Governed decision simulation with hybrid agents

From weak signal
to governed
strategic decision.

Rehearse an executive decision or buyer committee before the proposal is sent. KayrosLab combines governed agents with authorized stakeholder profiles to expose vetoes, objections and missing evidence.

Test the decision path Request trial → Opens directly
Already adopted by strategy & innovation teams in early access
REFERENCE PROTOTYPE
kayroslab-complete-with-ai-agents.html • Complete with AI agents
Architecture Plan-and-Solve + ReAct
Specialized Agents 6+ (Red Team, Devil's Advocate…)
Memory Shared + Vector
Deployment P0 Standalone → P2 Cloud
THE KAYROSLAB MODEL

A structured 8-step process.
Hybrid AI + Human governance.

Unlike free-form LLM conversations, KayrosLab follows a cyclical workflow inspired by best practices in strategy and foresight, industrialized by specialized AI agents and formalized Human-in-the-Loop. The 4th step Position confronts the idea with the market through automated competitive analysis, the 7th Project transforms the decision into a piloted trajectory, the 8th Realize ensures execution and measurement, then the cycle loops back to Listen.

01

Listen

Noise reduction, qualification and scoring of weak signals.

→ Qualified Signals
02

Map

Trend network, bisociative bridges and strategic relationships.

→ Network + Bridges
03

Build

Scenarios, Collision Mode and structured collaborative briefs.

→ Scenarios + Brief
04

Position

Ontological competitive analysis: Cytoscape.js graph, property inspector, query playground, OWL export.

→ Ontology + Gap Analysis
05

Test

Multi-agent future-proofing: Critic, Devil's Advocate, Red Team.

→ Challenged Ideas
06

Arbitrate

Multi-criteria votes, Working Groups and structured final decision.

→ Decision + Deliverable
07

Project

Trajectory, roadmap, probabilistic simulation and learning loop.

→ Roadmap + KPIs
08

Realize

Pilot, deployment, review: tracked milestones and actual impact vs projected.

→ Actual vs Projected → loop
04

Position

NEW

Ontological competitive analysis: web scraping + GitHub/GitLab KPIs (stars, forks, contributors, commits, freshness, issues) to position the idea in a 14 entity-type ontology (7 tech, 7 business) connected by oriented relationships — all visualized in a Cytoscape.js graph with property inspector, natural language query playground and OWL RDF/XML export compliant with the Microsoft Ontology Playground. The differentiation gap directly feeds the Kayros Index and prepares the Red Team.

Agents: Positioner, Ontology Designer, Query Engine Deliverable: Ontology + Graph + Gap Analysis + OWL Export
Launch positioning analysis

Each step is traceable, evaluated by agents and subject to human gates (COMEX veto, expert validation, etc.).

Try the standalone demo now No installation • Opens in browser
TOOL REGISTRY & INTEGRATED TOOLS

Specialized tools
for each process step.

KayrosLab has an extensible Tool Registry. Agents (via ReAct) can call these tools autonomously or semi-autonomously. Here are the main tools associated with each step.

01 Listen
  • signal_scoring
  • context_enrichment
  • search_regulatory_risks
02 Map
  • trend_network_builder
  • bisociation_detector
  • weak_signal_mapper
03 Build
  • scenario_generator
  • brief_structurer
  • collision_mode
04 Position
  • search_competitors
  • analyze_github_landscape
  • ontology_builder
  • graph_explorer
  • nl_query_playground
  • gap_analyzer
  • owl_exporter
05 Test
  • red_team_attack
  • calculate_ki_impact
  • risk_assessor
06 Arbitrate
  • ki_scorer
  • pdf_livrable_generator
  • vote_aggregator
07 Project
  • roadmap_builder
  • scenario_simulator
  • kpi_tracker
  • feedback_loop
08 Realize
  • milestone_tracker
  • impact_recorder
  • variance_analyzer
  • closure_report
All tools are declared in the Tool Registry and exposed to agents via function calling. The registry is extensible (you can add your own business tools).
Output from Step 06 • Project

Strategic Dashboard

Once the decision is arbitrated, KayrosLab transforms the Go into a piloted trajectory: roadmap, KPIs, probabilistic scenarios and automatic re-arbitration loop.

