$65M+ Revenue InfluencedAI Training for TeamsAI TransformationMarketing / Sales / HR / Finance6+ Years Building Growth Systems125+ Projects Delivered$65M+ Revenue InfluencedAI Training for TeamsAI TransformationMarketing / Sales / HR / Finance6+ Years Building Growth Systems125+ Projects Delivered$65M+ Revenue InfluencedAI Training for TeamsAI TransformationMarketing / Sales / HR / Finance6+ Years Building Growth Systems125+ Projects Delivered$65M+ Revenue InfluencedAI Training for TeamsAI TransformationMarketing / Sales / HR / Finance6+ Years Building Growth Systems125+ Projects Delivered$65M+ Revenue InfluencedAI Training for TeamsAI TransformationMarketing / Sales / HR / Finance6+ Years Building Growth Systems125+ Projects Delivered$65M+ Revenue InfluencedAI Training for TeamsAI TransformationMarketing / Sales / HR / Finance6+ Years Building Growth Systems125+ Projects Delivered$65M+ Revenue InfluencedAI Training for TeamsAI TransformationMarketing / Sales / HR / Finance6+ Years Building Growth Systems125+ Projects Delivered$65M+ Revenue InfluencedAI Training for TeamsAI TransformationMarketing / Sales / HR / Finance6+ Years Building Growth Systems125+ Projects Delivered
All 20 AI systems

Open-source playbook · 05 of 20 · For marketing

Content Engine

Your best ideas stay in meetings and voice notes because nobody has time to write them up.

The systemTurns one talk, call or voice note into posts, a newsletter and short video scripts in your voice.

// The problem

Good ideas, stuck in recordings.

The founder said something sharp on a customer call. A product lead explained the product better than the website does. It’s all recorded, and none of it gets written up, because writing it up is a half-day nobody has.

  • Talk at a customer event22 min · 6 weeks ago

    0pieces written
    from it so far

  • All-hands Q&A35 min · 5 weeks ago

    0pieces written
    from it so far

  • Customer call, Leeds depot48 min · 3 weeks ago

    0pieces written
    from it so far

  • Voice note on the walk back14 min · 2 days ago

    0pieces written
    from it so far

Sample recordings, fictional company. The last one, a 14-minute voice note, is the one we follow below.

// How it works

One recording. A dozen pieces.

The system finds the ideas worth sharing, writes each one up in every format you use, checks it against your voice and hands the batch to a person.

1The source

Voice note · 14 min

Dana Osei, Founder, Quillstone (fictional) · recorded on the walk back from a customer visit

2The ideas it finds

  1. 1Field teams don’t read dashboards. They read texts.
  2. 2The best request this year was “remove a step.”
  3. 3We stopped selling to owners and started selling to dispatchers.

3Written up in every format

  • 3

    Posts

    One per idea, each with its own opening line.

    “Field teams don’t read dashboards.”

  • 1

    Newsletter

    All three ideas as one issue.

    Subject: What a ride-along taught me

  • 2

    Video scripts

    30–45 seconds, with a hook and shot notes.

    0–3s: “Your field team has never opened your dashboard.”

  • 1

    Article draft

    The long version, for your blog.

    “Remove a step”: our best feature request

  • 5

    Quote lines

    Short lines for slides and graphics.

    “We stopped selling to owners.”

4Voice check, on every piece

  • Words Dana never uses
  • How she opens and closes
  • Sentence length and rhythm
  • Every claim traced to the recording
5

Dana approves, the team publishes. She approves, edits or drops each piece. The team schedules what she approves. Nothing posts on its own.

You sign off

// What comes out

One voice note. Five finished formats.

Sample outputs from one 14-minute voice note. The speaker and company are fictional. Each piece is written for its format, not cut and pasted between them.

Post 1 of 3

DODana OseiFounder, Quillstone

Field teams don’t read dashboards.

I rode along with a dispatcher on Tuesday. Forty jobs. Not one login. Everything that mattered arrived as a text.

We’d spent a year making our dashboard better. For the people doing the work, the best version of it was a message that says “next job, 2:40, gate code 4471”.

So that’s what we’re building now.

DraftWaiting for Dana’s approval

Quote line 2 of 5

“The best feature request we got this year was ‘remove a step.’”
Dana Osei · Founder, Quillstone

Newsletter

Quillstone Notes

What a ride-along taught me

Three things I learned in one day at a depot, and what we’re changing because of them.

