AI content production: Where does it work, and where doesn't it?

We have been actively using AI tools in our clients' content processes for a long time. The result shows that the claims at both extremes are wrong: AI is neither a "one-click agency" nor a "soulless text machine". Used in the right place, it multiplies production speed; used in the wrong place, it makes your brand sound like everyone else.

Where it works

• Content calendar and idea pool: Coming up with 20 post ideas for a month is now a half-hour job. A human still decides which ideas fit the brand.
• Generating variations: 10 different ad headlines and descriptions in 5 different lengths from a single approved message. Ideal for A/B testing.
• Turning long-form into short-form: A LinkedIn post from a blog article, Reels captions from a video script, an email summary from a press release.
• Visual drafts: Producing initial visuals to show a concept to the client. The final visual still usually comes from a designer.
• Interpreting data: Translating GA4 and ad reports into plain language and flagging unusual changes.

Where it doesn't work

• Brand voice: By default, these tools write in an "average internet" tone. Without a brand voice guide, the output sounds like everyone else's.
• Local and industry knowledge: On topics like prices, regulations and product details, it can be confidently wrong. Every figure and claim must be verified by a person.
• Community management: Auto-replying to comments is the quickest way to escalate a crisis.
• Originality: "Tips" content that everyone produces with the same tool no longer resonates with either algorithms or people.

A practical workflow

1. Create a brand voice guide: one page covering tone, banned words and sample sentences. Give this guide to the tool every time you produce something.
2. A human sets the monthly content themes; the tool generates post ideas and drafts from these themes.
3. Drafts are edited, and concrete examples and brand-specific details are added.
4. Visuals are produced with the tool at the concept stage and by a designer at the final stage.
5. Performance data is summarised with the tool every month, and the next month's themes are updated accordingly.

The bottom line: AI isn't replacing the agency; it enables the agency to produce more, and more consistent, content with the same budget. The winners are those who get the process right, not the tool.