Strategic Dashboard
Project "Energy Diversification" • Updated 2h ago
Active loop
Kayros Index 78
Strategic Steering Indicators
Kayros Index
78/100
+6 pts depuis arbitrage
Success Probability
68%
P50 • Intervalle 52–81%
Projected ROI (P50)
2.4×
P90 : 3.1× • P10 : 1.3×
Residual Risk
Moyen
3 red flags actifs
Time-to-Value
9 mois
Premier impact mesurable
Executive Alignment
Élevé
4/5 décideurs
Next Gate
J+45
COMEX • Go / No-Go
Budget Consumed
38%
142 k€ / 375 k€
Hypothesis Coverage
81%
17/21 validées
Leading Lagging Risk Budget / Hypotheses
Probabilistic Simulation (P10 / P50 / P90)
Base Scenario Expected Value
Optimistic Scenario P90
Adverse Scenario P10
Budget Impact Simulation
Total Allocated Budget
375 k€
Consumed to Date
142 k€ (38%)
Remaining to Commit
233 k€
Base Scenario
361 k€
−14 k€ vs budget
Optimistic Scenario
298 k€
−77 k€ (under-consumption)
Adverse Scenario
452 k€
+77 k€ (overrun)
Delay Impact Simulation (POC)
Hypothesis: marginal daily cost + opportunity loss
+15 jours
+18 k€
Budget → 393 k€
Prob. succès −3 pts
TtV +0,5 mois
+30 jours
+37 k€
Budget → 412 k€
Prob. succès −6 pts
TtV +1,0 mois
Référence
+45 jours
+52 k€
Budget → 427 k€
Prob. succès −9 pts
TtV +1,5 mois
+60 jours
+71 k€
Budget → 446 k€
Prob. succès −14 pts
TtV +2,0 mois
Alert Threshold
Beyond +45 days, ROI drops below 2.0× and the adverse scenario becomes the most likely.
Mitigation Lever
Partner mutualization + phasing → −18 % CAPEX and reduced delay risk.
Cost Sensitivity Analysis
Tornado Chart — ±20 % impact on total budget
Negative Impact (−) Base Positive Impact (+)
Partners / subcontracting
±48 k€
POC duration (timeline)
±37 k€
Internal headcount
±29 k€
Compliance & regulatory
±18 k€
Go-to-market / acquisition
±12 k€
Critical High Medium / Low
Critical
Une hausse de 20 % des coûts partenaires fait basculer le ROI sous 2,0× dans le scénario base.
Recommended Action
Négocier un plafond de dépenses partenaires + clause de révision à J+45.
Return on Investment Analysis
24-month horizon • Base vs ranges
Base ROI
2.4×
+140 % net
Payback
11 mois
Capital Recovery
NPV (approx.)
+412 k€
Over 24 months
ROI by Probabilistic Scenario
Optimistic Scenario
3.1×
P90
Base Scenario
2.4×
P50
Adverse Scenario
1.3×
P10
Coûts totaux (24 mois)
375 k€
Partenaires / sous-traitance192 k€
Équipe interne98 k€
Conformité & outils47 k€
Go-to-market38 k€
Bénéfices projetés (24 mois)
900 k€
Gains opérationnels410 k€
Nouveaux revenus320 k€
Évitement de coûts120 k€
Valorisation stratégique50 k€
Breakeven Point
The project becomes profitable from month 11. Beyond 18 months, ROI exceeds 2.8× in the base scenario.
Watch Point
In the adverse scenario (P10), ROI remains positive (1.3×) but payback extends to 19 months.
Execution Roadmap (Now / Next / Later)
Now
Team & budget structuring
D+0 → D+30
Next
Market POC & partners
D+30 → D+90
Later
Scale & industrialization
D+90 → D+180
Risk Matrix
Supplier Dependency High
Regulatory Medium
Market Adoption Low
1 active re-arbitration trigger
Feedback Loop

KPIs and signals are automatically reinjected into the Listen step. A breached threshold triggers a formal re-arbitration.

