12-step production pipeline · 6 specialized AI agents

AI white papers that survive
executive scrutiny

A structured AI pipeline that enforces argument integrity, evidence density, and style consistency from concept to final draft. No hallucinations. No opinion dressed as fact.

Built for B2B marketers, agencies, and content teams that need white papers to actually perform — not just exist.

The problem with AI white papers today

Every B2B team is trying to use AI for white papers. Most are producing the same thing: long, generic, uncited content that no buyer trusts.

Problem
Hallucinated statistics
How we fix it
Research-Analyst agent never fabricates. Every unsupported claim tagged [DATA NEEDED].
Problem
No argument structure
How we fix it
Argument Lock before writing begins. Core claim, evidence sources, and CTA locked — agents can't deviate.
Problem
Semantic paraphrasing
How we fix it
Semantic repetition scanner detects paragraphs that say the same thing in different words. Cut, shorten, or rephrase before final.

The 12-step production pipeline

Every white paper moves through four phases. You approve each phase before the next begins. No shortcuts.

Concept
1.Define argument & audience
2.Master outline
3.Phase approval
Planning
4.Detailed outline
5.Evidence plan
6.Phase approval
Drafting
7.Write 7 sections
8.Fill data gaps
9.Phase approval
Refactoring
10.Structural + clarity pass
11.Semantic repetition scan
12.Final + library update

Not “press go and get a paper”

The pipeline has 3 mandatory approval gates. You shape the argument and structure before any prose is written. Nothing advances without your sign-off.

Step 3
Approve master outline

Review section headers and one-sentence descriptions — the skeleton of the argument. Edit or reject before anything gets expanded.

Step 6
Approve detailed outline

Review 3–7 bullets per section plus the evidence plan. Every data gap is flagged. You confirm the structure before a single word of prose is written.

Step 9
Approve full draft

Review the complete draft before refactoring begins. Send sections back, request changes, or approve to move into the editing and quality phases.

You bring the research. You approve the structure. You control what gets published.

What the pipeline enforces

Not prompts. Not suggestions. Hard rules that agents check before every output.

Argument Lock

Lock your core argument, evidence sources, and CTA before writing begins. Every agent checks the lock — no scope drift, no contradictions between sections.

6 Specialized Agents

Outline-Architect, Research-Analyst, Section-Writer, Dev-Editor, Beta-Reader, Section-Refiner. Each with a distinct role and strict constraints. No generalist prompting.

Evidence = 32%

Every claim needs support or gets tagged [DATA NEEDED]. The evidence plan tracks source quality: peer-reviewed > analysts > government data > case studies > industry reports.

Style Learning

Upload a past white paper and the system extracts your style fingerprint: tone, sentence rhythm, citation style, header format, vocabulary. Every paper matches your brand.

Evidence Library

Every completed paper adds searchable evidence to your library. Reuse findings, statistics, and case studies. Your 5th paper on a topic benefits from 4 papers of curated evidence.

Semantic Repetition Scanner

Vectorized paragraph comparison. Detects LLM paraphrasing — paragraphs that say the same thing in different words. Tiered: cut, shorten, or rephrase before final.

Beta Reader Agent

Reads the draft as a skeptical member of your target audience. Flags credibility gaps, abandon points, missing objections, and weak CTAs — before you publish.

Section Draft Order

Problem → Evidence → Solution → Takeaways → Background → Conclusion → Executive Summary (last). The order that produces argument coherence, not narrative convenience.

Phase-Gated Approvals

You review and approve at the end of each phase. No accidental final drafts built on shaky outlines. No pipeline advancing without your sign-off.

6 agents. Each with one job.

Specialized agents outperform generalist prompting. Each agent has a defined role, strict constraints, and knows exactly when it runs in the pipeline.

Outline-Architect
Steps 2, 4
Argument structure, section weight, master and detailed outlines
Research-Analyst
Steps 5, 8, 11
Evidence mining, fact verification, data gap filling — never fabricates
Section-Writer
Step 7
Drafts each section from the detailed outline, tagged [DATA NEEDED] for gaps
Dev-Editor
Steps 10, 11
Three-pass review: structural integrity, clarity, engagement
Beta-Reader
Post Step 9, Post Step 12
Reads as a skeptical target reader — flags credibility gaps and abandon points
Section-Refiner
Steps 10+, defluff
Applies targeted fixes from review findings and defluff results

One price. No subscription.

