My AI content workflow as a solo marketer is built around three stages, research and outlining with ChatGPT/Claude, drafting in my own voice with heavy manual editing, and repurposing one core piece across formats, with strategy, brand voice, and client relationships kept entirely out of AI's hands. This isn't a "10 best AI tools" roundup; it's the actual week-by-week process I use running organic growth for edtech and startup clients. Below is exactly how it works, including where it breaks if you're not careful.
Every "AI marketing workflow" post I read is either an affiliate-driven tool list or so vague it's useless ("use AI to save time!"). I run content and strategy for multiple clients solo, so I don't have the luxury of a team to catch bad output, here's what actually happens on my end, week to week.
Monday: Research and Outlining (Where AI Earns Its Keep the Most)
I start the week by feeding Claude or ChatGPT a rough brief: the topic, the target keyword if it's SEO content, competitor content I've already found manually, and 2-3 bullet points on the angle I want to take. I never ask it to "write a blog post about X" cold, that's where generic, hollow output comes from.
What I actually ask for at this stage:
- An outline with H2/H3 structure based on real search intent, which I then heavily reorder based on my own read of what the audience actually needs.
- A list of counterarguments or gaps in the obvious take, which is genuinely useful for finding an angle competitors haven't covered.
- A first pass at data points or examples I should verify, I treat every factual claim AI gives me as unverified until I check it myself, because hallucinated statistics are still a real and common failure mode.
This stage takes about 20-30 minutes per piece and saves me roughly an hour of blank-page staring. That's the actual value, not "AI writes my content," but "AI removes the friction of starting."
Tuesday-Wednesday: Drafting (Where I Do the Heavy Lifting)
This is the stage most people skip past too fast. I don't ask AI to write a full draft in one shot, I write the first draft myself, using the outline as scaffolding, because voice is the hardest thing to get an AI to fake convincingly and it degrades fastest the more the tool is doing the actual composing.
Where I do use AI during drafting:
- Getting unstuck on a specific paragraph. If I know what a section needs to say but can't find the sentence, I'll ask for 3-4 versions of just that paragraph, then rewrite the best one in my own words rather than copy-pasting it directly.
- Checking logical structure. I'll paste a rough draft and ask "does this argument actually flow, or am I skipping a step?", this catches gaps in reasoning better than most self-editing does.
- Generating counter-examples to stress-test a claim before I publish something opinionated, so I'm not caught flat-footed in the comments.
What I never do: paste a full AI-generated draft and lightly edit it. It reads flat every time, no matter how good the model is, because it's missing the specific, slightly imperfect detail that comes from actually having done the work being described.
Wednesday-Thursday: Editing for Voice (The Step That Actually Matters Most)
This is where most "AI-assisted content" fails, and it's almost always a voice problem, not a facts problem. AI defaults to a particular cadence, heavy on "Moreover," "In today's fast-paced world," three-item lists everywhere, and a kind of relentless positivity that doesn't sound like an actual person with opinions.
My editing pass specifically hunts for:
- Any sentence that could apply to literally any brand. If a sentence about "authentic engagement" could be pasted into a competitor's blog with zero changes, it gets rewritten with something specific, a real number, a real example, a real client name where relevant.
- Removing the AI tell-tale rhythm. AI text tends toward evenly-paced sentences of similar length. I'll deliberately break that up, short sentence. Then a longer one that carries more of the actual argument and detail.
- Adding first-person specificity. "I've seen this work with edtech clients" reads completely differently from "this tends to work well for brands," even though they're saying almost the same thing.
- Cutting the hedge words. AI drafts lean on "can," "may," "often," "tends to", safe, noncommittal language. I replace as many as possible with direct claims, because that's closer to how I'd actually say it out loud, and it's more useful to the reader.
Friday: Repurposing One Piece Across Formats
Once a core piece (usually a blog post) is finalized, I repurpose it rather than starting from scratch for every channel, but each version still gets a manual voice pass, not a copy-paste.
Blog → LinkedIn post: I pull the single sharpest insight or contrarian point from the blog, not a summary of the whole thing. A LinkedIn post that tries to cover everything the blog covers reads like a press release; one built around a single sharp claim gets read and shared.
Blog → Instagram caption: I take the same core insight and rewrite the first line as a hook specific to Instagram's scroll-stopping format (a question, a stat, or a direct claim), then compress the supporting point into 2-3 short sentences instead of the blog's full explanation. I ask AI for 4-5 hook variations here, which is one of the genuinely strong AI use cases, fast divergent options for a single line, where I'm the one picking the best.
Blog → Reel script: I outline the talking points with AI's help (it's useful for pacing, "what's the fastest way to say this in 30 seconds"), but I write and deliver the actual script myself, because tone of voice in spoken content is even less fakeable than written tone.
This repurposing process takes about 45-60 minutes for three formats once the source piece is done, versus the 2-3 hours it would take starting each one from a blank page.
Where AI Should Never Make the Call
- Strategy. Which platform to prioritize, what a client's actual growth bottleneck is, whether to pivot content pillars, these require context AI doesn't have (client history, market nuance, what's actually happened in past campaigns) and shouldn't be outsourced to a model that's pattern-matching on generic best practices.
- Voice, at the final stage. AI can help find a voice-adjacent draft, but the decision of what "sounds like me" is a judgment call I make, every time, on every piece.
- Client relationships. I don't use AI to draft client emails about sensitive topics (missed deadlines, results conversations, pricing pushback), those require reading the actual relationship and history with that person, not a templated tone.
- Anything requiring a real, specific number or claim about a client's results. I've caught AI confidently inventing plausible-sounding statistics more than once. Every data point in client-facing or published work gets manually verified against the source.
The Honest Time Trade-Off
This workflow doesn't save as much time as the "10x your content output" claims suggest, for me it's closer to 30-40% faster on research and repurposing, and roughly the same amount of time on drafting and voice editing, because that stage needs to stay manual to not sound generic. The real win isn't speed on every stage, it's removing the two most energy-draining parts (blank-page starting and repurposing fatigue) so I can spend more actual thinking time on strategy and voice, which are the parts that were never going to be automatable anyway.
FAQ
Can AI completely write a blog post that doesn't sound like AI? Not reliably on its own, a manual editing pass focused specifically on voice, specificity, and removing hedge language is what actually removes the "AI-sounding" quality, not a better prompt alone.
What's the best AI use case for a solo marketer with limited time? Research, outlining, and repurposing existing content into multiple formats, these are high-friction, low-judgment tasks where AI genuinely saves time without risking voice or strategy quality.
Should I tell clients I use AI in my content workflow? Being transparent about using AI for research and drafting support (while being clear that strategy, voice, and final decisions are yours) tends to build more trust than it costs, especially with clients who already assume everyone's using it in some form.
How do I stop my content from sounding generic when I use AI for drafting help? Add specific numbers, named examples, and first-person detail that AI can't generate on its own since it doesn't have your actual experience, that's usually the fastest way to break the generic pattern.
Is it risky to use AI for client-facing statistics or case study numbers? Yes, treat every AI-generated data point as unverified until you check it against your actual source, since models can generate confident, plausible-sounding numbers that aren't accurate.
Where to Go From Here
This is the actual workflow behind the content and strategy work I do for clients like Masai School and Sparkling Sandset, AI where it removes friction, judgment everywhere else. If you're a solo marketer or small team figuring out where AI actually fits into your process, take a look at the case studies or get in touch and I'm happy to talk through what's working.