BRAIN
An integrated innovation platform with multi-agent execution of methodologies. Built by Quantum Leap for an innovation consultancy. In implementation.
An innovation consultancy runs structured methodologies for major brands — Open Innovation, Strategy & Foresight, value-creation methodologies, and others. Each methodology has stages, deliverables, decision points, and a rhythm. Each engagement requires consultants to execute those stages repeatedly, across many clients, with consistency.
The repetition was eating senior consultant time. Stage-by-stage execution is where the methodology lives, but it is also where the most automatable work happens. The consultancy wanted a platform that executes methodology stages with AI agents instead of with consultants, freeing senior practitioners for the work that actually requires judgment.
The constraint that made the project hard was customization. The consultancy does not run a fixed catalog of methodologies. Different clients require different blends. Sometimes a new methodology needs to be defined from scratch. The platform had to support both: out-of-the-box methodology execution and user-defined methodology authoring.
What was at stake: scaling the consultancy without proportionally scaling the headcount, and protecting the IP that the methodology library represents while making it instantiable on demand.
The engagement spans Horizon 02 and Horizon 03.
Horizon 02 mapped the existing methodology library. Each methodology was decomposed into stages, decision points, deliverables, and the cognitive tasks involved. The mapping surfaced which tasks were amenable to agent execution and which required human judgment. The result was a methodology meta-model — abstracting what a methodology is at a structural level, independent of any specific one.
Horizon 03 designed the multi-agent execution layer on top of that meta-model. Agents are role-bound (one agent per stage), context-aware (they know where they are in the methodology), and customizable (users can author new agents or modify existing ones to fit a new methodology).
The integration layer was designed in parallel. SharePoint for document storage and Teams for collaboration are the platforms the consultancy's clients already live in. ChatGPT, Perplexity, and internal tools integrate via API to extend agent capabilities. The platform meets consultants where they already work.
Implementation runs in active deployment across the consultancy's practice areas. Quantum Leap built the platform and continues to deliver new agents as new methodologies are authored.
The architecture has four core systems.
System 01 / Methodology engine. The structural meta-model. Each methodology is defined as a sequence of stages, each stage with input requirements, output deliverables, decision points, and assigned agents. The same engine executes Open Innovation, Strategy & Foresight, value-creation methodologies, and any user-authored methodology that conforms to the meta-model.
System 02 / Multi-agent execution layer. Agents specialized by stage:
01 RESEARCH AGENT gathers external signals via Perplexity API 02 SYNTHESIS AGENT compresses input into stage-appropriate output 03 IDEATION AGENT generates options under explicit constraints 04 CRITIQUE AGENT challenges output against methodology criteria 05 DOCUMENT AGENT formats deliverables for the stage 06 HANDOFF AGENT prepares the input package for the next stage
When a user authors a new methodology, they define which agents run at which stage. They can compose from the existing agent pool or author new agents with custom prompts and tool access.
System 03 / Intelligent project management. Project tracking embedded in the platform. Stages, deliverables, deadlines, decisions, and team allocations are first-class entities. AI surfaces at-risk projects, suggests stage transitions, and predicts effort.
System 04 / AI-managed Mind Map. A visual workspace where ideas, evidence, and decisions emerge across a project. The Mind Map is not a passive canvas. An agent reads the Mind Map continuously, suggests connections, surfaces gaps, and updates the structure as the project evolves.
Integrations. SharePoint for document storage. Teams for collaboration and conversations within projects. ChatGPT, Perplexity, and internal tools via API for extended agent capabilities. Admin dashboard for usage telemetry, agent performance, and consumption.
The platform is in active implementation across the consultancy's practice areas. The first methodologies — Open Innovation, Strategy & Foresight, value-creation methodologies — are running with agent execution. New methodologies are being authored by senior consultants in the user-defined methodology editor.
Three structural changes.
Senior consultant time recovered. Stages that previously required hours of execution now run in minutes with agent execution, leaving consultants free for client conversations and judgment-heavy work. The headcount-to-engagement ratio is shifting.
Methodology IP became instantiable. Methodologies that previously lived in slide decks and tribal knowledge are now formal artifacts in the platform, executable on demand. New methodologies can be authored as easily as new agents can be configured.
The agent pattern became reusable. The six core agents — Research, Synthesis, Ideation, Critique, Document, Handoff — are reused across all methodologies. Each new methodology composes from the same set, with custom prompts as needed. The platform is becoming a methodology execution engine, not a single-purpose tool.