Wiki:Decks/AI Infrastructure
AI Infrastructure is a deck rebuild kit. Each section below carries one slide's title, subtitle, illustration, footer line and speaker notes, re-illustrated in the blue house style and de-identified for external use.
| Source | Slides | Illustrations | Style | Aspect | Generated |
|---|---|---|---|---|---|
Generic-AI-Infrastructure-Presentation.pdf |
22 | 27 | Blue | 21:9 (slide content zone) | 2026-09-19 |
How to rebuild
For each slide: create a slide from the corporate template, paste the Title and Subtitle into the template placeholders, insert the illustration into the content zone at full width, and type the Footer line into the footer text box where one is given. The template supplies the legal disclaimer and the logo — the illustrations deliberately contain neither, and carry no title or subtitle text of their own.
Slides
Slide 01 — The Acceleration Methodology
| Title | The Acceleration Methodology |
|---|---|
| Subtitle | Proven at Client A — this is what I'm here to share |
| Key message | A consultant plus methodology infrastructure produces manager-level output, letting a lean team deliver depth that traditional firms need three times the headcount to match. |
| Footer | This methodology allows a lean team to deliver depth that traditional firms need 3x the headcount to match. |
Consultant plus methodology infrastructure yields manager-level output, evidenced by Client A engagement.
- Speaker notes
- No speaker notes in source.
Slide 02 — The Real Bottleneck
| Title | The Real Bottleneck |
|---|---|
| Subtitle | Data ingestion consumes the clock — insights get rushed |
| Key message | Without infrastructure, most of the engagement clock is spent ingesting and cleaning data and the insight work is rushed at the end; an automated pipeline inverts that ratio. |
| Footer | The infrastructure elevates every level — freeing consultants, senior consultants, and managers to focus on the thinking that drives value. |
Time allocation with and without methodology infrastructure.
- Speaker notes
- No speaker notes in source.
Slide 03 — What We're Going to Cover
| Title | What We're Going to Cover |
|---|---|
| Subtitle | none |
| Key message | The session runs in three parts: the six foundational principles, how the infrastructure performs on a real engagement, and how to get started. |
| Footer | none |
The three parts of the session.
- Speaker notes
- No speaker notes in source.
Slide 04 — The Foundation
| Title | The Foundation |
|---|---|
| Subtitle | Six principles that make AI infrastructure work |
| Key message | Six model-agnostic principles underpin working AI infrastructure, and they hold regardless of which AI vendor is chosen. |
| Footer | These principles are model-agnostic and platform-independent. They work regardless of which AI you choose. |
The six foundational principles.
- Speaker notes
- No speaker notes in source.
Slide 05 — Clarity Pipeline
| Title | Clarity Pipeline |
|---|---|
| Subtitle | Sequential refinement for AI excellence |
| Key message | Output quality is set upstream: clear thinking produces clear writing, which produces clear prompting, which compounds into high AI efficacy. |
| Footer | none |
Clarity compounds through four sequential stages.
- Speaker notes
- No speaker notes in source.
Slide 06 — Code Before Prompts
| Title | Code Before Prompts |
|---|---|
| Subtitle | Prompts wrap code, not replace it |
| Key message | Roughly eighty per cent of the work should be deterministic code and twenty per cent AI judgment; prompts are a wrapper around a code foundation, not a substitute for it. |
| Footer | Prompts wrap code, not replace it. |
An eighty-twenty split between deterministic code and AI judgment.
- Speaker notes
- No speaker notes in source.
Slide 07 — CLI as Interface
| Title | CLI as Interface |
|---|---|
| Subtitle | Comparing GUI complexity with CLI elegance |
| Key message | A graphical interface spreads one task across nested menus and cluttered panels; a command line collapses the same task into a single instruction with legible output. |
| Footer | Text beats clicks. |
Interface complexity versus command-line directness.
- Speaker notes
- No speaker notes in source.
Slide 08 — Deterministic AI Architecture
| Title | Deterministic AI Architecture |
|---|---|
| Subtitle | Code as factory, intelligence as manager |
| Key message | Deterministic code machines execute the repeatable work on a production line while an intelligence layer above decides what runs and when. |
| Footer | none |
Option A — simple
Code executes on the line; the intelligence layer orchestrates it.
Option B — detailed
Detailed option: what each code machine does, the checks between them, and the intelligence layer directing the line.
- Speaker notes
- No speaker notes in source.
Slide 09 — Deterministic + Probabilistic
| Title | Deterministic + Probabilistic |
|---|---|
| Subtitle | Each has its place — use both strategically |
| Key message | Deterministic code owns reproducible computation and probabilistic models own judgment and synthesis; human oversight belongs at the handoff between them. |
| Footer | More human oversight is required as you move toward probabilistic — but that is where synthesis and storytelling happen. |
Deterministic and probabilistic work, with human oversight at the handoff.
- Speaker notes
- No speaker notes in source.
Slide 10 — Modular Skill Composition
| Title | Modular Skill Composition |
|---|---|
| Subtitle | Specialized agents leverage foundational skills — composition over duplication |
| Key message | Specialized agents do not duplicate capability; each composes the foundational skills it needs, so a skill built once is reused across every specialization. |
| Footer | Each agent in the pipeline composes the exact skills it needs — no duplication, maximum reuse. |
Agents compose foundational skills rather than duplicating them.
- Speaker notes
- No speaker notes in source.
Slide 11 — File System as Memory
| Title | File System as Memory |
|---|---|
| Subtitle | Organized structure powers intelligent retrieval |
| Key message | A disciplined directory structure is what makes retrieval possible; when every file has a home, the AI can find context instead of searching through chaos. |
| Footer | When your files are organized, your AI knows where to look. |
Structured storage feeding search, filter and retrieval.
- Speaker notes
- No speaker notes in source.
Slide 12 — Architecture Over Model
| Title | Architecture Over Model |
|---|---|
| Subtitle | Why infrastructure matters more than AI selection |
| Key message | The model is the visible tip of the system; the architecture beneath it is the larger mass that persists while models are replaced. |
| Footer | Models change quarterly. Architecture compounds annually. |
The model is the visible tip; architecture is the mass beneath.
- Speaker notes
- No speaker notes in source.
Slide 13 — Scaffolding Over Model
| Title | Scaffolding Over Model |
|---|---|
| Subtitle | The hidden ninety per cent where actual value lives |
| Key message | The AI model is the ten per cent everyone discusses; the six infrastructure layers beneath it are where value is actually created and where most organizations are blocked. |
| Footer | Where most organizations are blocked. |
Option A — simple
Six infrastructure layers beneath the visible model.
Option B — detailed
Detailed option: what actually lives in each of the six infrastructure layers beneath the model.
- Speaker notes
- No speaker notes in source.
Slide 14 — Putting It to Work
| Title | Putting It to Work |
|---|---|
| Subtitle | How this infrastructure performs on a real consulting engagement |
| Key message | Section break: the discussion moves from principles to performance on a live engagement. |
| Footer | none |
Section opener motif.
- Speaker notes
- No speaker notes in source.
Slide 15 — Augmented Consulting Methodology
| Title | Augmented Consulting Methodology |
|---|---|
| Subtitle | Where domain expertise meets systematic acceleration |
| Key message | A five-stage decision method runs from clarifying the decision to acting on it, with the modelling and analysis stages accelerated by AI and the whole loop feeding back into learning. |
| Footer | Humans provide domain knowledge and strategic judgment. AI agents execute systematic analysis 10x faster. |
Five stages from clarifying a decision to acting on it, with AI acceleration in the middle.
- Speaker notes
- No speaker notes in source.
Slide 16 — In Practice: Agent Orchestration Pipeline
| Title | In Practice: Agent Orchestration Pipeline |
|---|---|
| Subtitle | A real consulting engagement with human checkpoints |
| Key message | Five specialized agents run in sequence inside an autonomous execution zone, passing artifacts between them, with human checkpoints placed at the two points where judgment is required. |
| Footer | none |
Option A — simple
Five agents in sequence with human checkpoints at the judgment points.
Option B — detailed
Detailed option: the five agents, the artifact each hands to the next, and the two human checkpoints.
- Speaker notes
- No speaker notes in source.
Slide 17 — The Infrastructure in Action
| Title | The Infrastructure in Action |
|---|---|
| Subtitle | Multiple execution environments — same framework |
| Key message | The same framework runs unchanged in a cloud development environment and on a local machine; the execution environment is a choice, not a constraint. |
| Footer | The framework runs identically in both — and there are more options. |
Option A — simple
The same framework in a cloud environment and on a local machine.
Option B — detailed
Detailed option: the same file tree, editor, agent terminal and framework layers in a cloud environment and on a local machine.
- Speaker notes
- No speaker notes in source.
Slide 18 — What the Infrastructure Produces
| Title | What the Infrastructure Produces |
|---|---|
| Subtitle | Real consulting artifacts from a real engagement |
| Key message | The infrastructure produces five classes of auditable consulting artifact, from causal structure maps to data-quality audits, each reproducible and each adding to institutional knowledge. |
| Footer | Every artifact is auditable, reproducible, and builds institutional knowledge. |
Five classes of auditable artifact produced by the infrastructure.
- Speaker notes
- No speaker notes in source.
Slide 19 — Getting Started
| Title | Getting Started |
|---|---|
| Subtitle | The framework, your architecture choices, and what adoption looks like |
| Key message | Section break: the discussion moves from evidence to adoption. |
| Footer | none |
Section opener motif.
- Speaker notes
- No speaker notes in source.
Slide 20 — The Framework
| Title | The Framework |
|---|---|
| Subtitle | Daniel Miessler's blueprint for building persistent AI infrastructure |
| Key message | The framework has five components — identity, skills, hooks, tools and memory — and each maps onto something a firm already has informally. |
| Footer | You already have all of this — in people's heads and shared drives. This architecture makes it executable. |
The five components of the framework.
- Speaker notes
- No speaker notes in source.
Slide 21 — Your Architecture, Your Rules
| Title | Your Architecture, Your Rules |
|---|---|
| Subtitle | No single blueprint — right-size to your requirements |
| Key message | Hosting model, AI provider and data-security posture are three independent choices an organization controls; the framework stays constant across all of them. |
| Footer | The framework is the constant. The infrastructure choices are variables your organization controls. |
Option A — simple
Three independent architecture choices: hosting, provider and data security.
Option B — detailed
Detailed option: three independent decision bands, with what each option implies and the control-versus-effort trade-off beneath.
- Speaker notes
- No speaker notes in source.
Slide 22 — What Adoption Looks Like
| Title | What Adoption Looks Like |
|---|---|
| Subtitle | A phased approach — start small, prove value, scale |
| Key message | Adoption runs in three phases: a contained four-to-six week pilot, a three-to-six month practice rollout, and ongoing institutional capability, with a go or no-go decision at the end of phase one. |
| Footer | Phase 1 is low risk, low cost, and proves the concept on real work before any broader commitment. |
Three adoption phases from pilot to institutional capability.
- Speaker notes
- No speaker notes in source.
De-identification
The source deck was checked for third-party branding; 3 substitutions were made across 2 slide(s). The substitution log is held with the source material and is not published.
Source
Rebuilt from Generic-AI-Infrastructure-Presentation.pdf on 2026-09-19. Original slides are preserved outside the wiki.