Next injection: D+7
Generated Deliverables
Project Sheet + Roadmap
Gantt & RACI
KPI Table

This dashboard is automatically generated at the end of the Project step and stays alive through the feedback loop.

TECHNOLOGY CORE

Governed Agentic
Orchestration

The orchestrator transforms a simple conversation into a structured, cyclical, resilient strategic decision system across 8 steps.
Risk Reduction
Systematic challenge via Red Team agents + mandatory human gates. Fewer blind spots and reinforced protection against classic LLM biases and hallucinations.
Full Auditability
Complete timeline, KI scoring, agent logs and human decisions. Compliant with Board, NIS2, DORA and GDPR requirements.
Controlled Scalability
From standalone pilot to sovereign local or private cloud deployment. Governance and control maintained at every step.

Your sales oracle knows where the deal can stop.

Upload the RFP, proposal, business case and constraints. KayrosLab turns them into cited evidence, then builds a governed decision twin from system agents, your experts and consented hybrid profiles. Run the same offer through the people who approve it, fund it, secure it, operate it or veto it — then strengthen it before the real committee meets.

Hybrid agent connecting an executive committee and a buyer veto network to a proposal
Live secure workspace

Start a Sales Oracle case

The tool is now embedded here: connect to your tenant, define the decision to test and upload the evidence corpus directly to the governed API.

01 · Workspace

Connect securely

Use the account provisioned for your organization. Access remains isolated to your tenant.

Request access

The session token remains in memory only and disappears when the page is refreshed or closed.

02 · Decision case

Frame the decision

Reuse an existing case or create the precise decision question the hybrid committee must test.

03 · Evidence

Upload the corpus

Files are fingerprinted locally, uploaded through a short-lived signed URL and verified before ingestion.

PDF, DOCX, TXT, Markdown or CSV · 50 MB/document · 20 documents/case · 250 MB/case.

    Rehearse your company decision

    Model a strategic decision with the perspectives of the executives who will challenge it. The result is a decision dossier, not a synthetic endorsement.

    1. 1Upload the decision corpusAdd the proposal, business case, metrics and constraints; every fact keeps its source.
    2. 2Compose the augmented committeeSelect system or custom agents and assign consented LinkedIn or Crystal Knows hybrid profiles.
    3. 3Run the cited swarmEach role evaluates the same evidence and captures GO, conditional GO, NO-GO, vetoes and missing proof.
    4. 4Arbitrate as a humanCompare positions, expose friction and let the accountable executive make the final decision.

    Pressure-test the customer RFP

    Turn the buying committee into a pre-mortem for the deal. Rehearse the proposal against procurement, finance, security, legal, operations and the business sponsor.

    1. 1Upload the RFP and proposalBuild one controlled corpus from requirements, response, pricing, security, contractual and delivery evidence.
    2. 2Map the veto chainIdentify buyers, influencers, budget owners, technical evaluators and veto holders; apply only authorized personality signals.
    3. 3Simulate objections and evidence gapsMake every objection traceable to source evidence, a role rule or an explicitly labelled assumption.
    4. 4Rewrite and negotiateProduce a stronger proposal, objection matrix, conditional-GO checklist and negotiation brief.
    See vetoes before submissionFind the committee member, rule or missing proof most likely to stop the deal.
    Tailor evidence by buyerGive Finance, Security, Legal and Operations the proof each function actually needs.
    Prepare the negotiationEnter the meeting with anticipated objections, fallback conditions and clear owners.
    Keep the simulation governedConsent, provenance, tenant isolation, audit logs and human arbitration stay attached to the run.

    The Sales Oracle is a decision-rehearsal system, not a mind reader or prediction engine. It labels simulated stakeholder feedback, uses official connectors or authorized exports, and never replaces the real buyer's decision.

    Prepare a Sales Oracle analysis
    TECHNOLOGY ARCHITECTURE

    A zero-dependency engine.
    A governed architecture.

    KayrosLab is built around a pure Node.js core (ESM, zero-dependency) that orchestrates AI agents, applies Human-in-the-Loop governance, and computes trajectories deterministically — without ever inventing numbers.

    Zero-Dependency Core

    Pure Node 20+ ESM engine — zero npm dependencies. 81+ unit tests. Usable client-side (browser imports) and server-side (Fastify).

