Who should own AI transformation on your marketing team?

Marketers have marketing to do. Campaigns need to launch. Customers need to be understood. Someone has to decide where the next dollar goes.

Now they’re also being asked to figure out how AI changes all of that. Someone has to choose which work to hand to agents, build those agents, get the team using them, and keep them working as the business changes.

So who owns that? Whoever does needs three things:

Domain expertise to know which work is worth handing to agents and what good output looks like.

Technical expertise to build agents that hold up on real work.

Ownership to keep them working as the business changes, with someone accountable and the time to do it.

We see marketing teams try four approaches. Each one works somewhere, and each one has a ceiling. The dots under each are our own read on these three.

1. Your internal domain expert

A marketer who knows the business inside out and becomes your AI champion.

Best for: Small, nimble teams where one motivated person can help everyone become more productive. The champion already knows which reports eat the week and which numbers the team trusts, so the first few agents tend to be the right ones.

The ceiling: AI adoption is asymmetric, and your champion still has a day job. One or two people take to it and pull far ahead while the rest of the team keeps working the old way. The champion ends up building for everyone, maintaining what they built, and fielding the questions when a number looks wrong.

Setting up agents and skills once can make you feel more productive. Keeping them useful is a different job. Do you know how to improve a skill? How to fix a broken agent? Whether your agents are reliable enough to trust? What happens when they go stale?

Building, maintaining, and improving an entire agent workforce becomes a big responsibility as the team grows.

2. Your internal engineers

Put dedicated technical talent on it: your engineering team, or an engineer hired for the job, builds agents, automates workflows, and supports the team.

Best for: Organizations that can invest in a permanent internal AI capability. Engineers know your systems and your data, and what they build tends to stay running.

The ceiling: Too often, we see engineers ship one-size-fits-all tools that don’t meaningfully improve how domain experts work. Without deep business context, it’s easy to optimize for what can be built rather than what actually creates value.

Which projects do they start on? How do they get enough time to handle everything and also learn the domain expertise stakeholders need for agents to work at a human level? Who handles enablement for the tools they ship? Is there a PM involved? Engineers building for customers usually have one. If nobody plays that role for marketing, that’s a strange way to build.

3. AI consultancies

Hire a firm to assess where you stand, set your AI strategy, and recommend what to do next.

Best for: Leadership teams that need an outside view and a plan: aligning executives, prioritizing use cases, making the case for budget, setting governance. You get what the firm has seen at other companies, on a fixed scope, without adding headcount.

The ceiling: Consultants are expensive, and their fees aren’t directly tied to outcomes. A consultant’s deliverable is usually a roadmap, a deck, or a training session.

So the consultant audits your work, and then what? Do they hand it to your agency or your internal team? Does that team have the bandwidth to build what you need, or the AI expertise? Do you bring in the consultancy’s own implementation team to put the work into action, and get charged twice? Then do you keep paying consulting prices forever to maintain it and keep improving it?

Most AI consultancy engagements we’ve encountered aren’t focused on marketing. Why would a CMO hire someone who doesn’t understand marketing to transform their organization? A use case can look right on a slide and still be wrong for the brand, and it takes a marketer to see that.

4. Forward Deployed Marketing Engineers

Combine marketing domain expertise, AI engineering, and ongoing enablement in one team working alongside your marketers.

The frontier already tells us the optimal way to build an AI-native team. Stripe’s marketing organization is hiring what it calls Forward Deployed AI Accelerators, “embedded directly with marketing cohorts to make AI the default mode for all work.” The job listing asks them to “prioritize what to build, and be clear about what is and isn’t worth building.”

At Uber, CTO Praveen Neppalli wrote in July that the company handpicked about 30 of its most AI-proficient engineers and paired each with a domain expert from a business function, marketing included. Each pair got two weeks, and the first two days went to shadowing the expert. His biggest lesson: “The best AI opportunities are rarely visible from the outside. You discover them by sitting next to the people doing the work, understanding every friction point, and building with them, not for them.”

Not every business is Stripe. This team is a hybrid of domain experts and technical people, and hiring and training that hybrid internally is really hard. Uber could handpick 30 engineers who already knew its systems. Most marketing teams are deciding whether they can justify one hire.

At Bamboo, we scale this model to all businesses, bringing the best of the frontier to your marketing team. And we take the model a step further: both people on the team know marketing, and one of them is a dedicated PM.

The forward deployed PM is a marketing domain expert embedded with your team who speaks your language. They prioritize the work and identify the hidden steps in your workflows.

The AI engineer, who also has marketing experience, builds your agents with true engineering rigor, treating skills and context as natural language code. The PM writes a brief, and you approve it. The engineer breaks it into a design doc and tickets, so the gaps show up before anything is built.

Here’s what else we’ve found makes this work.

Truly embed in the business. We work side by side, on-site multiple times a week, to learn the workflows, constraints, and context that never make it into a requirements document. Sitting with the people who do the work is how you find the step that isn’t written down anywhere, the one nobody, human or AI, was doing.

Make upskilling easier. Build reliable systems that marketers can use, adapt, and improve themselves, rather than creating more dependence on engineers. Before we schedule a report, the marketer runs it against their own numbers. Once it matches the report they already produce by hand, it goes on a schedule. When a later output falls short, they explain what’s wrong in marketing terms and see how that correction changes the agent. On a recent build, the marketer was the one who asked whether a large request should be one skill or several.

Own the outcome. Agents break, workflows change, and business context evolves. Someone needs to be accountable for continuous improvement, team-wide productivity, and bottom-line impact. When a daily report understated refunds, we added the refund rule to the brain the agents run on, so the correction applies to every report that follows. That’s the difference between an agent that’s good on launch day and one that’s better a month later.

Report every week in your language. Each week we tell you what improved, what we checked, and what still needs your judgment.

Keep everything yours. Your brain, agents, skills, and evals live in your repo. If we stop working together, they stay with you.

Best for: Marketing teams that want this model now but can’t hire and train the hybrid team themselves. You get a domain expert and an AI engineer working with your marketers, without building a new function first.

The ceiling: The model only scales as far as the domain expertise does. Each new vertical or team needs people who know that kind of marketing, and they’re hard to find. We’re also an outside team, so we have to go through procurement, legal, and credentialing before we can start.

Side by side

The paths also combine. The engagements that go best for us have an internal champion on the customer’s side.

If you’re trying one of these paths, or you think I missed one, I’d like to hear how it’s going.