Medium: A scalable AI writing workflow for freelancers

Florence de Borja is a freelance writer and AI content strategist with fourteen years of writing experience and a background in IT. Her February 2026 article addresses a problem common to freelancers who have adopted AI tools but not a system: productivity gains disappear when writers switch between research, drafting, and editing simultaneously, using AI at each step without clear handoffs between them.

Her central argument is that the bottleneck for most AI-assisted writers is not capability but friction. De Borja cites two data points: a Nielsen Norman Group study finding that generative AI increased throughput by an average of 66%, and an MIT study showing ChatGPT reduced completion time for workplace writing tasks by 40%. She argues these gains are only realizable when AI is deployed within a structured process, not used opportunistically across all phases at once.

The system she proposes has four stages with explicit boundaries between them. The first is research: before touching any drafting tool, the writer builds a structured brief that defines the core claims the piece needs to make, the intended audience, available evidence, likely counterarguments, and the expected takeaways. This brief becomes the prompt input for everything that follows, which is why de Borja calls the principle “clean inputs produce clean drafts.” The second stage is outlining: the brief feeds into a hierarchical outline, with each section assigned a target scope of 500 to 800 words and a clear objective. AI is used here to improve structure and flag gaps, not to write.

The third stage is drafting, which de Borja treats as the highest-stakes phase. The writer drafts first; AI enters only to tighten sentences or flag unclear passages. This sequence — human first, AI second — is what her system is designed to enforce. Inverting it produces output that reads smoothly but lacks the specific judgment, angle, or voice that came from the writer thinking through the material. The fourth stage is editing, structured as four sequential passes: structure, then voice, then precision, then polish. Running these simultaneously is what she identifies as the primary cause of revision cycles stretching beyond two rounds.

The article also includes a measurement component: de Borja recommends tracking revision round count and drafting speed on a weekly basis and adjusting the system based on what the data shows. This makes the workflow adaptive rather than prescriptive.

Useful for freelance writers working across multiple clients, and for editors managing AI-assisted contributors who are producing first drafts that require heavy revision. The method applies most directly to long-form work where voice and argument quality matter more than speed.