You have a growth target. You haven't been allowed to add people all year. Somewhere on LinkedIn, somebody is explaining how they run an entire marketing department with agents.

It's understandable that the budget conversation turns into: how do we do that?

Meanwhile, your team might be making more things than ever and rewriting half of them. Or the work gets approved, but nobody can tell you whether it contributed to the business goal. Or one person spends the whole day reviewing what the agents produced.

Those are different problems. You would spend your money differently depending on which one you actually have.

I'm a fractional growth leader helping founders and CMOs build AI-enabled marketing teams, and I'm making these decisions in my own business. For 2027, I'm looking at the people, systems, and implementation work that turn AI output into something the business can use.

Start with one growth outcome you need the team to deliver. Then look at where the work gets stuck today.

Watch this week’s Weekly Signal on YouTube.

Let's use a fictional team. This isn't a client case or a result I'm claiming.

The team wants more qualified sales conversations from one customer segment. It has started using AI to make ads. Drafts arrive faster, but the creative lead keeps sending them back because the message doesn't match the customer's problem.

The team says it needs more production capacity. If we doubled that capacity, the creative lead might have twice as much work to send back.

I'd pull a few recent examples and ask the person rejecting them to explain why. Was the intended customer wrong? Was the offer unclear? Were the product facts inaccurate? Or was the idea good, but the execution wasn't there?

Then I'd look at the work that reached the market. Can we connect it to qualified conversations, or are we judging everything by whether we liked the creative?

Maybe the creative lead has the knowledge but no time to turn it into examples and standards. Maybe nobody knows how to improve the concept. Maybe the tracking is too inconsistent to diagnose performance confidently.

Don't turn those possibilities into three hires yet. Write down what you know and what you need to check.

For this fictional example, assume the approved product information is available and tracking is good enough for the question we're asking. The recurring problem is the concept: the team needs better ideas and a clearer explanation of why those ideas should work.

Now we have something specific an investment needs to change.

Making 50 versions of a weak idea doesn't give you 50 strong bets.

I want people who understand the customer deeply, pay attention to the market, and can turn that knowledge into something worth testing. That could be a hook, an offer, a format, or another reason for somebody scrolling the feed to stop.

The title matters less than whether the person can do the work. In our example, I'd want someone to develop a testable concept and explain who it's for, why that person would care, and what we're trying to learn.

AI can help develop variations around that direction. The person still needs to recognize which variations preserve the idea and which ones lose it.

In my own business, I'm hiring for creative that carries emotion and feels different. I'm also hiring YouTube expertise because I want someone who understands how to shape an idea for that channel.

Creator and affiliate relationship management is another real job. Recruiting people, understanding what they're good at, helping them make work for a brand, and sustaining those relationships takes ownership. AI can help with research, briefing drafts, and organizing information. I still want someone who knows the creator and is accountable for how that relationship goes.

If community matters to your growth strategy, somebody needs to make that experience worth returning to as well.

The budget question is which of these capabilities you're missing. Creative direction, channel expertise, production, and relationships aren't interchangeable just because they all contribute to content.

“Write down how we know an ad is good” is less exciting on a planning slide than “launch an autonomous marketing department.” Somebody still has to do it.

Experienced marketers often carry the reasoning in their heads. They can look at a piece of work and know something is off. Ask them why, and you start hearing things that never made it into the brief.

I'd fund the work of getting those decisions into examples. Show what passed, explain what failed, and turn the parts you can check into repeatable evaluations.

An evaluation is a test against a defined standard. Did we use approved product facts? Address the intended buyer? Include the required information? Which decisions still need an experienced person?

AI can perform some checks. We then need to check whether it's applying them correctly. A system approving its own work doesn't establish that the work is good.

And an ad passing a quality check doesn't establish that the ad will perform.

Someone needs to own the performance data too: what a campaign is connected to, what counts as a qualified result, and whether the information is trustworthy enough for the decision.

Our fictional team's tracking is adequate, so I'd focus its next investment on creative direction and documenting the standard. For a team with unreliable tracking, the data work might need to come first.

