Skip to main content

​
The Operator Prompt

​
A System for Content That Compounds

Most AI-generated content resets. It is produced for this week’s campaign, consumed or ignored, and then next week’s prompt starts from zero. The context disappears. The insights disappear. The institutional knowledge that should be compounding is instead evaporating at the end of every conversation. This guide is a system for the opposite. The Operator Prompt is a framework for DTC brand operators who want content creation to compound — where each piece builds on the intelligence from the last, where the AI has more context over time rather than less, and where the output of creative work feeds back into the customer intelligence that drives the next creative brief.

​
Part 1: The Context Architecture

The gap between generic AI output and genuinely useful output is context. The gap between useful output and compounding output is persistent context — context that grows over time rather than resetting with each conversation. Build five living documents. These are not documents you create once and file. They are documents you maintain and feed into every AI content session.

​
Document 1: Brand DNA

Everything a content creator needs to know about the brand in a single document that can be uploaded to any AI session. Contents:
  • Brand voice principles (3–5 rules, written as constraints — what the brand never says, not just what it says)
  • ICP definition — specific, behavioral, not demographic. Not “women 35–55” but “CMOs at $25M brands who have been burned by agencies that prioritized vanity metrics”
  • Key messaging pillars — the 3–4 ideas the brand owns and returns to consistently
  • Vocabulary list — the specific words the brand uses and avoids
  • Tone examples — 2–3 samples of writing that represents the brand voice at its best
The test: If someone unfamiliar with the brand reads this document and then writes a piece of content, would it sound like the brand? If not, add more specificity.

​
Document 2: Customer Intelligence Brief

The rolling weekly analysis of the customer file. This is the document built from the Claude agent workflow in the agents playbook — the Monday morning intelligence brief that accumulates over time. Contents:
  • Current RFM distribution (Champion segment size and revenue share)
  • Most recent cohort LTV findings
  • Current at-risk customer volume and revenue at stake
  • Key behavioral insights from the last 30 days
  • Any anomalies or trend shifts
Why it matters for content: When the content team briefs creative, they should know which segment they are writing for. The Customer Intelligence Brief tells them who the Champion is, what the at-risk customer looks like, and where the file is growing or shrinking. Content that is not segment-aware is not intelligence-driven content.

​
Document 3: Customer Voice Library

The living repository of verbatim customer language. Built from review mining (see Agent 13 in the Claude Agents Playbook), support ticket analysis, post-purchase survey responses, and community engagement. Contents:
  • Top 15 phrases customers use to describe product transformation
  • Top 10 objections customers voice before purchasing
  • Top 5 pain points the product resolved (verbatim)
  • Language customers use when they are disappointed vs. delighted
  • Any phrases that appear 3+ times verbatim across reviews
The rule: At least 3 verbatim customer phrases must appear in every creative brief. Not paraphrased. Verbatim. This is the single most reliable way to make creative resonate with the audience that matters.

​
Document 4: Competitive Landscape

Updated weekly (see Agent 17 in the Claude Agents Playbook). What competitors are saying, where they are weak, what customer needs they are not meeting. The content application: Every campaign brief should include one line: “What this says that competitors can’t say.” If it doesn’t exist, the campaign is producing parity content. Parity content is expensive slop.

​
Document 5: Performance Archive

What has worked and what has not, written in enough detail to be useful. Not just metrics — context. Contents:
  • Best-performing campaigns with context (what was the segment, the offer, the format, the timing)
  • Worst-performing campaigns with diagnosis (why it did not work — wrong segment, wrong message, wrong timing, wrong platform)
  • Subject lines ranked by open rate and revenue per recipient
  • Ad creative ranked by ROAS and 90-day cohort LTV (not the same list)
The compounding effect: When a content creator has access to a Performance Archive with 18 months of annotated history, they are not starting from zero. They are starting from the accumulated knowledge of everything the brand has tried and learned.

​
Part 2: The Brief-First Workflow

Most content creation starts with the output. Write the email. Write the ad. Write the post. Then figure out what it was supposed to accomplish. The Operator Prompt workflow starts with the brief. The brief is the intelligence layer. The output is downstream.

