Ongoing

CADE Production Tools

AI Workflow Designer · 2026 · Supporting toolset · 3 min read

Designed three AI-assisted production workflows that fill gaps in the CADE source package — an op-board generator, an annex generator, and a briefing builder — so exercise production isn't blocked when a unit's inputs are incomplete.

Overview

CADE requires an OPORD-quality source package before anything else can be produced. In practice, units don't always arrive with complete operational graphics, annexes, or briefing materials. I built three supporting workflows to close those gaps: an op-board generator that produces operational graphics when none exist, an annex generator that builds annexes on demand, and a briefing builder that turns Markdown slide scripts into PPTX and HTML outputs. Each workflow is AI-assisted and operator-reviewed before the output enters the source package.

Problem

CADE's source-truth discipline depends on having complete, coherent inputs before downstream products are generated. When a unit doesn't have operational graphics or annexes, the entire artifact chain stalls. When briefing materials need frequent revision, manual formatting slows the iteration cycle. These gaps needed filling without bypassing human review or degrading source integrity.

Constraints

  • Op-boards and annexes must be coherent with the OPORD before they enter the source package.
  • Briefings must remain easy to revise as CADE evolves.
  • AI output in all three workflows requires operator review before use.
  • No workflow is allowed to introduce drift into the source layer.

Approach

I treated each workflow as a distinct production problem with its own input, output, and review gate. The op-board generator takes scenario and operational context and produces graphics to spec. The annex generator builds supporting annexes when the unit hasn't produced them. The briefing builder keeps slide content in editable Markdown and handles PPTX and HTML output, so revision effort stays on content rather than formatting.

AI Operator Skill Demonstrated

  • Identified where the production chain was most likely to stall and built targeted workflows for each gap.
  • Kept AI output at draft status until operator review cleared it for the source package.
  • Separated content authoring from formatting so revision cycles stayed fast.
  • Designed each workflow to feed the source layer, not bypass it.

Key Decisions

Build gap-filling tools rather than require complete unit inputs

Reasoning:

Waiting for a unit to produce missing op-boards or annexes before exercise production could begin would block CADE adoption. Controlled AI generation with operator review is faster and doesn't sacrifice source integrity.

Alternatives considered:
  • Require units to supply complete inputs before production starts
  • Let controllers improvise missing materials at execution time

Use Markdown as the briefing source layer

Reasoning:

Markdown keeps slide content readable, reviewable, and version-controlled before it hits any presentation format. Revisions stay fast because content and formatting are separate.

Alternatives considered:
  • Author directly in PowerPoint
  • Maintain separate scripts for each output format

Tech Stack

  • Claude
  • ChatGPT
  • Markdown
  • PPTX
  • HTML

Result & Impact

These tools removed the production blockers that would otherwise stall CADE when unit inputs are incomplete. The briefing builder reduced revision cycles from format-heavy editing to content-only changes. The op-board and annex generators gave the design authority the ability to complete a source package on demand rather than waiting on the unit.

Learnings

  • The most valuable AI production work is often in the gaps — the inputs no one thinks to provide until the chain is already blocked.
  • Operator review at the edge of each workflow, not at the end, is what keeps AI output from drifting into the source layer unchecked.
  • Briefing automation pays for itself in the second revision, not the first.
All projects