What my content team looks like in 2026

We had ClaudeCode running, the repo was set up, and we genuinely believed that once the team saw how powerful this was, they'd be in. Some of them were. Plenty weren't - because plenty of the team didn't want to build in a terminal environment, and that reality changed how I think about content engineering entirely.

So let me tell you what my content team looks like in 2026, and why it's a more honest version of what a modern content operation should be.

Content strategy still sits with me, the CMO, and the CEO. Where are we going? What are we speaking about? How do we translate the business narrative into stories that are worth a target audience's time? How do we reach them, distribute, scale? That's still human work. If it fails, I'm accountable. That hasn't changed.

What changed is everything underneath it.

There's a layer now that I call content engineering. There's no better word for it, so here we are. Every process with safe steps, a clear input, clear templates, clear examples, and a defined output can probably be handed to an agent. It gives the team back the 70% of their week that was grunt work rather than strategy or craft.

Before all of this, a solo content operator's week was mostly execution: a lot of copy-pasting, report-building and CMS-loading, with a rushed quality check bolted on at the end. The copying and pasting from a Google Doc into a CMS, the building of reports, the putting-together of decks, the loading of a blog post into Webflow which still takes fifteen to twenty minutes per post per person - time freed from that goes to higher-value work. Removing that execution grunt work creates space for the thinking that actually moves the business forward.

Agents are structured workflows - archetypal sequences of steps that carry a lot of context, move between systems, and handle multidisciplinary tasks without needing someone to sit in the middle passing information between windows, doing the job.

Here's one I run at Backbase. We've got a full blog agent workflow. It starts with a search analysis, surfaces the topics we can credibly own, runs a SERP analysis on the top five blogs, then pulls a Google AI Overview analysis via API to understand what Google is surfacing for that topic. It saves that into a database. It scrapes about 30 links from those results so there's no hallucinating - the agent has to find the page and use something real from it before it's allowed to cite it. Then it goes into our knowledge base and looks for a proprietary angle: a point of view, relevant data, a transcript from someone who has spoken about this topic. It builds a brief, writes the draft, does internal linking based on our cluster strategy, runs an external link sweep, and then passes everything through our tone of voice framework. Final output. About twelve minutes. Someone spends five minutes editing. We publish.

I remember the first time it ran end-to-end without intervention - I checked the output expecting to find something broken, and it was just... good. That moment is the kind of thing that's hard to explain to someone who hasn't sat through years of briefing writers, chasing drafts, and fixing formatting errors at eleven at night.

The split now looks more like: 40% strategy, 25% quality control, 35% content engineering. The grunt work is gone, and what replaced it is the architecture - building the workflows, maintaining the system, making sure the agents have what they need to do the job well. That is now a core function of a content operation, and the sooner teams accept that, the faster they'll move.

Consistency deserves its own mention. AI does what it's told. If you confine it to a knowledge base, a rule set, tone of voice examples, and a defined brief, it will stick to brand with a reliability that's genuinely hard to match in a team of humans. Humans have off days, different writing styles, different levels of care on different mornings. An agent, set up well, doesn't drift. Which means brand consistency has become an engineering problem, and a solvable one.

What we realised at Backbase was that the platform needed to be separated from the terminal. Leadership was serious about scaling this. We brought in serious AI developers to maintain the infrastructure and guardrails, and then built a workspace where I could create agentic workflows that don't break inside a controlled system. The team executes on top of that, without ever needing to touch the underlying code.

That's the model. I build so they don't have to copy-paste. They execute and edit and add the judgement that a model can't replicate.

And this isn't just an internal restructure story. It's a signal about what the content function is becoming. The orgs building this architecture now - rethinking their tools, their SOPs, their job descriptions, their workflows - are moving at a serious pace, and they're compounding that advantage every month.