Hi Reader,

A prospective client asked me how much content a company needs before it can get found through Google and AI chat.

Most companies already have more useful content than they realize.

Last year, I wrote an email about this. I called it “Scrooge McDucking in content gold.” (I still love that title) 😆

This content lives in sales calls, project reviews, podcast interviews, presentations, team meetings, internal documents, and all the other places where experienced people explain how they do the work.

In that email, I shared a quick prompt for analyzing a transcript and finding content ideas inside it. That’s a great starting place.

Today I want to show you the larger workflow I use to create content at scale in order to increase AI citations.

This is the second issue in my series on getting found through AI chat. Last time, I explained how to choose a source your team can produce consistently.

This week, I’ll show you how I turn that source material into useful content. The three prompts below will give you a structured source note, three grounded article briefs, and a list of the questions you still need to answer before drafting. They give you a practical way to test one part of the knowledge-base workflow.

1/ Prepare the source material

A transcript is raw source material.

It contains the useful answer. It also contains small talk, repetition, unfinished thoughts, unsupported claims, confidential details, references to things on a shared screen, and ideas that made sense to the people on the call but will mean nothing to a reader.

I keep the original transcript and add enough context to make it useful later:

• The date

• The speakers and their roles

• Why the conversation happened

• The subjects covered

• Any confidentiality limits

• Links to related documents, recordings, or pages

Then I extract the useful knowledge into a standard format.

For me, a knowledge base is an organized collection of source material and structured notes that AI can search. Using one standard format lets AI compare conversations and find repeated questions, examples, and judgments.

What I build for clients often includes years of articles and newsletters, client conversations, project documents, presentations, and the context that explains the audience, offers, point of view, and editorial standards.

But the best place to start is to start small. Pick one transcript where a buyer asked good questions or an expert explained something clearly. Save the original transcript, then create the first structured note beside it.

2/ Structured knowledge notes

Each AI agent should have one narrow job. The first job is to turn each source into a structured knowledge note.

By agent, I mean an AI workflow with a defined job, the context and tools required for that job, and a clear handoff to whatever happens next.

The source note separates the buyer’s questions, the company’s judgment, the evidence behind it, and anything that still needs verification or a human answer.

Start a new AI chat and attach your company's Core Context. Then paste the prompt below with your transcript.

(If you need a starting point for that context, you can download my Core Context files.)

Copy/paste prompt 1: Create the source note

Analyze the transcript below as source material for our company knowledge base.

Use the attached Core Context to interpret the company's language, priorities, and audience.

Do not draft content or propose topics yet.

Create a structured source note with these sections:

1. Source context: date, speakers, roles, and why the conversation happened

2. Buyer questions: what the buyer asked and what decision they were trying to make

3. Expert judgments: recommendations, tradeoffs, corrections, and points of view

4. Firsthand evidence: stories, examples, results, and project details

5. Repeated patterns: problems or questions that may connect to other sources

6. Claims to verify: facts, numbers, and outside assertions that need research

7. Confidential material: names, commercial details, or examples that should stay private

8. Open gaps: questions a subject-matter expert still needs to answer

For every finding, include a timestamp, line reference, or short source excerpt so another person can return to the relevant part of the transcript.

TRANSCRIPT: [Paste transcript here]

A timestamp, line reference, or short excerpt lets the next agent return to the exact conversation and use the original source as evidence.

One note from the conversation behind this series could read:

Buyer question: How much content does the company need?

Point of view: Choose a format the team can sustain, publish the work on its own website, and use AI to increase the pace while people maintain the editorial standard.

Source: A link to the relevant transcript passage.

3/ Turn the knowledge base into topic briefs

In a full system, the next agent searches across the knowledge base for buyer questions that keep coming up, different explanations of the same tradeoff, and examples that support the company’s point of view. For this quick example, one source note is enough.

Ask it for three briefs, then use buyer research to decide which ideas deserve priority. I’ll cover that buyer-question map in the next issue.

Copy/paste prompt 2: Create three grounded article briefs

Review the structured source note below and propose three article briefs.

