Module 1.2 · Anatomy of the Factory

📦 The 19 Deliverables

Two text boxes go in. Nineteen files come out. The math is simple: 15 Markdown documents + 2 PPTX presentations + 2 Word documents. Each piece is a deliverable a consulting firm would charge a lot for—and your Factory generates the whole set for less than US$1.

📋 Assessment

•Technology and data infrastructure inventory
•Department pain point matrix
•AI maturity and readiness assessment (1-5)

🗓️ Planning

•30/60/90/180/360-day roadmap
•Quick wins list
•ROI calculator and cost analysis

🛠️ Implementation

•Vendor comparison · build vs. buy
•Software license consolidation
•Library of department-specific use cases

🛡️ Governance + resources

•AI policy and acceptable use · data governance
•Change management and training manual
•Prompt library + glossary of terms

🎁 The 4 polished deliverables

The 15 Markdown files are the raw knowledge. On top of them, the Factory generates the 4 files the client actually opens in the meeting:

•PPTX — Strategic AI presentation (board deck)
•PPTX — Current-state vs. future-state diagrams (Mermaid rendered as an image)
•DOCX — Final AI strategy report
•DOCX — Statement of work (SOW), with scope and proposal

15 Markdown

internal knowledge

2 PPTX

executive decks

2 DOCX

report + SOW

= 19

complete package

🔗 The 3-phase pipeline

The 19 files don't appear all at once by magic. They go through a three-phase assembly line: Research → Synthesis → Generation. Each phase has a dedicated tool and delivers its output to the next—exactly like a factory assembly line.

💡 Why split into phases

Each phase uses the right tool for the right job: Perplexity finds current facts, Gemini writes affordably with a huge context window, and Python formats. Separating responsibilities keeps the pipeline predictable—and shows you exactly where to step in when you want to improve a part.

Research

Perplexity

Synthesis

Gemini

Generation

Local Python

Chained

output → input

🔍 What Each Phase Does

You won’t use the Factory as a black box. Understanding each phase’s work keeps you in control — you’ll know where the quality comes from and where you can improve it.

Research — gathers current facts

Perplexity runs 9 to 18 live searches about the company: business vision, technology stack, competition, AI initiatives, pain points, and priorities. It doesn’t make things up—it brings recent data and sources. Output: a research JSON.

Synthesis — turns facts into documents

Gemini 2.5 Flash receives the facts + consulting frameworks and runs 15 specialized prompts, one for each deliverable. It’s the writing engine: affordable, huge context window. Output: 15 Markdown documents and the Mermaid diagram code.

Generation — formats for delivery

Running locally (with no API cost), Python takes the Markdown files and produces the 2 PPTX files, the 2 DOCX files, and renders the Mermaid diagrams as PNGs. Finally, it packages everything into folders ready for download.

// the 3 phases become 3 linked data models

CompanyInput  (nome, descrição, modo)
   → research()    → ResearchOutput   (fatos + fontes + custo)
   → synthesize()  → SynthesisOutput  (15 entregáveis + diagramas)
   → generate()    → GenerationResult (decks + relatórios + imagens)

Current facts

search, don't guess

Guided synthesis

15 prompts

Formatting

local, no cost

No black box

you step in

🏆 Why it feels like elite consulting

The secret isn't the AI model—it's what went into the prompts. The Factory was fed the real playbooks that McKinsey, BCG, KPMG, OpenAI and IBM publish, distilled into cheat sheets. The output inherits that rigor.

🧠 Borrowed authority

These PDFs are huge (28, 46 pages) and don't fit in the context all at once. The solution was a lightweight RAG: Gemini processed each PDF in chunks, extracting only what's valuable and cutting the buzzwords. What remained was pure knowledge—ready to guide every deliverable.

•These firms' frameworks become the foundation for the 15 prompts
•Each document comes out with the structure a senior consultant would use

✓ Why the output is convincing

✓Cutting-edge consulting language and structure
✓Perplexity's current facts, not generic ones
✓Established frameworks (maturity, quick wins, ROI)

✗ What still requires you

✗Review the numbers before sending them to the client
✗Validate what makes sense for that business
✗Adapt the tone and cut what doesn’t apply

💡 A lesson worth its weight in gold

Perceived quality comes from embedded knowledge, not the button. That's why all of Track 2 is about building that cheat sheet foundation: it's the asset that turns an ordinary AI into a consultant who seems elite.

Playbooks

McKinsey, BCG…

Distilled

Lazy RAG

Authority

borrowed

Perception

elite quality

🎬 Demo: a real package

Seeing the destination before the journey gives you a quality benchmark. Imagine running the Factory for "Stripe". In a few minutes, the output folder fills up—organized by file type. See what shows up, piece by piece.

// output/stripe/ — generated structure

output/stripe/
├── markdown/          # 15 .md (inventário, dores, roadmap, ROI…)
├── presentations/     # 2 .pptx (resumo executivo + análise)
├── documents/         # 2 .docx (relatório + SOW)
├── mermaid_images/    # diagramas atual vs futuro (.png)
├── research_cache.json
└── state.json         # progresso + custo total (~US$0,08)

Roadmap + ROI

Open the 30/60/90 roadmap: each time frame has prioritized initiatives. The ROI calculator projects the return—which justifies the investment in a board meeting.

AI policy + governance

The acceptable-use template and data governance framework—the kind of document companies pay lawyers and consultants to write from scratch.

Deck + SOW

The PPTX you present and the SOW that closes the contract—scope, timeline, and price ready. Together, they turn analysis into a closed deal.

💡 Use this as a benchmark

Remember the feeling of opening this package. Each course module builds one of these pieces—and the quality standard you just saw is the benchmark to reach when you run your own Factory.

Folders

organized by type

~US$0,08

package cost

Minutes

not weeks

Scale

quality reference

🧭 The course map

Now you’ve seen the whole Factory. This course takes you from here to building it piece by piece. Each track is a layer of the pipeline — you always know which part of the machine you’re assembling.

🥉

Track 1 · Fundamentals — the vision, anatomy, and setup (you are here)

🥈

Track 2 · Synthesis — the frameworks and knowledge base (Phase 1)

🥇

Track 3 · Skills & Agents — reusable resources and document generation (Phase 2)

💎

Track 4 · The Factory — Perplexity research, the 15 prompts, and orchestration (Phases 3-4)

👑

Track 5 · Consultant — frontend, sales, and the capstone with a real client (Phase 5)

🧱 Incremental construction

The sequence isn’t decorative: knowledge → skills/agents → the factory → the client. Each piece you build becomes input for the next track. By the end, the 3-phase pipeline you saw here is entirely in your hands—and you know how to fix and evolve each part.

Sequence

logical, not standalone

Layers

1 track = 1 piece

Incremental

each piece adds up

Capstone

real client

✅ Module summary

✓

19 deliverables — 15 Markdown + 2 PPTX + 2 DOCX, from inventory to SOW.

✓

3-phase pipeline — Research (Perplexity) → Synthesis (Gemini) → Generation (Python).

✓

Each phase has a job — collect facts, write, format. No black box.

✓

It seems elite because it is — BCG/KPMG playbooks distilled into the prompts.

🎯 Mission 1.2 — The 3 deliverables that sell best

Think of a company you know well (yours, a friend's, or a local brand). From the 19 deliverables, choose the 3 most likely to sell for it and write, one sentence each, why:

Deliverable 1: what pain point does it solve for this company?
Deliverable 2: why would the decision-maker pay for it?
Deliverable 3: what makes it urgent now?

Success: 3 deliverables selected with a 1-sentence justification. What you gained: clarity about which part of the package is your sales hook — and the foundation for the pitch we'll build in Trail 5.

Next module:

1.3 — Building the arsenal (Claude Code, API keys, and running the first package)

← Previous module Next Module →