AI for VPs of Marketing: Where to Start

The VP of marketing is the seat where an AI strategy either becomes an operating model or dies as a memo. The CMO sets the budget and answers to the board. The managers run their teams’ weeks. The VP owns the layer in between: several teams, several managers, one pipeline number, and the question of whether all that AI activity underneath adds up to a function that actually works differently.

Most AI advice skips this seat. It’s written for the executive deciding how much to spend, or for the practitioner deciding which prompt to use, and I’ve watched the gap between those two audiences swallow whole adoption efforts. This guide is for the person accountable for closing it. If you run marketing at a company where the CMO title doesn’t exist, or where you are the most senior marketer under a CEO who wears the CMO hat, this is also your guide, with the CMO playbook as the companion read for the budget and board layer you’ve absorbed.

What this seat owns that the seats around it don’t

The clean division of labor, in my experience: the CMO owns the why and how much, the managers own the daily how, and the VP owns the how it fits together. That means the VP’s AI questions are structural. Which workflows get standardized across teams and which stay team-local. Whose quality bar applies when content, lifecycle and paid are all generating drafts with the same models. Which of the six overlapping tool subscriptions survives the next renewal cycle.

None of those questions can be answered one level up or one level down. A CMO ruling on tool consolidation without knowing the workflows underneath picks wrong. A manager can standardize their own team’s recipes but can’t make the lifecycle team’s quality bar match the content team’s. Cross-team consistency is the work only this seat can do, and it’s exactly the work that determines whether AI adoption compounds or fragments.

The adoption numbers say fragmentation is the default outcome. Per Salesforce’s State of Marketing research, 75% of marketers have adopted AI, yet 84% admit to running generic campaigns, and 48% haven’t figured out how to adapt their strategies to widespread AI use. Read those together and you get the honest picture of the average marketing org in 2026: everyone is using the tools, the output got more plentiful and more average, and the strategy layer hasn’t caught up. Closing that gap is a VP-shaped job.

Start with an audit, not a tool

My strong recommendation for the first month: buy nothing, mandate nothing, and find out what’s actually happening. At 75% adoption, your teams are already using AI, and a meaningful share of that use is invisible to you because nobody asked. The audit runs through your managers, not around them. Ask each one to bring three things to your next one-on-one:

1. What their team uses AI for today, gathered amnesty-style, no penalties. The point is a map, not a compliance exercise.
2. The one workflow where it clearly works, with whatever numbers exist: time saved, volume shipped, rework rate.
3. The one place it’s causing quiet damage: review time ballooning, quality drifting, a tool nobody remembers approving touching customer data.

Two or three weeks of this gives you something no vendor deck can: a real inventory of where AI already lives in your function, which teams are ahead, and where the risk sits. In every version of this exercise I’ve run or seen, the same two surprises show up. The heaviest users are rarely the people who talk about AI the most. And at least one team has quietly built something excellent that the team next door is paying a SaaS subscription to do badly.

The audit also hands you your baseline. Every AI decision you make afterwards gets measured against it, which is what separates an operating model from a vibe. If your function doesn’t yet have agreed AI marketing KPIs, setting them is part of this same first month, because you cannot manage cross-team adoption on self-reported enthusiasm.

The three decisions only this seat can make

Once the map exists, the VP work starts. I’d frame it as three decisions, made in this order.

Decision one: what standardizes, what stays local. Some workflows deserve a single cross-team recipe: reporting narratives, brief formats, competitive research synthesis, anything where inconsistency between teams creates downstream cost. Others should stay team-local, because prompt quality is downstream of domain knowledge and the paid team’s ad-variant workflow has nothing to learn from the content team’s outline workflow. My rule of thumb: standardize wherever two teams’ outputs meet the same audience or the same reviewer, localize everywhere else. Write the standardized set down as shared recipes with named owners. The marketing manager’s guide describes what a recipe looks like at team level. The VP’s version of that job is deciding which recipes are function-wide law.

Decision two: who owns quality. This is the decision most VPs skip, and it’s the expensive one. When every team can produce five times the drafts, review becomes the constraint, and review capacity has to be assigned like budget. Name a quality owner for everything customer-visible, usually your strongest editor or most senior manager per surface, and make review time a tracked cost rather than an invisible one. A workflow that saves four hours of drafting and adds five hours of review is a loss that self-reported time savings will record as a win. You only catch it if someone in your seat insists the whole loop gets measured, which is the same discipline a proper AI ROI calculation enforces at the spend level.

Decision three: the tool consolidation call. The audit will surface overlap. Three teams with three subscriptions that all wrap the same underlying models is the standard finding, and the renewal calendar is your leverage. My bias, argued at more length in the CMO playbook’s build-buy-prompt section: consolidate toward fewer general tools plus written recipes, and make any specialized tool defend itself against the question “what does this do that a frontier model with our recipe doesn’t?” The VP is the right altitude for this call because you can see usage across teams, and vendors know it, which is why they prefer selling to your managers one team at a time.

Generic output is your problem now

Here is the uncomfortable stat pairing again: 75% adoption, 84% running generic campaigns. Adoption made average output cheaper for everyone, including every one of your competitors, so average now buys less than it did. I treat this as the central quality argument a VP has to internalize: the tools raise the floor and lower the ceiling at the same time, unless someone in the org is explicitly paid to defend the ceiling.