    Multi-Agent Orchestration

    Plan-and-Solve + ReAct. 6 specialized agents: Planner, Critic, Devil's Advocate, Red Team, Bisociator, Synthesizer — each with a defined role in the ideation pipeline.

    Human-in-the-Loop Governance

    4 gate types (expert_review, red_team_veto, comex_arbitrage, output_censor) and 4 human roles. Absolute veto with mandatory justification. Gates persisted to disk.

    Deterministic Calculations

    Monte-Carlo (seeded PRNG), Kayros Index scoring (5 strategic × 6 technical dimensions), ROI, TCO, NPV — all computed, never hallucinated. The LLM provides hypotheses; the engine calculates.

    Pluggable LLM Providers

    Mock (offline), Ollama (sovereign), Anthropic Claude (proxy), HTTP Backend. RoutingPolicy with automatic failover and Circuit Breaker per provider (3 strikes, 30s cooldown). Client never sees API keys.

    Vector Memory

    InMemory or Qdrant. Semantic recall via cosine similarity with ideaId filtering. Used for agent context (ReAct loop) and the Monitor/Listen feedback loop.

    Resilience & Circuit Breaker

    Exponential backoff with full jitter. 3-state Circuit Breaker (CLOSED/OPEN/HALF_OPEN). withResilience() composes retry + breaker + fallback. Everything is non-blocking.

    Ontological Positioning

    14 entity types, typed properties, oriented relationships. Web + GitHub scraping. OWL RDF/XML export. Cytoscape.js graph with inspector, query playground, competitive gap analysis.

    Conversational Connectors

    Slack (Block Kit, modals, slash commands), Teams/Discord (Adaptive Cards). Abstract View/Action translated to native formats. Account linking via single-use cryptographic tokens.

    Strategic Steering & KPIs

    Kanban with gates, per-stage configurable scorecards, weighted votes, impact tracking (projected vs actual), leaderboards, campaigns with moderation, CSV export, funnel analysis.

    Explore the source code
    COMPLIANCE & SOVEREIGNTY

    An architecture designed for the most
    demanding compliance environments.

    KayrosLab was designed from the ground up to meet the sovereignty, traceability and security requirements of regulated organizations and strategic sectors.

    Data Sovereignty

    • Sovereign local deployment (Ollama + Qdrant) with no data transfer outside
    • Zero dependency on US hyperscalers for strategic data processing
    • Data exploitation and valorization exclusively by your organization

    Certifications & Alignments

    • Local architecture designed to facilitate HDS compliance (Health Data Hosting)
    • Alignment with the European Health Data Space (EHDS)
    • Compatible with French and European digital sovereignty requirements

    Regulatory Compliance

    • Native GDPR compliance (data minimization, right to erasure, consent)
    • Aligned with NIS2 and DORA requirements (resilience, traceability, risk governance)
    • Relevant ISO standards: ISO 27001 (information security), ISO 27701 (privacy management) and ISO 42001 (AI management systems)
    • Preparation for SOC 2 Type II certification (security, availability, confidentiality and data integrity)
    • Complete traceability of decisions and processing for internal and external audits

    Auditability & Governance

    • Complete timestamped timeline of all steps and decisions
    • Traceable and exportable Kayros Index (KI) scoring
    • AI agent logs + human decisions for audits and risk committees
    • Structured human gates (COMEX veto, expert validation) with traceability
    WHY KAYROSLAB IS UNIQUE

    Not just another chatbot.
    A governed ideation system.

    Classic LLMs excel at text generation. KayrosLab industrializes a complete strategic decision-making process where AI is a controlled, traceable and arbitrated collaborator.

