How we cut a startup's content cost by 60% without losing quality

The starting position

The client — a 24-person B2B SaaS company — was spending roughly $18,000 per month across three freelance writers, an agency retainer for design, and a part-time editor. Output was twelve articles a month. Organic traffic had been flat for three quarters and only two pieces had ever generated a demo request.

The problem wasn't writing quality. It was that nobody could say which pieces were supposed to do what.

Diagnosis: volume without a job description

We audited 140 published pieces and mapped each to a job: rank, convert, enable sales, or build authority. Sixty-one percent had no job at all. Those pieces consumed budget and diluted internal linking.

• 61% of published content had no defined commercial job
• Average brief length: 3 lines; average revision rounds: 3.4
• No topical clustering — 140 pieces, 118 orphan URLs

The rebuilt pipeline

We replaced the freelancer-per-article model with a four-stage pipeline: structured briefing, AI-assisted first draft, subject-matter enrichment by an internal expert, and a single editorial pass against a published quality rubric.

Crucially, we cut the writer count from three to one senior editor and moved the budget into research, original data and design. The model drafts; the human owns the argument, the examples and the claims.

Quality control that actually holds

Every piece had to pass a five-point rubric before publication: a defensible thesis, at least one non-obvious insight, original or primary-sourced data, a named reviewer, and a specific next action for the reader.

Anything failing two points went back. In month one, 40% failed. By month four, 9% did — the briefs had improved, not the model.

Results after six months

Cost fell from ~$18,000 to ~$7,200 per month, a 60% reduction, while output rose to eighteen pieces a month. More importantly, the work started producing.

• Organic sessions: +118% in six months
• Content-attributed demo requests: 2 → 27 per month
• Average time from brief to publish: 19 days → 6 days
• Editor satisfaction: unchanged (we asked — it mattered)

What we'd do differently

We under-invested in design for the first two months and it showed in social performance. Visual systems should be built in parallel with the writing pipeline, not bolted on after.

The transferable lesson: AI didn't cut the cost. Deleting work that had no job did. AI just made the remaining work faster.