Defending it looks unglamorous. A written voice standard that every team’s AI output gets reviewed against, not per-team interpretations of the brand. A “kill the generic” pass in review, where the question is whether a competitor could have shipped the same asset by changing the logo. And a bias, which I hold strongly, toward using the saved production hours on the work AI can’t do: sharper positioning, better offers, actual customer conversations. HubSpot’s data shows 80% of marketers now use AI for content creation, which means content production is no longer where marketing advantage comes from. The advantage moved upstream, to deciding what deserves to exist. That decision lives in your seat and your managers’ seats, and no model makes it for you.

The cadence: where AI fits the rhythms you already run

A VP already has an operating cadence, and my advice is to bolt AI accountability onto it rather than inventing a parallel process that will die by December. Concretely, the version I’d run:

• Weekly, in existing one-on-ones: one standing question per manager. What did the team ship AI-assisted this week, and did anything fail review? Thirty seconds when things are fine. An early-warning system when they’re not.
• Monthly, in the leads meeting: one recipe reviewed, added or retired across the function. Models change fast enough that a static playbook rots in a quarter, and 61% of marketers call this the biggest disruption the field has seen in 20 years. Disruption-era playbooks need maintenance windows.
• Quarterly, in planning: the portfolio view. Which workflows scaled, which pilots died, what the review-pass rate and cycle-time numbers did against the audit baseline, and what the next quarter’s single new pilot is. One pilot at a time per team is my firm ceiling. Sequential adoption trains the org’s evaluation muscle. Simultaneous adoption trains its excuse muscle.

The quarterly view is also where governance stops being a document and becomes management. If your function doesn’t yet have a written policy covering what AI may produce unreviewed, what data never enters a public model, and where disclosure applies, the AI governance framework is the template I’d start from, and the VP is usually the right owner for it. The CMO sponsors it. You enforce it.

Managing up: feed the board answer

Sooner or later your CMO gets asked “what are we doing about AI?” in a board meeting, and the quality of their answer depends almost entirely on what you hand them. The strong version of that answer is workflow-shaped and evidence-backed, and the CMO playbook describes its four parts: where AI is deployed, what it’s returning, what it’s not allowed to do, and what’s next. Every one of those parts is manufactured in your seat. The deployment map is your audit. The returns are your baseline deltas. The guardrails are your governance policy. The pipeline is your quarterly pilot list.

I’d go one step further and volunteer the package before it’s requested: a one-page quarterly summary in exactly those four parts. It’s the cheapest credibility a VP can buy, it makes your CMO better in the room, and it quietly establishes that AI in this marketing org has an operating owner. When headcount or budget conversations arrive later, being the person with the evidence is worth more than being the person with the opinions.

Hiring and skills, through your managers

The VP rarely interviews the individual contributor anymore, so your influence on skills runs through what you tell your managers to hire and train for. The list I’d give them is short. Problem framing, because the person who writes a good brief writes a good prompt. Output evaluation, because calibrated judgment of “is this actually good” is the scarce function when production is cheap. And domain depth, because context is what separates a useful AI workflow from a generic one.

What I’d tell them to stop hiring for: tool-specific skill as a primary criterion. The tools turn over in months and the skills transfer in weeks. A candidate who can precisely critique a mediocre AI-produced asset in your domain is worth more than one who lists every model by name. At the function level, resist the urge to create an “AI marketer” role before you have multiple scaled workflows that need a governor. Grow the capability inside the teams first. The other role guides on this site are, in effect, the training syllabus for that: each one maps where AI lands in a specific job, and handing the right guide to the right lead is a cheap way to level-set a team.

What stays in the seat

Every role guide I write ends with this section, because the automation-resistant core is what defines the job. For the VP of marketing, my list: the cross-team resource calls, because trade-offs between teams are political and contextual in ways no model sees. The performance conversations with your managers, because an AI-drafted difficult conversation reads exactly like what it is. The quality ceiling, because someone has to be paid to say “this is generic” while every dashboard says throughput is up. And the accountability, because “the model produced it” is not a sentence a VP gets to say to a CMO, a customer or a regulator.

None of that shrinks the opportunity. It locates it. The VPs I see winning with AI treat it as an operating-model project with tools attached, run it through the cadence they already own, and spend the recovered capacity on judgment. The ones I see struggling bought the tools first and hoped the operating model would emerge. It doesn’t.

FAQ

How is this different from what the CMO should be doing? Division of altitude. The CMO decides how much to spend, carries the board narrative and sponsors governance. The VP builds and runs the machine: the audit, the standardization calls, the quality ownership, the cadence. In orgs where one person holds both titles, do the CMO layer first, budget and governance, then this guide’s sequence. My full view of the executive layer is in AI for CMOs.

My teams are at wildly different AI maturity. Do I level them? Not by decree. Use the audit to find your most mature team, turn their working recipes into the first function-wide standards, and pair their recipe owner with the lagging team’s lead for one quarter. Uneven maturity is an asset if the VP moves the knowledge sideways, and a liability only when each team is left to rediscover everything alone.

What’s the first metric I should put on my own dashboard? Review-pass rate: the share of AI-assisted work accepted without major rework, tracked per team. It catches quality drift months before customers or your KPI stack do, it exposes where recipes are broken, and it keeps the function honest about whether time saved producing is being spent re-reviewing.

The seat above: AI for CMOs. The seats you manage: AI for marketing managers. The foundations everyone shares: the complete guide to AI marketing.