acts, four questions: who is it and who authorised it (a named owner); what is it allowed to do (channels, budgets, audiences, claims in scope); is it performing as expected; and can we audit what it did. In practice that means brand voice, approved claims and banned phrases in shared context files every workflow reads; spend caps, send limits and audience exclusions enforced in the platform rather than by asking the model nicely; and a clear split where legal, compliance and ops write the rules and pods own the workflows. Nobody approves every asset by hand, or they become the bottleneck.

If you want the detail on the shared context files, I wrote it up in The Marketing Shared Brain.

How to start: 30, 60, 90

Order matters. Prove one workflow before you build the structure around it, and don’t hire the org chart on day one.

First 30 days. Choose one workflow and one buyer segment. Decide whether it runs on the AI built into your tools or a composed workflow with your own context. Map the current team against the five capabilities above. Audit which decisions in the team belong to AI and which to people. Adopt a brief template that asks, before anything else, what AI should accelerate and what needs human judgment.

Next 60 days. Stand up the pilot pod: two or three generalists plus agents, and a junior. Pick the segment where the work is already live and the outcome is measurable; for most B2B teams that is existing-customer growth. Set one to three outcome KPIs and track review time alongside output. Name an accountable human for every agent. Rewrite the hiring scorecard around domain expertise, AI orchestration, cross-functional fluency and audience insight.

Next 90 days. Review pilot outcomes and review load before adding pods or roles. Collapse the remaining channel silos into pods, one buyer at a time. Protect entry-level hiring and redesign the junior roles as apprenticeships. Move reporting from monthly to weekly. Surface the pod KPIs in executive review, because the measurable version of marketing is what gets the CMO back into the room where decisions are made.

Courses, mapped to the capabilities

You cannot send a team on a course and get a pod. But you can close specific capability gaps quickly, and most of these take a day or less. I have grouped them by the capability they build. Prices and dates are as published on 2 October 2026; check before you book.

Foundations, for everyone in the team

Course	Provider	Format and time	Cost
AI Fluency: Framework and foundations	Claude Academy (Anthropic)	14 lessons, about 4 hours, self-paced	Free
Generative AI for Everyone	DeepLearning.AI, Andrew Ng	About 5 hours over 3 weeks	Free to audit; certificate paid
Google AI Essentials	Google via Coursera	5 short courses, under 10 hours	Coursera subscription, 7-day trial

Prompt and context engineering, the skill every marketer builds

Course	Provider	Format and time	Cost
Google Prompting Essentials	Google via Coursera	4 courses, under 10 hours	Coursera subscription, 7-day trial
AI capabilities and limitations	Claude Academy (Anthropic)	13 lessons, about 3.5 hours	Free
Intro to AI Prompt Writing	Section	2-hour live workshop with recording	Per workshop or membership

Applied AI in marketing, for pod leads and generalists

Course	Provider	Format and time	Cost
AI in Marketing	Chartered Institute of Marketing	1 day, virtual or in person, introductory	From £499
AI Marketing Strategy and Tools	Chartered Institute of Marketing	1 day, virtual or in person, intermediate	From £525
AI for Marketers	Section, taught by Tahnee Perry	2-hour live workshop with recording	$195 list, or membership
Piloting AI	Marketing AI Institute	18 on-demand courses, self-paced	Inside the AI Mastery membership, $999 a year

AI search, for whoever owns the website and content

Course	Provider	Format and time	Cost
AEO Fundamentals Certification	HubSpot Academy	Self-paced certification	Free

Leading pods and the platform, for the strategic core

Course	Provider	Format and time	Cost
Building AI-Ready Teams and AI for Team Leaders	Section	1.5 to 2-hour live workshops	Per workshop or membership
Building Your Team’s Agentic Roadmap	Section	2-hour live workshop	Per workshop or membership
Building effective human-agent teams (beta)	Claude Academy (Anthropic)	5 lessons, about 45 minutes	Free
AI Strategy	Reforge	Live, 4 weekly sessions	Reforge membership, $1,995 a year

Two things no course covers. The first is your own context: the personas, the messaging house, the approved proof points and the language rules, written down where every person and every agent can read them. The second is the pilot itself. Pick the workflow, run the pod, measure review time as closely as output, and let the results decide what you build next.

If you want help building it

Working out the model is the easy half. The harder half is standing up the pilot pod inside a live team, writing the context the agents run on, wiring the guardrails into the platform, and proving the outcome KPIs before the next headcount review. That is the work I do with marketing teams, usually starting with one pod and one workflow, with first deliverables in the first fortnight.

If your AI budget and your headcount plan are currently pointing in different directions, book a strategy call and we will work out which pod to start with.

Sources: Harris Beber, CMO of monday.com, on AI and marketing briefs; Steven Brovich (AWS), talk on building teams for the agentic era, including the Matt Garman quote and the four team shapes; Customer.io on the three universal marketing capabilities; Isabelle Guis on CMO reporting lines; Academy of Continuing Education on the pod-based org chart and implementation principles; Singapore IMDA’s Model AI Governance Framework for Agentic AI (January 2026). Course details from each provider’s website on 2 October 2026.