Skip to content

robertholzer42-stack/applied-innovation-platform

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

36 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Applied Innovation Platform

An open-source, AI-powered innovation analysis system that applies Future Thinking, Design Thinking, and Systems Thinking together through 14 specialized agents. Inspired by my 8+ years leading Innovation teams, IFTF's frameworks and Braden Kelley's Resilient Innovation framework and designed to run inside Claude (Anthropic).

The Problem

Most organizations have a federated or disconnected model to drive innovation, don't connect innovation back to strategic goals and often don't think about the various innovation skills needed to be successful. They run design thinking workshops but ignore the systems their ideas have to survive in. They do signal monitoring and scenario planning but never ask whether anyone wants what they're building. One-dimensional innovation fails.

The Solution

This platform orchestrates 14 AI agents, each with a distinct personality, analytical framework, and output format, into a repeatable engagement pipeline. Each agent is a Markdown skill file that Claude reads and follows. No infrastructure, no API keys, no code to deploy. Just structured prompts that turn Claude into a multi-dimensional innovation practice.

What it covers:

  • Product and service innovation
  • Business model disruption analysis
  • Process and operational redesign
  • Strategic foresight and scenario planning
  • Innovation portfolio management
  • Change readiness assessment

Prerequisites

  • A Claude Pro, Team, or Enterprise subscription (claude.ai)
  • OR Claude Code / Claude Cowork (desktop)
  • Basic familiarity with Claude Projects

Quick Start (10 minutes)

1. Clone this repo

git clone https://github.com/robertholzer42-stack/applied-innovation-platform.git

2. Create a Claude Project

In Claude (web or desktop app), create a new Project. Upload all 14 files from the skills/ folder (14 agent skills plus the setup guide) as project knowledge.

3. Add the project instructions

Copy the contents of CLAUDE.md into your project's custom instructions field.

4. Run your first analysis

Run a full applied innovation analysis on [your challenge here]

That's it. The Conductor agent will route your challenge to the right agents and produce an integrated strategy.

See the Setup Guide for detailed instructions, alternative deployment options (Claude Code, Cowork), and a walkthrough of your first engagement.

Agent Architecture

Tier 1: Client Interface

Codename Role
Navigator Challenge discovery, innovation maturity assessment, engagement scoping
Publisher Board-ready decks, strategy reports, workshop guides

Tier 2: Core Thinking

Codename Framework
Scout Future Thinking - IFTF-aligned Prepare-Foresight-Insight cycle
Empathy Design Thinking - human-centered design process
Architect Systems Thinking - causal loops, leverage points, consequence mapping

Tier 3: Intersection

Codename Intersection
Visionary Future + Design - future personas, speculative design, scenario prototyping
Integrator Design + Systems - scalable solutions, pilot-to-production pathways
Sentinel Future + Systems - resilience mapping, stress testing, adaptive strategy

Tier 4: Operational

Codename Role
Radar Competitive intelligence, market monitoring, positioning analysis
Banker Innovation portfolio management, Horizon 1/2/3 balancing
Scorekeeper DVFA scoring (Desirability, Viability, Feasibility, Adaptability)
Bridge Change management, adoption readiness, transition design

Tier 4.5: Quality Assurance

Codename Role
Critic Independent quality review with five-criterion evaluation (Completeness, Evidence Quality, Fabrication Audit, Writing Standards, Integration Readiness), PASS/REVISE/FLAG verdicts, web verification of numbers and citations, [HUMAN-VERIFY] flags for inconclusive items

Tier 5: Orchestration

Codename Role
Conductor Cross-agent synthesis, conflict resolution, unified strategy

How It Works

The Conductor manages an 11-stage pipeline with Critic checkpoints:

