A content designer I spoke to recently told me her company made an overnight switch to a Claude Code-first workflow. Designers, PMs — everyone was told to start pushing code.

Seemingly overnight, she was working 17-hour days.

The volume of strings being created jumped 3x. Every code change had new copy in it. Her day turned into @-mentions in GitHub PRs and late nights combing through the codebase — figuring out what had customer-facing text in it, what was risky, what had already shipped before she could see it.

The tools made everyone else faster. And that speed became her problem.

The narrative doesn't match what's actually happening

There's a story being told about content design and AI: as generation gets easier, the need for content designers goes down.

The content designers I've talked to on the front lines are describing something completely different.

In AI-first workflows, the volume of copy explodes. It's in every PR, every prototype, every agent spinning up a new UI. Copy is being generated continuously, by everyone, often without anyone flagging that a content decision was made at all.

And those teams are discovering that AI isn’t thinking about cohesive language across a product.

This is the same reason design systems have become so critical in AI-first workflows — to keep the UI visually consistent when a dozen different people and agents are building in parallel. Copy has the same problem, and far fewer teams have solved for it.

Agents don't build intuition across sessions. Each project starts fresh. They're optimized to produce output for the task in front of them — a single screen, a modal, a feature — without any sense of how that piece fits into the larger user journey. They don't think in terms of cross-functional workflows or multi-step user experiences. They think linearly: what does this page need?

Meanwhile, these things are a content designer’s superpower. Contextualization, user empathy, the ability to look at a string in isolation and know whether it fits — or breaks — the experience users are actually having across your product. That's not a skill you can prompt into existence. It's built from experience, from knowing the product, from caring about the human on the other end.

The problem isn't that this skill is less valuable. It's that the conditions for applying it have gotten significantly harder.

The work doesn't shrink. It shifts.

Just because speed is increasing doesn't mean quality standards are decreasing. Every one of those strings still needs to be right — helpful, on-brand, compliant, contextually appropriate.

What's changing is where the work happens.

Content designers are no longer primarily responsible for writing or approving individual strings. They're responsible for the systems that make review possible at dramatically increased volume and cadence. That's a different job. It requires different infrastructure. And most teams haven't built it yet.

Governance — the unglamorous, structural work of making sure the right copy ships reliably — has become the bottleneck. Not writing. Not even reviewing. Governance: the scaffolding that determines whether review is even possible at scale.

What governance actually needs

There are four things content designers need to build to make this work.

1. A source of truth that agents and humans can both reference

Not a PDF that lives in a shared drive. Not a stagnant markdown file. An accessible, connected set of rules, guidelines, terminology, and approved language that exists where the work is happening. When an agent generates a screen, it should be pulling from this. When an engineer writes a string, they should be able to check against it. The source of truth has to travel with the workflow.

2. AI as your own enforcer

The monotonous reviews — does this follow capitalization rules? Is this terminology on the approved list? Does this error message follow the format we use everywhere? — shouldn't require a human every time. A well-configured AI can catch the obvious errors and filter out the noise, so the humans in the loop are spending their time on the decisions that actually need human judgment.

3. A single place for high priority work

When something does need human attention — a compliance-adjacent string, a new pattern that hasn't been established yet, copy in a high-stakes flow — there needs to be a centralized place to surface it, discuss it, and resolve it with the right people in the room. One place, with all the context needed to make the right call.

4. A system that compounds

The biggest failure mode in content systems isn't building them — it's letting them go stale. Every time your team ships something, that's an opportunity to refine a rule, add a terminology entry, update a guideline based on what you learned. A source of truth that doesn't grow with you stops being true. Build the habit of maintenance into the workflow itself, not as a separate quarterly audit.

Where to start

These things can be built with individual skills, markdown files, GitHub repos, and a well-configured CLAUDE.md. That's a legitimate starting point and, for some teams, the right one.

For teams that want a purpose-built system — one where all four of these come together, where the source of truth connects directly to the agentic tools your team uses, and where governance is infrastructure rather than a manual process — that's what we're building at Ditto.

But regardless of where you build it: the teams that figure out governance now will have a significant advantage over the ones that wait until the volume becomes unmanageable.

AI made generation cheap. It made oversight expensive. The content designers who thrive in this era won't be the fastest writers. They'll be the ones who build the best systems.

Interested in what a governed content system looks like in practice? See what we're building on our public roadmap →