PRISM AI

Agentic collaboration like never before

Comment

Inspiration

Organisations often face challenges with project coordination, fairness, and knowledge continuity. Work gets duplicated across teams, deadlines slip due to poor dependency tracking, and leadership changes disrupt progress.
We wanted to create an AI-driven multi-agent system that acts like an organisational nervous system - ensuring smooth workflows, fair recognition, and strategic foresight.

What it does

PRISM AI is a multi-agent ecosystem powered by a knowledge graph and temporal automata engine.

  • Tracks projects, teams, milestones, and deliverables in real time.
  • Enforces temporal and logical rules (e.g., Milestone B must follow Milestone A within 10 days).
  • Detects conflicts like overlapping responsibilities or fairness imbalances.
  • Suggests leadership-level actions for risk mitigation and future planning.
  • Integrates with tools like Slack, Jira, and Confluence.

How we built it

  1. Knowledge Graph (Neo4j) to model entities (Projects, Teams, Milestones, Deliverables, Risks).
  2. Temporal Automata + Model Checking using Büchi automata to enforce Linear Temporal Logic (LTL) rules.
    • Example:
      LTL property: Always (AlphaComplete → Eventually BetaStart)
      Meaning: Whenever AlphaComplete happens, it will always be followed at some point by BetaStart.
  3. Multi-Agent Architecture where each agent has a role (Lifecycle, Coordinator, Fairness, Compliance, Leadership).
  4. Integration Layer with Slack/Jira/Confluence to feed real-world data into the system.
  5. Event-Driven Communication (Kafka/Redis streams) to let agents interact without central bottlenecks.

Challenges we ran into

  • Encoding real-world organizational rules into temporal logic & automata.
  • Preventing infinite feedback loops between agents.
  • Handling messy, inconsistent data from Slack/Jira integrations.
  • Balancing ML-driven insights with strict rule-based reasoning.

Accomplishments that we're proud of

  • Designed a scalable agent architecture that can expand with organizational needs.
  • Successfully modeled temporal constraints in real-world workflows.
  • Built a prototype that integrates with Slack and provides actionable AI-driven recommendations.
  • Achieved a system where multiple agents collaborate rather than act in isolation.

What we learned

  • How to apply temporal logic and model checking for real-world project workflows.
  • The power of knowledge graphs for unifying data across tools.
  • Designing agent interaction protocols to avoid deadlocks and redundancies.
  • Practical integration of symbolic AI + machine learning in one ecosystem.

What's next for PRISM AI

  • Expand integrations (e.g., Trello, GitHub, Notion).
  • Add predictive modeling for project risks using ML.
  • Introduce explainable fairness metrics for transparent recognition.
  • Deploy as a SaaS platform for organizations to plug-and-play.
  • Scale agent reasoning using Watsonx.ai / LLMs for richer contextual insights.

Built With

Share this project:

Updates

Submission history