  1. Intake (Navigator) - Scope the challenge, assess innovation maturity
  2. Research Foundation (Conductor) - Shared fact base so all agents work from the same data
  3. Core Analysis (Scout + Empathy + Architect) - Three independent analytical lenses, run in parallel
  4. Critic Review - Quality check on core analysis outputs
  5. Preliminary Scoring (Scorekeeper) - Early DVFA scores to guide downstream agents
  6. Intersection Synthesis (Visionary + Integrator + Sentinel) - Where the lenses overlap
  7. Critic Review - Quality check on intersection outputs
  8. Operational Context (Radar + Banker + Bridge, then Scorekeeper final) - Real-world grounding + final DVFA with delta tracking
  9. Critic Review - Quality check on operational outputs
  10. Orchestration (Conductor) - Cross-agent synthesis, conflict resolution
  11. Deliverables (Publisher) - Board-ready artifacts with reusable design system

You can run the full pipeline, use individual agents directly, or ask for a quick 2-3 agent assessment.

Engagement Depth Tiers

Every engagement runs at one of three depth levels. Each level controls how many agents participate, how deeply each agent digs, and how many sessions the analysis spans.

Depth Agents Sessions Time When to Use
Quick scan 2-3 agents + Conductor 1 session 15-30 min Initial exploration, gut-checking an idea, single-lens question
Standard 6-8 agents + Conductor + Publisher 2 sessions 2-4 hours Most real decisions, product strategy, market entry, process redesign
Deep dive All 14 agents + Publisher + QA 4 sessions Full day+ High-stakes strategy, portfolio transformation, board-level decisions

Quick scan skips the research foundation, runs 2-3 agents at their lightest toolkit, and produces a focused brief in a single conversation. Good for: "Should we even look at this?"

Standard includes the shared research foundation, two-pass DVFA scoring, and the full intersection layer. Agents run their mid-depth toolkits with proper evidence chains. Produces an integrated strategy and board-ready deck. This is the right choice for most real work.

Deep dive runs every agent at maximum depth. Scout builds full scenarios with cross-impact analysis. Empathy creates detailed validated personas. Architect runs complete causal loop analysis. The Devil's Advocate pass is recommended. Produces the full deliverable suite. Reserve this for decisions where the stakes justify a full day of analysis.

Standard and deep dive engagements span multiple sessions using checkpoints to maintain continuity.

To set depth, include it in your prompt: "Run a standard analysis on [topic]" or "Deep dive on [topic], with Devil's Advocate."

DVFA Scoring Framework

Every innovation opportunity is scored across four dimensions:

  • D - Desirability (from Design Thinking): Do people want this?
  • V - Viability (from Systems Thinking): Can this work in real-world systems?
  • F - Feasibility (from operational reality): Can we build and deliver this?
  • A - Adaptability (from Future Thinking): Will this survive across multiple futures?

Each dimension scored 1-5 with confidence levels. The result: an Innovation Score that goes beyond the typical desirability-viability-feasibility triangle.

Verification Framework

The platform applies fact-checking discipline at every stage. AI output sounds confident regardless of whether it is correct. Confidence is not correctness. The platform's principle: capability without verification is a liability.

Stake Level Drives Verification Effort

Every engagement carries a stake level set by Navigator or Conductor at intake:

Stake Level Description Critic Mode
Not Much Internal exploration, draft for review Quick: red flag scan only
Embarrassing Client-facing, named publication, board update Standard: web verify all numbers and citations, flag inconclusive items
Real Damage Board decision, regulatory filing, financial commitment Deep: full verification + cross-model second-opinion pass

If stake level is unspecified, the Conductor defaults to Embarrassing and announces the assumption.

Three Layers of Verification

  1. Producer-side honesty. Scout, Empathy, Architect, and Radar each run an Honesty Protocol at output time, scanning their own drafts for six fabrication risk patterns and applying citation hygiene rules. The cheapest place to prevent fabrication is at the source.
  2. Critic Fabrication Audit. A dedicated fifth evaluation criterion that hunts the three error types (phantom statistics, fake citations, reasoning errors in true data), checks four canonical logic failures, and at Standard or Deep depth performs web verification of every specific number and citation. Items that cannot be confirmed are flagged with [HUMAN-VERIFY] rather than passed silently.
  3. Conductor Verification Governance. Pre-synthesis triage of all unresolved items, three-prompt stress test on the draft synthesis (argue opposite, identify falsification conditions, surface strongest counterargument), and autonomy-level classification (Draft / Direct Action / Chained Action) with reversibility checks.

