Steps to create an AI-assisted writing workflow
This resource is based on The inventor's mindset that will save freelance writers in the AI era, featuring Jennifer Goforth Gregory , published on the AI Lab by ActiveCampaign.
Get the checklist
By the end of this checklist, you’ll have a repeatable workflow for using AI to brainstorm, outline, and research an assignment without ever using it to write. It’s based on the approach Jennifer Goforth Gregory uses for her B2B tech writing clients. Setup takes about 30 minutes; the workflow runs in about the time a single assignment takes.
Before you start
- Confirm client permissions: read the client’s AI policy. Some clients ask writers to sign a form stating they won’t hand in AI-generated content. If the policy isn’t clear, ask.
- Pick your tools: you’ll need ChatGPT (or another LLM) for brainstorming and outlining, and Perplexity for research with cited sources.
- Open a fresh doc: keep AI outputs in a separate tab or scratch doc. The rule is to never copy and paste from the AI into your draft.
The workflow
Phase 1: Brainstorm pitches and angles
After this phase, you’ll have: a shortlist of story angles you can pitch or turn into an outline.
- Prompt for topic trends: use AI the way you’d use Google — to surface ideas, not to pull quotes from.
Starter prompt
What are some trends in [INDUSTRY] right now? What are the biggest challenges [ROLE] is facing with [TOPIC]?
- Read what the AI surfaces: treat the output like a search results page. Read more on the topics that interest you, then let your own reading spark the angle.
- Write the pitch in your own words: none of the AI output goes into the pitch verbatim. Gregory has never taken AI output straight — she uses it to spark her own creative spin.
Phase 2: Outline the piece
After this phase, you’ll have: a section-by-section outline with working headlines you’ll rewrite before filing.
- Confirm the outline is fair game: outlining with AI is more common than brainstorming. If the client’s policy allows it (or encourages it, as one of Gregory’s does), move forward.
- Prompt for an outline: feed the angle and ask for main headlines, sections, and supporting points.
- Edit the outline heavily: Gregory writes from the outline, changing it as she goes. The AI draft is a starting point, not a contract.
Phase 3: Brainstorm headlines and subheads
After this phase, you’ll have: a working headline and a set of subheads you can revise during drafting.
- Ask for headline options: request 10–15 variations in different angles (question, how-to, counterintuitive).
- Combine and remix: Gregory never takes a headline verbatim. She picks phrases she likes, combines options, or lets the AI’s attempt spark a better one of her own.
- Draft subheads the same way: use the AI output as a set of raw materials, not a final product.
Phase 4: Research and source-hunt
After this phase, you’ll have: a list of statistics and supporting sources, each verified against the original report.
- Switch to Perplexity for research: Perplexity is Gregory’s research tool of choice because it surfaces cited sources alongside answers.
- Ask for statistics on a topic:
Starter prompt
What are the latest statistics on [TOPIC]? Include the original source and year for each.
- Verify every statistic at the source: open the original survey, report, or article. Never quote the AI’s summary. Gregory treats the AI output like Wikipedia — useful for finding leads, never taken as fact.
- Cite the primary source: link the reader to the original research, not the AI answer.
Quick reference
- Total time: ~30 minutes setup, then runs assignment-by-assignment
- Tools needed: ChatGPT (or equivalent LLM), Perplexity
- Key output: a writing workflow where AI speeds up brainstorming, outlining, and research, but never drafts
Related
More data from the AI Lab.