TL;DR

AI workflows for content teams were supposed to fix the capacity crunch. Instead, for many marketers, they've become a second job layered on top of the first, and when the numbers slip, it's the individual explaining themselves to their manager, not the leadership decision that created the squeeze. The setup work alone (prompting, refining, troubleshooting, designing repeatable processes) eats hours before a single deliverable ships. Do that on top of a full workload and the tool sold as the cure for burnout becomes the cause of it. There's a second cost leadership rarely prices in: turn your best marketers into full-time prompt engineers now, and in two years you'll have a real shortage of people who can actually write, position, and tell a story, the skill AI still can't replace. The real fix is resourcing, clear ownership, and leadership actually owning the trade-off instead of quietly passing the risk down to whoever's closest to the keyboard.

Intro

"I don't want to start new stuff with AI anymore... I spend four hours arguing with a machine and I could've just spent that time being useful."

Sound familiar? It's the exhaustion you don't hear in the keynote slides. AI adoption is failing marketers, not the other way around, and the rollout is the reason: full-time output, plus full-time experimentation, plus full-time workflow building, with no corresponding capacity decision to match. That gap between what's expected and what's resourced is the part that gets skipped.

This piece unpacks the organisational cost of AI adoption, and specifically, who's accountable for it. Not the philosophical case for humans in the loop (we've covered that in AI Content Workflow: Why Humans Stay in the Loop) and not a how-to guide. This is about what happens when you treat AI workflows as a side project instead of a resourced, leadership-owned rollout, and the workforce risk that creates if nobody course-corrects.

‍

AI workflows for content teams feel slower before they feel faster

The pitch is speed, but the reality is delay. AI content workflow setups take longer than manual execution of the task they're supposed to automate. One of the podcast hosts put it plainly: the real payoff only lands “in six months when you've finished fixing and making this really solid workflow.”

Until then, you're iterating. Prompting. Fixing outputs that almost work. Rebuilding workflows when context changes. The math the podcast floated: spend three hours now, save a hundred later, in theory.

That's a good trade if someone actually budgets the three hours as work, and not as something a marketer finds in the gaps between everything else already on their plate.

The problem starts the moment AI gets bolted onto an existing workload with no corresponding capacity decision. Teams expect gains from day one, and when that doesn't happen, they assume the problem is skill, or effort, or the tool itself.

Rarely do they ask: did we give this work the time it needs, and did anyone with budget authority actually sign off on that?

‍

The setup work nobody budgets for

Here's where it gets uncomfortable.

When a podcast guest described the ask (step back from output, learn the tools, build workflows, turn them into repeatable systems, train the rest of the team), the host's response was immediate and flat: “But that's a full time job.” And then: “It's another job on top of your job.”

That's the crux. AI adoption for marketing teams has been framed as a productivity gain. However, the work of designing, testing, and maintaining a repeatable AI workflow is real work. It requires time, skill, and someone's attention, consistently, over months.

And when that work gets assigned as overtime, it gets done resentfully. Or incompletely. Or not at all.

TLDR: The workflow build is a job. If you don't resource it, you're asking someone to do two jobs for the price of one.

‍

Who absorbs the cost when AI adoption isn't resourced

A marketer spends hours wrestling with a tool that was supposed to save them time. The workflow doesn't pay off yet. Then she's the one in a one-on-one, explaining why her numbers dipped.

The interesting question is who absorbs the cost when AI adoption isn't resourced properly, not just whether it got resourced in the first place. If leadership chooses the rollout but an individual marketer absorbs the lost hours, the missed targets, and the performance conversation that follows, an organisational trade-off has quietly become a personal one. The cost of the decision doesn't disappear just because nobody wrote it down. It runs downhill until it pools on whoever has the least power to redirect it (usually, the person closest to the deadline).

If a business decides content workflows should use AI, that's a resourcing decision, the same category as buying a new CMS or restructuring a team. Someone with the authority to move targets, timelines, or headcount needs to own what it costs, out loud, before the work starts:

• The trade-off gets written down before rollout, not implied.
• A dip in output during the ramp-up counts as expected, not as a red flag.
• If a manager brings up someone's numbers, the AI-adoption cost is already priced into that conversation.