Cross-Cutting Rules

  • No false precision. Confidence drives precision: H confidence allows 0.5 increments, M confidence whole numbers, L confidence ranges only.
  • [NEED: data from X] not fabrication. When evidence is unavailable, mark the gap explicitly. Never invent.
  • [HUMAN-VERIFY] items must surface. Items the Critic could not confirm appear in the final deliverable's appendix, not quietly dropped.
  • Confidence inherits the weakest input. Aggregate scores cannot exceed the precision of their underlying evidence chains.

Repository Structure

skills/              - 14 agent SKILL.md files + design system + setup guide
docs/                - Example deliverables, architecture diagram, PPTX template
engagements/         - Your engagement workspaces (one folder per challenge)
  [challenge-name]/
    checkpoints/     - Checkpoint files for multi-session runs
    reviews/         - Critic quality reviews
knowledge-base/      - Reusable patterns, case studies, frameworks
  patterns/
  case-studies/
  frameworks/
CLAUDE.md            - Project instructions (paste into your Claude Project)

Key Reference Files

File Purpose
docs/publisher-deck-template.js Reusable PptxGenJS template with design system, slide builders, and validation functions
skills/publisher-design-system.md Visual identity spec: colors, typography, layout rules, QA checklist
knowledge-base/patterns/TEMPLATE.md Template for capturing reusable patterns from engagements
knowledge-base/case-studies/TEMPLATE.md Template for documenting completed engagements
skills/critic-SKILL.md Quality evaluator - independent four-criterion review of agent outputs

Multi-Session Engagements

Standard and deep engagements span multiple sessions using checkpoints to maintain continuity across conversations. The Conductor writes checkpoint files to engagements/[challenge-name]/checkpoints/ at the end of each session.

Each checkpoint captures:

  • Which agents have completed their analysis
  • Current DVFA snapshot and scoring state
  • Next steps and which agents are queued

To resume an engagement: start a new conversation, load the platform, and read the latest checkpoint file. The Conductor picks up where it left off.

Customization

This platform is designed to be forked and adapted. Common customizations:

  • Add industry-specific context to agent skills (e.g., add healthcare regulations to Architect)
  • Adjust DVFA weights for your organization's priorities
  • Create new agents following the SKILL.md pattern
  • Integrate your own tools - the Architect agent supports pluggable analytical frameworks
  • Modify writing rules in CLAUDE.md to match your organization's voice

Methodology Foundations

  • Future Thinking: Based on the Institute for the Future's Prepare-Foresight-Insight cycle, with 50+ years of proven foresight methodology
  • Design Thinking: Human-centered design with empathy research, personas, journey mapping, and prototyping briefs
  • Systems Thinking: Causal loop analysis, leverage point identification, and consequence mapping

The innovation at the core of this platform is that these three disciplines are applied together, systematically, every time. The intersection agents (Visionary, Integrator, Sentinel) exist specifically to find insights that no single lens would reveal.

Contributing

Contributions welcome. Some ideas:

  • New agent skills for additional analytical dimensions
  • Industry-specific templates and context
  • Example engagements that others can learn from
  • Improvements to existing agent frameworks
  • Documentation and tutorials

Please open an issue to discuss significant changes before submitting a PR.

Credits

Created by Bob Holzer. Inspired by IFTF and Braden Kelley's Resilient Innovation framework, which articulates how Future Thinking, Design Thinking, and Systems Thinking intersect to produce innovations that are desirable, viable, and resilient.

License

Apache License 2.0 - See LICENSE for details.

About

AI-powered multi-dimensional innovation analysis using Future Thinking, Design Thinking, and Systems Thinking with 14 specialized agents. Runs inside Claude.

Topics

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors