Venture Studio

Six specialist agents

The Orchestrator plans before executing. Five specialist agents produce tangible, ownerable artifacts with full provenance.

Three-question intake form: describe your device, select regulatory pathway (e.g. FDA De Novo), choose development stage. The Orchestrator then activates six specialist agents in an animated build sequence, each reporting its first action. A unified workspace assembles around your venture.

• 3-step wizard: device → pathway → stage
• 6 agents animate in: Orchestrator, Research, Product, Engineering, Growth, Operations
• Pre-loaded with GlucoSense venture (non-invasive glucose monitoring wearable)
• Active venture workspace appears at completion

Narrative briefing card synthesises agent activity into 2–3 sentences: 'Research Agent completed market sizing and flagged a new Apple competitor. Engineering is blocked waiting for your product spec decision.' Decision cards show agent reasoning, recommended option, and consequences of inaction.

• 4 inbox tabs: Primary, Other, Later, Cleared
• Each decision card: agent reasoning, recommended badge, 'What happens if I skip?' collapse
• Founder Memory context on each decision
• Dependency chain indicators between decisions

Parallel venture streams — Science & Innovation (3 sub-spaces: PPG Signal Processing, Sensor Hardware, Biocompatibility), Clinical Development (5 sub-spaces including FDA Submission, Quality System), Commercial (4 sub-spaces), Operations (4 sub-spaces). Each space has contextual insight banners and agent skill indicators.

• Per-space insight banner: 'EXP-041 has been running 48h. Preliminary MARD: 14.2%'
• 3-panel layout: Resources | Canvas | Agents
• Suggested canvas prompts per space
• Agent skill relevance tooltips showing which agent handles what

AI-synthesised conversation summaries: '2 artifacts shared, 3 messages backed by citations, 4 thread replies.' Messages tagged with importance (Finding, Needs Input, Update). Interactive citation pills link to source details with quality ratings and confidence scores.

• Channel summary card with key metrics
• Importance tags: Finding, Needs input, Update
• Inline citation pills — click to expand source details
• Agent status indicators per channel

Three-layer provenance across every artifact: source count, confidence level, and ownership status. Provenance breadcrumb trails show the full chain from source to output. Every claim traceable. Cross-surface related links connect artifacts across spaces and channels.

• Provenance pills: source count + confidence % + owner badge
• Interactive citation popovers with full source details
• Cross-surface 'Related' section linking artifacts across modules
• Append-only audit trail for all agent actions

Low-stakes outputs proceed automatically. High-stakes outputs pause for founder approval with full reasoning shown. The system knows its confidence level and surfaces it transparently. GxP compliance enforced from day one. IP ownership transfers fully by month twelve.

• Autonomy levels per agent skill, visible in tooltips
• High-stakes decisions show: agent reasoning + recommended option + skip consequences
• Compliance state panel: GxP ENFORCED, QbD register LIVE, PCCP scaffold DRAFT, Citation guard ON
• Co-development independence gauge: tracks IP transfer progress toward month 12

Research paper browser with venture relevance scoring (0–100%). Six pre-loaded papers with citation counts, tags, and a relevance sentence like '95% Core CGM review — MARD benchmarks directly apply to EXP-041'. Convert papers to working proofs-of-concept via the PoC Builder.

• 3 tabs: Papers, Data, PoC Builder
• Venture relevance % with colour-coded badges
• 'Ask Research Agent' opens workspace with pre-loaded context
• Active PoC tracker: 'PPG Glucose Estimation Model' at 35% with Engineering Agent working

Screenshots