Drill into a single task

Example task decomposition

Tools daedarch builds (n=8)

[{'name': 'Adaptive Differentiation Engine', 'description': 'Harvester ingests student assessment data (formative quizzes, prior grades, reading level, IEP goals) via school system connectors. Agent orchestrator generates personalized learning pathways—custom worksheets, video recommendations, scaffolded problems—for each student tier. Content production engine adapts difficulty, modality (text/video/audio), and pacing. Teacher approves pathway templates once; system generates individualized mat…

Tools generic-AI (n=5)

[{'name': 'ChatGPT/Claude for Lesson Planning', 'description': "Teachers prompt an LLM to generate lesson outlines, discussion questions, or activity ideas. Vendor pitch: 'Save hours on lesson prep.' Reality: Teacher still writes the actual lesson, adapts it to their class, and validates pedagogical soundness.", 'automation_level': 'augment'}, {'name': 'Grading Assistant (e.g., Turnitin + AI)', 'description': "AI flags plagiarism and provides rubric-based scoring suggestions. Teacher reviews and…

Why drill this deep

Aggregate "X% of jobs automated" claims hide the granularity where displacement actually happens — at the task level within an occupation. The workbench surface lets you compare what daedarch would build (contract-gated, traceable, reusable) vs what a generic chatbot deployment would do (ad-hoc, untraceable, single-use). The contrast is what justifies the daedarch architecture.

Programmatic access

POST /api/v4/execute
{ "tool_id": "daedarch.business.atlas_intelligence_v1",
  "inputs": { "naics_code": "<6-digit>", "include_transformations_full": true } }
Returns tools_daedarch_builds[] + tools_generic_ai[] per scenario.