None of this needs a task force. It needs one sentence from someone senior: “this will cost time, and I'm accounting for it.” Say that before the work starts, and nobody has to defend themselves for a decision they didn't make.

‍

Workload inflation isn't transformation

Here's the trap: a team adopts AI tools, leadership sees “efficiency gains” in the pitch deck, and suddenly the expectation is the same output plus new capabilities, all from the same headcount.

That's workload inflation dressed up as transformation.

The podcast hosts called it out directly, citing teams “acting like it's easy” while expecting marketers to build and run AI agents as “some side project while doing all their jobs.” The result? As one host put it: “the antidote to burnout is the cause of the burnout.”

It's a serpent eating its tail. The tool meant to reduce pressure increases it, because the labour of making it work hasn't been accounted for. And the people caught in that bind are the same ones who were already stretched.

‍

Who should own AI workflows in content teams

The podcast made a clean argument here: “let the engineers be engineers and let the marketers be marketers.” Marketers bring the strategy, the use-case knowledge, the editorial instinct.

That's a useful starting instinct, but the real fix isn't about which job title owns the workflow. Plenty of teams don't have dedicated engineers or AI ops staff, and plenty of AI workflows need someone who understands both the marketing use case and how to build and iterate the system, which isn't a job that splits cleanly along a marketer/engineer line.

What actually matters is that workflow ownership becomes an explicit, resourced responsibility, assigned to a specific person with the time to do it. That person can be a marketer who wants the role. It just can't quietly become everybody's second job by default.

One example from the episode: a creative pitches an idea, and the head of AI turns it into something usable. Whoever holds that responsibility, the point is that it's a named, resourced role, not an assumption that whoever's nearest the tool will absorb it.

Not everyone has to build an AI agent, and not everyone should, but that's a different claim from saying marketers shouldn't develop these skills. Comfort and interest in this kind of tooling vary by role and by person, and choice matters here: a marketer who wants to specialise in workflow building should be able to. The problem is making workflow engineering a blanket expectation for every marketer while still expecting their previous output at full volume. Some of your strongest storytellers will never want to touch a workflow diagram, and forcing the ones who don't want to doesn't make them better at either job.

This matters because AI doesn't change what needs to happen in your workflow. It changes who does it. The six-step production process (briefing, research, drafting, review, localisation, publishing) stays the same. What changes is the actor inside each step: human, freelancer, or AI agent. And that structural shift makes workflow design even more important.

A broken process run by AI breaks faster and at higher volume.

TLDR: Workflow ownership needs to be explicit and resourced. Nobody should be quietly pressured into a second job they didn't sign up for, and nobody who wants the role should be discouraged from it either.

‍

The workforce risk nobody's pricing in

There's a longer bill that may be coming due here, and it's worth naming as a risk, not a certainty.

Say the path of least resistance wins: your marketers spend this year becoming prompt engineers and agent builders. Fine, short term. But that's time not spent on something else, and the podcast's own prediction is worth treating as a hypothesis worth taking seriously, not a forecast:

In two years, we're gonna see all those people who trained to be AI... agent builders or whatever, who were marketers. It's gonna be a real gap for real marketers in two years who can actually do marketing work.

The sharper version of that risk: what capabilities are teams choosing not to develop while everyone learns workflow building instead? If junior and mid-level marketers spend less time on research, positioning, interviewing, editing, and storytelling now, the open question is where tomorrow's senior marketers acquire those skills at all.

Every hour a strong writer spends debugging a workflow is an hour not spent on editorial judgement, the thing nothing else replicates. And the market doesn't stop wanting that skill just because AI got louder. If anything, the whole argument of this episode, that fundamentals win and commodity content is worthless, means real writing gets more valuable exactly as companies risk training theirs out of existence.

This is the same accountability question from a different angle. Protecting time for the people whose job is explicitly not to build workflows is a resourcing decision. Ask who's left to actually write the content, and who's positioned to grow into a senior storytelling role, if everyone on the team spent this year learning to prompt instead.