Buy once. Bring your own Anthropic API key. A finished white paper costs $1–2 in API fees — paid directly to Anthropic, not us.

Founding member
White Paper System
Mac & Windows · Lifetime licence
$37
$67
Get it on Gumroad

Founding member price ends at 50 sales. After that: $67.

Built for teams that publish, not experiment

White Paper System is not a general writing assistant. It is a production system for one specific content format.

B2B Marketing Teams

Produce thought leadership white papers at scale without losing brand voice or argument quality across team members.

Marketing Agencies

Deliver client white papers on deadline with consistent quality. The style learning system adapts to each client's voice.

SaaS & Tech Companies

Build the evidence-backed technical white papers and market research reports that enterprise buyers expect before purchase decisions.

Consulting Firms

Turn proprietary research and methodology into white papers that position your firm as the category authority — with your evidence, your argument, your voice.

Frequently asked questions

Everything B2B teams ask before committing to an AI white paper tool.

What is an AI white paper generator?

An AI white paper generator uses artificial intelligence to plan, research, outline, draft, and refine white papers. White Paper System goes further than basic AI tools — a structured 12-step pipeline with 6 specialized agents enforces argument integrity and evidence density so the result survives scrutiny.

How is this different from ChatGPT for white papers?

ChatGPT produces unstructured output with no evidence enforcement, no argument lock, and no quality gates. White Paper System locks your argument before writing begins, tags every unsupported claim [DATA NEEDED], runs a semantic repetition scanner to remove paraphrasing, and requires phase approvals. The output is a white paper — not a long blog post.

Will the AI hallucinate statistics?

No. The Research-Analyst agent is explicitly instructed never to fabricate statistics. Any claim without a verifiable source gets tagged [DATA NEEDED] for you to fill before publication. The Argument Lock also specifies required and forbidden claims upfront.

Can it learn my company's writing style?

Yes. Upload past white papers (PDF, DOCX, or Markdown) and the system extracts a style fingerprint: tone, reading level, citation style, header formatting, sentence length, preferred vocabulary, and forbidden terms. Every agent uses it. Your 10th paper sounds like your first.

How long does it take to produce a white paper?

A 3,000–5,000 word white paper takes 2–4 hours of guided pipeline work, including your review and approval at each phase gate. Pure AI generation per section takes minutes.

What is an Argument Lock?

A per-paper configuration that locks the core argument, audience, word count target, CTA, required claims, and forbidden claims before any writing begins. Every agent checks the lock — if a claim is forbidden, it doesn't get written. This prevents scope drift and contradictions across sections.

What is the semantic repetition scanner?

It embeds every paragraph using sentence-transformers and computes cosine similarity across the draft. Paragraphs that say the same thing in different words (similarity ≥ 0.80) are flagged and tiered: T1 (≥0.90) cut, T2 (0.85–0.89) shortened, T3 (0.80–0.84) rephrased. LLMs paraphrase — this catches it.

Does it search the internet for research?

No. The app works entirely from the source material you provide — PDFs, reports, data files, interview notes, competitor papers, anything you drop into the project's research folder. It doesn't scrape the web or pull live data. Your content stays on your machine.

Who is this built for?

B2B content marketers, marketing agencies, SaaS companies, consulting firms, and professional services teams that produce white papers regularly. Not a blog post tool — the pipeline is designed specifically for research-heavy, argument-driven white paper format.

Is this a 'press go and get a paper' tool?

No — and that's by design. The pipeline has 3 mandatory approval gates: after the master outline (step 3), after the detailed outline and evidence plan (step 6), and after the full draft (step 9). Nothing advances without your sign-off. You shape the argument and structure before any prose is written.

What AI model powers it?

Claude Opus 4.6 by Anthropic for all writing, editing, and analysis. Opus is optimized for long-form reasoning and document production — the right model for sustained argument construction.

How much does it cost to run?

The app is a one-time purchase ($37 founding member, $67 standard). You pay Anthropic separately for API usage — typically $1–2 per finished white paper. No subscription, no markup, no metering on our end.

Your next white paper. Done right.

Download the app. Bring your own Anthropic API key. Run the full 12-step pipeline locally — your data never leaves your machine.

Download the app

Mac & Windows. Bring your own Anthropic API key.