Real cases · Marketing · First job
Ananya, 22: forty posts a month and none of them sound like anyone
A content calendar that eats her week, a tool that fills it in an hour, and engagement that has quietly halved.
Planned for release phase 2. The page is complete as a draft; interactive parts run in the browser and do not yet save to an account.
In brief
The situation. Ananya joined a Mumbai fintech as a content marketer eight months ago. The brief is forty social posts a month across three platforms. For the first three months she wrote them; then she found a tool that wrote them in an hour and started sleeping again. Engagement has halved since. Her manager has noticed. Ananya has noticed something worse: she can no longer tell which of the last hundred posts she wrote and which the tool did.
Step 1 — Define the End State
"A reader who follows us because the posts sound like a specific person who knows fintech and is a bit funny about it — not a brand account. Ten posts a month that get replies, not forty that get nothing. My manager can point at any post and I can say why it exists." Audience: one imagined reader, a 26-year-old saving for the first time, and her manager.
Step 2 — Map the Flow
Her list: define the voice, cut the volume, build a topic bank from real customer questions, draft, edit, schedule. AI review adds: look at the ten posts that performed best in the last year and find what they share — she had never done this; and a monthly interview with a customer-support colleague for real questions. It suggests "repurpose each post into five formats"; that is the volume trap again. Declined.
Step 3 — Assign Tasks
The voice document: hers, written by hand — it is the one thing the tool cannot know. Analysing the top ten posts: AI, verifiable. Turning support questions into a topic bank: AI generates, she chooses. First drafts: AI, against the voice document. Every final line: hers. Deciding which ten of the month's topics are worth a post: hers.
Step 4 — Select the Tool
A conversational tool with a project that holds the voice document and the top-ten analysis, because it is used weekly and context is everything. An image generator for the occasional visual, with a rule that anything showing a person is not generated.
Step 5 — Prepare and Direct
Criteria for a draft: contains one specific number or example a real reader could check; contains one sentence I would say out loud to a friend; contains no phrase from the banned list (unlock, seamless, journey, empower); could not be posted by a competitor without changing it. That last one is the whole job.
Steps 6 and 7 — Execute and Evaluate
Most first drafts fail the fourth criterion. She learns to see it in seconds — the tell is a post that could be any fintech's. The ones that pass are the ones built on a real support question. The banned list grows. Her editing time goes up; her drafting time goes down; the net is less than forty-posts-in-an-hour but far more than writing from scratch.
Step 8 — Integrate and Stress-Test
A month's ten posts, read together as the reader. Small fix: two posts in one week on the same theme. Structural: the posts are good individually and, read in sequence, have no throughline — a Step 1 problem; she had defined a voice but not what the reader should come to believe over a month. She adds one line to the end state and reorders.
What changed
Ten posts, replies on six. Her manager stops asking about volume. She can point at any post and say why. And she can, again, tell which ones she wrote — all of them, in the sense that matters.
Write the voice document by hand; it is the context only you have. "Could a competitor post this unchanged?" is the best single criterion for any marketing draft. Volume is the Dependent's metric.
A composite case, not a client engagement; the person is invented, the situation is not.