‍

AI workflows for content teams still need human oversight after launch

Here's the fantasy: build the workflow once, set it running, walk away. Here's the reality: you're still babysitting it months later.

The podcast cited technical SEOs at a meetup who'd spent over a year building workflows and the realisation that local tools still require supervision. The reason: AI outputs remain probabilistic and can be wrong, so workflows need QA and oversight designed into them from the start, not bolted on after something goes awry.

The workflow is the product: a one-off piece of output is not what you're actually building.

Even a well-built AI content workflow needs ongoing input: prompt adjustments as the product changes, edge-case handling when the model confidently produces nonsense, and editorial sign-off to catch the outputs that shouldn't ship. That work doesn't disappear. It shifts.

What resourcing looks like when you do it properly

The good news: some leaders have figured this out.

The podcast highlighted one: a leader who “built in hours into everybody's work so that everybody had a dedicated amount of time every week to do AI training.” Not optional lunchtime learning. Not a Slack channel of links. Paid, scheduled, protected time. Plus workshops, “fully paid for, built into their schedule.”

That's the difference: AI workflow setup becomes a real line item and training becomes a legitimate use of working hours, and leaders take the accountability question above, and actually answer it.

Even with that structure in place, it's still hard. People adapt to new tooling at different speeds, and a newly resourced rollout needs more hands-on support than a mature one does.

What your next staffing conversation needs to include

AI won't resource itself, and neither will the workflows that make it useful, or the accountability for getting this right.

If you're a content leader planning AI adoption, ask: who owns this, how much time does it take, where does that time come from, and who answers for the trade-off if it goes sideways?

If the honest answer to that last one is “whoever's numbers slip will have to explain it,” you haven't made a resourcing decision, and it will surface as a personnel problem instead of a strategy conversation.

The teams that get this right treat AI workflow work as operational infrastructure and staff it accordingly. Just as importantly, they protect a deliberate share of their team's time for the storytelling and editorial judgement AI still can't replicate.

Want to learn more about getting back to your storytelling roots? Watch the full episode.

‍

FAQs

What if leadership won't approve extra hours or headcount for AI setup and training?

Then you have a choice: reduce output expectations during the ramp-up, or accept that the rollout will take longer and produce more friction. Framing AI as “free productivity” is the root of the burnout problem. Show the setup work as a line item, and make the trade-off visible, and owned by someone with the authority to approve it, before it becomes a personnel issue.

If my output slips while I'm learning to use AI, is that on me?

Not if leadership asked you to adopt AI without adjusting your targets. That's a resourcing decision made above you. The accountability for its cost belongs with whoever made it. If your manager raises a dip in output during a documented AI ramp-up, that's the moment to point back to the resourcing conversation that should have happened before you started, and to have it now if it didn't.

Do all of our marketers need to learn how to build AI agents?

No. Treating it as a blanket requirement causes its own problem. The podcast's argument is blunt: let engineers be engineers and let marketers be marketers. Plenty of your strongest writers won't want to build workflows, and forcing them trades away the skill that's hardest to replace. And if too many marketers spend their time becoming agent builders instead of writers, you risk a shortage of people who can actually tell a story right when that skill gets scarce and valuable again.

We don't have engineers. Who builds the workflows?

You have a few options: hire or contract someone with AI ops skills, train a curious generalist already on your team, or work with an external partner who specialises in workflow design. The point is that someone needs to own the wiring as their actual job, with time carved out for it. That person does not have to be an engineer, but they do need protected hours and accountability.

How do we know when an AI workflow is mature enough to trust with less oversight?

You don't remove oversight entirely. You reduce it once the workflow produces consistent output across edge cases: different topics, formats, and inputs. The podcast describes this as a six-month runway before workflows become “really solid.” Even then, expect ongoing prompt adjustments and editorial sign-off for the outputs that need a human eye.

Isn't hiring or upskilling a dedicated AI ops person just more cost?

In the short term, yes. In the medium term, it's cheaper than the alternative: a team that burns out, workflows that never mature, a productivity tool that delivers negative ROI because nobody had time to make it work, and a talent gap in the one skill, real storytelling, you'll need most once the market floods with generic AI output.