1. Nobody opened the dashboard

Forty jobs went out on Tuesday. Every one of them ran on texts.

2. “Remove a step” beats any new feature

The most useful request we heard all year asked us to take something away.

3. The dispatcher decides

Owners sign the contract. Dispatchers decide whether it gets used.

Reply and tell me which step you’d cut. I read every one. Dana

Video script 1 of 2 · 45 seconds

TimeOn screenDana says

0–3sDana to camera, depot behind her“Your field team has never opened your dashboard.”
3–15sPhone screen: a text from dispatch“I rode along with a dispatcher on Tuesday. Forty jobs, zero logins. Everything ran on texts.”
15–35sWalking through the yard“So we stopped building screens for them, and started sending the one thing they need, when they need it.”
35–45sBack to camera“What does your team actually look at all day?”

Cut from the voice note’s 3:10–4:05, in Dana’s own words

// Why it sounds like you

It checks the voice before you read a word.

Your voice guide and your best past pieces live in the Company Brain. Every draft is checked against them, and anything that sounds like a press release, or claims something the recording doesn’t, is fixed or flagged first.

  1. DraftWe’re thrilled to announce an exciting new chapter for our customers!

    Dana never says “thrilled” or “exciting”.

    AfterWe changed one thing this month: dispatchers now get a text, not a dashboard.

  2. DraftTeams using Quillstone cut their admin time by 40%.

    That number isn’t in the recording.

    AfterNumber removed. Ask Dana for the real figure, or leave it out.

  3. DraftExciting times ahead for everyone in field service!

    Dana ends on a question, not a cheer.

    AfterWhat’s the one step your team would remove tomorrow?

Review boardVoice note, 14 min

  • Post 1: dashboards vs textsApproved
  • Post 2: “remove a step”Approved
  • Post 3: selling to dispatchersEdit asked
  • Newsletter issueApproved
  • Video script 1Approved
  • Video script 2Dropped
  • Article draftEdit asked

Nothing is scheduled or published until a person approves it.

// For your tech lead

Can we build it? Here’s what it takes.

Most of the work is writing down your voice: the guide, the words you never use and ten pieces you’re proud of. The pipeline around it is a small build.

Tools you’ll use

JobUseOr
Where recordings landA Google Drive or OneDrive foldera Notion page, Zoom or Meet recordings
TranscriptsWhisper (OpenAI)Deepgram, AssemblyAI, your meeting tool’s transcript
Finding ideas and writingClaudeChatGPT, Gemini
Voice guide and past piecesCompany Braina Google Doc or Notion page
Running the stepsn8nMake, Zapier
Review and schedulingA Notion or Google Docs review boardBuffer, Typefully, your newsletter tool’s drafts

Access it needs

  • Read access to one recordings folder, and nothing else
  • Your voice guide and a set of approved past pieces
  • Permission to create drafts, not to publish

Three ways to build it

  1. Start simple1 day

    A saved Claude or ChatGPT project with your voice guide and ten of your best pieces. Paste in a transcript and get the set of drafts.

  2. Production1–2 weeks

    n8n watches a folder: new recording, transcript, ideas, drafts in each format, voice check, then a review board where a person approves.

  3. Custom3–4 weeks

    Adds templates per channel, clip suggestions with timestamps from video, and a loop that learns which ideas and formats did well.

Guardrails

  • Nothing publishes on its own; a person approves every piece
  • Claims, numbers and names must come from the recording or an approved list
  • Never invents quotes, customers or stories
  • Customer calls are used only with the speaker’s okay, and customer names stay out unless approved

What to measure

  • Pieces published per recording
  • Time from recording to first piece published
  • Share of drafts approved with light edits
  • How each format performs, so the next batch leans on what works

Who builds and runs it

A content or marketing lead owns it, with someone comfortable in n8n for the first one to two weeks. After that, the speaker spends a few minutes approving each batch, and the content lead edits and schedules.

// Want help building it?

We’ll build your Content Engine with your team.

In 30 minutes with our founder, we’ll look at your tools and your week, and work out what your first AI systems should be and how long they’ll take.

Book a Strategy Call
Pallav KaushishFree 30-minute call with our founder · No commitmentPallav Kaushish Co-Founder, AIxGrowth