​
The Brief Template

Every content piece — email, ad, social post, landing page — should be preceded by a brief that answers these questions: Who is this for? (RFM segment, not demographic) What do they already believe? (What is their current understanding of the product or offer — informed by the Customer Voice Library) What do we want them to do? (One action. Not “engage with the content” — a specific action with a specific outcome.) What is the one insight this piece of content will deliver? (The thing they did not know before that makes the action logical. If there is no insight, there is no reason to produce the content.) What customer language must appear verbatim? (3+ phrases from the Customer Voice Library) What is this content’s relationship to the customer file? (Is this a Champion retention piece? A second-purchase driver for Recent customers? A winback for At Risk? If the answer is “general audience,” reconsider whether to produce it.)

​
Part 3: The Prompting Framework

The difference between prompts that produce usable content and prompts that produce generic output is specificity of context. The template: 
You are writing [content type] for [RFM segment].

Context documents I'm providing:
- Brand DNA (uploaded)
- Customer Voice Library (uploaded)
- Campaign brief (below)

Campaign brief:
- Objective: [specific action]
- Segment: [Champion / Loyal / Recent / At Risk]
- Insight to deliver: [one sentence]
- Customer phrases to include verbatim: [3 phrases from Customer Voice Library]
- Offer or CTA: [specific]
- Platform/format: [email, Meta feed, SMS, etc.]
- Tone constraint: [reference Brand DNA voice rules]

Do not use:
- Generic DTC opening lines ("We noticed you haven't purchased recently")
- Discount-leading copy for Champion segments
- Industry jargon not found in the Customer Voice Library
- Any phrase that could appear in a competitor's content

Output: [specific format — email with subject line and preview text, or 3 ad headlines with primary text]
The specificity of the brief determines the quality of the output. Garbage in, garbage out is true for AI as it is for every other creative process. The Operator Prompt is a system for ensuring the context going into the AI session is never garbage.

​
Part 4: The Feedback Loop

Content that compounds has a feedback loop built in. After every campaign: What went into the Performance Archive:
  • What the brief was (not just the creative)
  • Which segment received it
  • The specific customer phrases that were used
  • What the open rate, click rate, and revenue per recipient were
  • What the 30-day repeat purchase rate was for customers who received it
  • The diagnosis: why did it perform or underperform?
What gets updated in the Customer Voice Library:
  • Any new phrases from replies, replies, or support conversations that the campaign triggered
  • Any objections that surfaced in customer service after the campaign
  • Any language that generated outsized engagement
Over 12 months, this feedback loop produces an institutional knowledge base that no competitor without the same system can replicate. That is the compounding. Not more content — smarter context.

​
Part 5: The Anti-Patterns

Five content behaviors that signal a team is producing volume rather than intelligence: The generic win-back. “We miss you!” sent to every lapsed customer regardless of value, purchase history, or behavioral segment. This is not a retention strategy. It is a list management activity. The discount-first Champion email. Sending a 20% off code to a Champion-segment customer who would have purchased without it. This is not generosity. It is margin destruction. The campaign that forgets the brief. Content that is well-crafted but has no clear connection to a specific behavioral segment or business outcome. Beautiful slop. The A/B test without a hypothesis. “Let’s test two subject lines and see which performs better.” Without a specific hypothesis about why one would outperform the other — and what that result would tell us — the test produces data without intelligence. The reset. Starting each campaign from zero without referencing the Performance Archive, the Customer Voice Library, or the Customer Intelligence Brief. Each piece of content should be informed by everything that came before it.

​
Part 6: The Minimum Viable System

For teams that do not have the capacity to implement the full framework immediately, the minimum viable version:
  1. One document: Build the Customer Voice Library first. Extract verbatim phrases from reviews. Require 3 per brief. This change alone will produce measurably different creative output within 30 days.
  2. One rule: Every email goes to a segment, not a list. Define the segment before you write the email.
  3. One metric: Track revenue per recipient by segment, not open rate. The segment that receives the right message will show it in revenue per recipient first.
Build from there.
Click Open in Claude above to start building your Brand DNA document and Customer Voice Library with Claude’s help. Upload your last 100 customer reviews and Claude will extract the verbatim phrase library that drives the rest of the system.