Stop asking what AI can do: start asking what humans can do now that they could not do before

There's a point in every content leader's career where the job stops being about writing and starts being about architecture. I hit that point, and it changed how I think about everything from org structure to what "grunt work" even means anymore.

AI is more consistent than any roomful of humans producing content together. That's true. The nature of a large language model is to predict the most probable next word in a sentence - that's it. So when you pull it away from the open internet, give it a confined knowledge base, a rule set, writing examples, tone of voice guidelines, and you say "this is how you write" - its job is still just to predict the best next word in that context. That's what makes it so consistent. AI sticks to brand. If humans are only editing at the end, there's far less brand skew than in a traditional team-based workflow. That's a genuinely underappreciated point.

So here's what my team looks like in 2026.

Content strategy still sits with me. Where are we going, what are we speaking about, how do I translate the business narrative into stories the market actually cares about, how do we reach the audience, how do we distribute - that's mine. I own that entirely. But then what we've layered in is the content engineering side. Every process that has clear steps, clear inputs, clear templates, a defined output, and specific steps the team needs to follow - that's something an agent can do. And my job has become building those systems so my team doesn't have to copy paste things into a CMS, doesn't have to manually build reports, doesn't have to track down data before they can write a single sentence.

We run agentic workflows for everything from SEO planning to content drafting, insight surfacing, analytics and reporting. One example that illustrates the scope: someone presses one button with a brief document for a webinar, and the system produces the full content package for that asset - saving the team serious time that would have otherwise gone into manual coordination, formatting and distribution.

We run two systems in parallel - my team in one, me in ClaudeCode. We started by getting everyone into the terminal and said "use this, it's powerful" - and then realised pretty quickly that building in an agentic environment is genuinely complex, and not everyone wants to live there. So we brought in some AI engineers, got leadership bought in, and built our own platform. I'm doing the "vibe coding", the devs are holding the guardrails, and the team just executes in the platform we provide them.

This shift demands you rethink everything first. You have to rethink the org, the processes, the SOPs, the tooling, the jobs. Everything looks different through this lens.

Here is what things looked like in 2023 - just before everyone and their mum started using ChatGPT.

Back then, a content marketer's time broke down roughly like this: 15% went into strategy, 70% into what I call groundwork, and the remaining slice into QA. That 70% wasn't wasted in any meaningful sense - historically, it was just how the work got done. Loading a blog into a CMS, even with a decent one, took 15 to 20 minutes per post per person when you factor in the author profile, the image flow, the metadata. That's copy pasting, moving things from a Google Doc into a different form, finding data, building reports, putting slides together. Nobody loved that part, and nobody pretended to.

Agents changed the calculation on that 70%. They're not little minions doing the washing up while you play video games. What they actually are is archetypal workflows - a set of steps that carry context and move between systems, handle multi-disciplinary tasks, and do it without dropping the thread. As one real example: our blog agent starts by identifying keyword opportunities, runs a SERP analysis on the top five results, then does a Google AI snippet analysis through the API three times to understand what Google is surfacing as the consensus answer for that topic. It saves that into a database, scrapes around 30 links that show up in the results, and analyses what each one is bringin to the table, and wha'ts missing. Then it writes findings into a knowledge base, and another one looks at that knowledge base and asks: do we have a proprietary angle here? Do we have data? Do we have a transcript from someone speaking on this? It finds the relevant source, builds a brief that says "here's the topic, here's the knowledge base material, here's the transcript extract, here's the data" - and then writes the draft from that, doing internal linking across our cluster strategy and external linking through scraping (so it can't hallucinate a URL that doesn't exist). Then there's a swap-and-clean pass that runs tone of voice checks against our framework and eliminates certain writing patterns we've trained it to avoid. Final output, H1 written last, specific template, about twelve minutes total. Someone spends five minutes editing it, and it's done, and published to Webflow with a click of a button.

So what does a content marketer's time look like now? Way more strategy and meaningful QA, with almost none of the grunt work that used to eat the day. The ratio has flipped.

The instinct a lot of leaders have when they hear this is "so you need fewer people." The answer is different people doing different things. The content marketer who used to spend 70% of their day moving files between systems now spends that time thinking and sharpening the strategy. That's a better use of a skilled person. It's also a better use of the budget.

I personally believe it's an exciting time for all of us, but it requires us to accept that AI is not the enemy. AI does not create slop, humans do. But AI can take a serious amount of grunt work off the table if you are willing to rethink and reinvent.