These capabilities can be shared. You don't necessarily need a full-time hire for each one, but you do need an owner and time to do the work.

In my practice, I'm seeing more overlap between work that used to sit in demand generation, outbound, and revenue operations.

Customer research, lists, messaging, ads, outreach, and follow-up are all trying to move the same prospective customers toward a buying conversation. For a defined segment, I'd consider giving someone responsibility for connecting that whole path.

That doesn't make one person the marketing department. It gives them a scope they can understand and improve.

If the person owns qualified pipeline for a segment, they need to know how someone gets on the list, what messages they see, what happens when they respond, and where the handoff to sales breaks.

An agent might help answer preliminary buying questions on a website. A straightforward landing page and form might work better for a particular audience. I'd want the owner to judge those options by what customers do and what sales receives, rather than by which department built the tool.

The role also needs working dependencies. Can this person access the research and data? Change the campaign? Get the CRM work done? Do they have the time and authority to fix a broken handoff?

Writing “own pipeline” in a job description doesn't answer any of that.

I've seen the limits of telling people to use all the tokens they want and go build stuff.

They still have existing responsibilities. Being interested in AI doesn't mean they have time to pull their whole process apart and work out a better one.

Someone needs to sit with the team, understand the work, build the workflow, document it, and help people learn to run it. I'm a little biased because this is what I do for a living. I'd still ask whether the person doing it understands the function they're trying to help.

Take a pitch deck builder for sales. It needs a narrative, product facts, pricing information, examples of the design standard, and guidance for different buyers. If those inputs don't exist, part of the assignment is working with the team to create them. Then somebody has to decide who reviews the deck and what that review covers.

That's why I want functional expertise involved. If the assignment needs a custom application, substantial integrations, or engineering that must be maintained, I'd bring engineering expertise in too. An engineer and a functional expert can be a strong combination.

Building custom software and teaching a marketing team to work with AI can involve different work. We need to be clear about which assignment we're funding.

Make the handoff part of what you're paying for. Can the team run the workflow and update it? Do people understand why it rejected something? Do they know when to ask for help?

The implementation should leave the team more capable.

Our fictional team's request began as “we need more ads.” After looking at the rejected work, the need became more specific: improve the concepts and make the reasoning clear enough for production to follow.

The human investment might be a creative strategist. It might also be protected time from the creative lead already on the team, paired with someone who can document and test the workflow.

The AI investment is capacity to develop and check variations within that direction. I'd look at the tools the team already has before assuming it needs another platform.

Implementation includes assembling approved inputs, creating examples, setting up checks, and teaching the team to run the process.

Put those investments together. What will the person do? What will the technology do? What has to be built or changed before the combination works? Who looks after it afterward? Because it's going to break.

Include review time. If generation is fast but approval takes half of someone's week, that time is part of the system's cost.

I wouldn't give every business the same percentage split between people and AI. A team with reliable data and strong creative leadership has a different gap from one that hasn't agreed on what a qualified lead is. Price the actual options using your costs, capacity, and required work, then compare them against the same outcome.

A short investment brief should make these points visible:

For our example, I'd establish the current revision effort and how often concepts are approved for testing. After the change, I'd look for evidence of usable work with less rework. I'd track qualified conversations from those tests separately. A better production process doesn't automatically mean the campaign wins.

If the evidence points elsewhere, change the investment. Good concepts waiting to be produced may call for production support. Untrustworthy campaign data may call for measurement work. Losing creators because nobody owns the relationships is unlikely to be fixed by another content generation tool.

That also helps distinguish a permanent hire from a scoped engagement. Is this an ongoing responsibility, or a concentrated piece of setup and teaching with a named person ready to take over?

Before your next planning meeting, put one role request and one AI request through this exercise together. Explain what each contributes to the same business outcome, what it costs to make them work together, and what evidence would change your mind.

That's a stronger budget conversation than debating whether a robot can replace a job title.

What are you planning to invest in for 2027? Is it a hire, a workflow, or help getting the team to actually use what you've built?

I'd love to hear where the work is getting stuck and what you're planning to change.

Email me what you're working through.