    Criteria ChatGPT / Claude
    Enterprise / Gemini    		 Innovation Platforms
    (IdeaScale, Viima…)    		 KayrosLab
    Process Structure Free linear conversation Idea collection + voting 8-step cyclical process
    Reasoning & Agents Single generic model Limited AI assistance Specialized multi-agents
    (Red Team, Devil's Advocate, Bisociator…)
    Memory & Continuity Limited session context Simple idea database Shared Memory + Vector Memory
    Persistent semantic search
    Robustness & Resilience None (frequent hallucinations) Low Retry + Circuit Breaker + Fallback
    Traceability & Audit Low / ephemeral conversation Basic history Complete timeline + KI scoring + agent logs
    Human Governance Informal (prompt engineering) Simple validation workflows Structured Human-in-the-Loop
    Multi-criteria votes, COMEX veto, Working Groups
    Data Sovereignty Cloud only (dependency) Cloud SaaS generally Local LLM-ready (Ollama)
    P1 sovereign mode: zero data egress
    Deployment SaaS only Cloud SaaS Standalone (1 HTML file) → Local → Cloud
    Primary Use Case Content generation & assistance Collaborative idea box High-level strategic ideation + auditable decisions

    Equip your strategy and innovation teams
    with a next-generation ideation workshop.

    KayrosLab is deployed in 3 tiers adapted to your context. No vendor lock-in: all tiers share the same core.

    P0 — Standalone
    Free

    Complete ideation workshop in a single HTML file.

    • 8-step cyclical workflow
    • 6 specialized agents
    • Mock mode (zero setup, zero cost)
    • Vector memory emulated
    • Local PDF export
    P1 — Sovereign
    On request

    Local deployment on your infrastructure.

    • Ollama LLM + embeddings (local)
    • Qdrant vector memory (local)
    • Zero data egress
    • HDS compliant architecture
    • Full governance: real humans gates
    P2 — Cloud
    Subscription

    Cloud deployment with maximized AI capabilities.

    • Anthropic Claude (API key on server)
    • Consolidated vector memory
    • User authentication + multi-tenant
    • Collaboration: campaigns, comments, moderation
    • Reporting, dashboard, impact tracking
    • Extensibility: Slack/Teams connectors
    Request a personalized demo

    Or contact us directly

    Concrete results your teams can achieve.

    KayrosLab is deployed on real strategic challenges. Each business case illustrates a typical use case by sector.

    Pharmaceutical Industry

    R&D portfolio prioritization

    Analyze 150+ research projects, apply multi-criteria scoring (KI), identify 12 high-potential programs and prepare 4 complete files for the investment committee.

    −60% analysis time
    +85 files processed simultaneously
    3 hidden risks detected (Red Team)
    Banking & Insurance

    Regulatory innovation monitoring

    Deploy a cyclical weak signal monitoring system to anticipate regulatory developments (BCBS 239, DORA, Solvabilité II) and transform them into strategic innovation programs.

    48 signals qualified per month
    3 regulatory watch alerts anticipated
    +22% innovation rate in compliance
    Industry & Construction

    Industry 4.0 technology roadmap

    Build a 3-year strategic roadmap integrating IoT, digital twin, predictive maintenance and ESG tracking. Position competitors, test scenarios and prepare investment files.

    5 technology roadmaps explored
    2 competitor gaps identified (Positioner)
    Budget variance < 8% (Project)
    Retail & e-commerce

    New generation customer experience

    Design and test a generative AI-powered hyper-personalized customer experience concept. From idea generation to business case construction and execution monitoring.

    3 concepts tested in 2 weeks
    −75% ideation-to-business case cost
    ROI 3.4× at 18 months (simulated)

    Whitepapers

    White Paper — Listen: From Weak Signal to Strategic Decision

    How to detect, qualify and amplify weak signals using hybrid AI-human governance.

    Read the white paper →

    White Paper — Position: Ontological Competitive Analysis

    Position your innovation through automated ontological analysis of the competitive landscape.

    Read the white paper →

    KayrosLab — Complete White Paper

    The governed ideation methodology: from weak signal to strategic decision.

    Read the white paper →

    Augmented Hackathon — From Idea to Market in 8 Governed Steps

    How KayrosLab transforms hackathons into structured innovation pipelines: ideation, validation, governance, and accelerated go-to-market.

    Read the white paper →

    Transform weak signals into governed strategic decisions.

    Ready to equip your teams with a next-generation ideation workshop?

    Request a personalized demo

    Or contact us directly:
    contact@kayroslab.com

    Geoffroy de La Tournelle — Founder & Director