The Content Engineer leads the system and still ships.
Hire one person who is equally comfortable writing a prompt, sketching a content model, and shipping a page. They manage systems and agents, not people. They own the outcome end-to-end and keep their hands on the keyboard.
What they own
- The content model. Types, fields, taxonomy, relationships. The structural spine everything hangs off.
- The agent fleet. Prompts, system instructions, eval rubrics, and the workflows that route work through Claude.
- Quality & sign-off. Final judgment on what ships. The buck stops here.
- The pipeline. Brief → draft → structure → variant → QA → publish, wired together and measured.
- SME orchestration. Pulling the right human in for the right 20 minutes, not running a standup.
What they don't
- Run a team of writers, editors, and coordinators.
- Hand-write every word at scale. Agents draft, they direct.
- Own brand strategy or the GTM plan (that's the founder/CMO).
- Build the CMS platform from scratch (composable, off-the-shelf).
- Sit in a coordination layer. There is no layer.
In week two they've already stood up a draft→QA agent loop, can read a content model and tell you what's missing, ship a page without asking permission, and get visibly impatient with meetings. They talk in systems, not headcount.
The 30-second JD
This isn't a traditional content writer. It's someone with a content background who thinks in systems, as comfortable with the model, the pipeline and the integrations as with the prose. The craft still matters, but the bigger half of the job is now technical and systems-minded: structure, process, data, and the agents that run on top of them.
A content strategist who learned to build. Treats content as structured data, designs the model and the pipeline, and directs AI agents to produce against it, shipping to production without a department behind them.
- A draft → QA → publish agent loop, running.
- A read of the content model, and a list of what's missing.
- A page shipped to production without asking permission.
Illustrative profile of the single hire this edition is built around, the colour split shows how much of the modern role is systems and technical (purple) versus traditional content craft (teal).
One human in the middle, a team of agents around them.
The org chart is flat, not a pyramid. The Content Engineer directs a fleet of AI agents (the volume) and pulls in part-time SMEs (the judgment) only when needed.
Five agent roles do the volume.
Don't think of these as "tools." Treat them as roles with clear ownership, prompts, and eval criteria. Each maps to how Claude is wired into the pipeline. The human directs and the agents execute.
Drafting Agent
Turns a brief + brand voice + source notes into a structured first draft. Owns the blank-page problem so the human never starts from zero.
Owns · first-draft volumeStructuring & Tagging Agent
Maps prose into the content model: fills fields, applies taxonomy terms, sets relationships and metadata. Keeps everything queryable and reusable.
Owns · structure + tagsVariant Agent
Spins one approved core asset into channel and audience variants (email, social, ad, snippet, long-form) without redoing the work.
Owns · channel variantsTranslation / Localization Agent
Produces locale variants that respect glossary, tone, and market nuance, well beyond literal translation. Flags terms that need a native reviewer.
Owns · localized outputQA / Eval Agent
Scores every draft against a rubric (accuracy, brand, structure, claims, links) and surfaces only what a human needs to look at. The safety net before sign-off.
Owns · automated checksHuman-in-the-loop
The Content Engineer reviews QA flags, makes the judgment calls agents can't, and signs off. Agents propose the work and the human decides what ships.
Owns · judgment & sign-offEnhance the creative parts. Don't automate them away.
Agents are brilliant at volume, structure and speed. What they don't have is an original idea or taste, the judgment of what is actually good, on-brand and worth saying. Those are the most human parts of the work, and the whole point of handing the grind to agents is to give the human more time for them, not less.
Idea generation and taste are a duet: the AI widens the funnel, the human's judgment picks the winner. That's where creativity gets amplified, not handed over.
Stays human
Enhance these, they're the edge.
- Original ideas and the angle worth taking
- Taste, knowing what is actually good
- Point of view and brand personality
- Emotional nuance and reading the room
- The final call on what ships
AI amplifies
Hand these over, they're the volume.
- First drafts and the blank-page problem
- Channel and audience variants at scale
- Structure, tagging and metadata
- Research, options and first-pass QA
- Speed, consistency and reach
Borrowed brains, not new headcount.
The human team beyond the one hire is part-time and purposeful. You rent expertise in small slices and give it back.
Founders & marketers as SMEs
Your domain experts review in slices, not shifts. The founder validates a positioning claim, the product lead checks a feature detail, sales flags what resonates. Twenty focused minutes, not a seat on the team.
Taxonomy / knowledge-graph specialist
Bring in a specialist for the setup spike. They design the taxonomy, the content model, and the knowledge-graph relationships properly the first time. A few weeks of expert work, documented, then they're gone. You don't keep them on payroll.
The fractional specialist is a spike, not a hire. If the engagement quietly turns into an ongoing retainer, you've recreated the department you were trying to avoid, just with a contractor invoice. Time-box it and capture the work as docs the Content Engineer owns.
Who's accountable vs. who does the work.
One human is Accountable for every activity. The buck stops with them. AI agents are the Responsible executors that carry the volume. SMEs are Consulted at the points where human judgment matters. Hover a row to focus it.
| Activity | Content Engineer (human) |
AI Agents (Claude) |
SMEs (part-time) |
|---|---|---|---|
| Model content | A | C | |
| Create / draft | A | R | |
| Tag / structure | A | R | |
| Variant | A | R | |
| Translate / localize | A | R | C |
| QA / eval | A | R | C |
| Publish / sign-off | A |
Read down the human column: Accountable for everything, Responsible for almost nothing. That's the whole model. Agents take the volume, the human keeps judgment and the final yes.
One owner, no handoffs.
Running the whole content workflow through one owner and a set of agents means there are no seams to hand work across, and no coordination layer to keep in sync. That's where most of the speed comes from, and it's a deliberate trade worth naming out loud.
You give up deep specialisation in every seat. You gain speed, far lower coordination cost, and one owner who's accountable end to end. For a funded startup or mid-market team, that trade is almost always worth it, until content volume genuinely forces a second hire.
Before you move on
By the end of this page you should know exactly who runs your content engine, one accountable human, a fleet of agents, and a couple of borrowed brains. Hold the four takeaways below before you turn the page; the next chapter assumes you've settled the team shape.
What you should have now
- ✓ One accountable owner, a single Content Engineer who owns the model, the agents and the final yes, with no coordination layer underneath.
- ✓ The Content Engineer profile, a content background wired to a systems brain, where most of the modern job is structure, pipeline and agents, not prose.
- ✓ The human edge, named, ideas and taste stay human and get amplified by agents, while drafting, variants and tagging are handed over for volume.
- ✓ The trade-off, accepted, you give up deep specialism in every seat to gain speed, low coordination cost and one owner accountable end to end.
Why this sets up the next stage
A lead owner and a fleet of agents only get to work once they have something to run on. Next, The Stack lays out the modern, AI-native tooling, CMS, agents, integrations, that turns this team shape into a working pipeline.
So the order holds: know the team, then choose the stack. You've decided who's accountable and what the agents do; the next chapter wires them into the systems they operate.