Each brief must include:

1. The specific reader

2. The question the article will answer

3. The decision it will help the reader make

4. Our point of view

5. The source evidence available, with a reference to the relevant section of the source note

6. What is still missing

7. Claims that require outside research or citations

8. Confidential details that must stay out of the article

9. The practical takeaway the reader should be able to use

Reject any brief that depends mainly on general AI knowledge. Every proposed article must have a meaningful point of view or firsthand evidence from the source note.

SOURCE NOTE: [Paste the source note here]

A person who understands the buyer and the company chooses which brief is worth developing.

4/ Fill the gaps in the knowledge base

When you review the brief, you can see what information is missing before anyone drafts the article.

• When the company’s point of view is missing, the subject-matter expert needs to fill it in.

• A missing example may require another search across calls, projects, or presentations.

• A research agent verifies outside factual claims and collects reliable citations.

• You can clarify a vague explanation with a few follow-up questions by email or a short voice-agent interview.

• A qualified human reviewer checks sensitive legal, financial, medical, safety, regulatory, or technical claims.

Add the expert answers and verified research back to the knowledge base with their sources, so it becomes a record of what the company knows and can support.

Copy/paste prompt 3: Turn the missing information into a plan

Review the article brief below and focus only on the “What is still missing” and “Claims that require outside research or citations” sections.

Create two lists:

1. Expert questions: Write the shortest set of specific questions I need to ask a subject-matter expert. For each question, explain which part of the article the answer will support.

2. Research tasks: List each outside claim that needs verification, the kind of source required, and what would count as adequate evidence.

Put the questions in the order I should ask them during a 15-minute interview. Do not invent answers or sources.

ARTICLE BRIEF: [Paste the selected brief here]

This third prompt gives you an interview guide and a research checklist. When you collect the answers, add them to the source note with the speaker, date, and source attached. Then update the brief before anyone starts drafting.

5/ The agent team and human review

You can run these three steps manually for one transcript. Repeating the same review across dozens of calls, briefs, claims, and source links is where a series of focused agents becomes useful.

The real magic begins when you have a pipeline of agents collaborating together. You can then do this at scale.

As I mentioned, each agent should have a narrow job:

1. The source-note agent turns each transcript into a structured record of the buyer's questions, the company's judgment, the evidence, and the open gaps.

2. The brief agent assembles the reader, question, point of view, sources, missing information, and review requirements.

3. The research agent verifies outside claims and collects reliable citations.

4. The drafting agent develops the article from the approved brief and the sources attached to it.

5. The fact-checking agent compares the draft against the sources and flags anything unsupported or inconsistent.

6. The page agent checks the title, headings, structure, internal links, and the technical details that help people and machines use the page.

Some of my pipelines have as many as 12 agents with multiple revision loops and pass/fail gates. A human reviewer still decides what the company is willing to publish and checks judgment, risk, confidentiality, and any regulated or claim-sensitive material. Agents handle the repeated work, while people remain responsible for the final piece.

Your Turn

Take a few minutes and try the workflow with one transcript.

1. Choose a recent sales call, client meeting, or expert interview. Run prompt 1 and save the response as a source note.

2. Start a new AI chat with that source note attached. Run prompt 2 and save the three briefs.

3. Choose the brief with the strongest point of view and source support. Run prompt 3 to create the expert questions and research checklist.

Stop before drafting. Look at what you have and ask three questions:

• Does this article answer a question a buyer has asked?

• Does it contain a point of view or firsthand evidence that belongs to your company?

• Do you know exactly what you still need from an expert or outside source?

If you can answer yes to all three, you have enough to develop the article. You should now have one saved knowledge-base entry, three source-linked briefs, and a specific plan for filling the gaps in the strongest one.

In the next issue, I’ll show you how to use buyer questions to choose which topics deserve priority and how much content your company may need.

- Dan

P.S. If you want help building an agentic content system that gets you found in search, book an AI strategy call. We’ll look at your source material, review requirements, and where the first agent should begin.