🚩 Inspiration

Teams waste hours manually triaging emails, coordinating schedules, and resolving conflicts across tools. Most AI assistants stop at suggestions — they don’t take action.

FlowPilot was built to change that.

🧠 What It Does

FlowPilot is an Active AI Agent built on Airia that transforms incoming emails into structured, prioritized, and executable workflows.

Instead of stopping at classification, FlowPilot:

Extracts intent from scheduling emails Calculates a weighted priority score Detects calendar conflicts Applies approval logic (Human-in-the-Loop when necessary) Executes actions across integrated systems (e.g., Slack, Calendar) Tracks automation metrics and ROI Each decision is transparent and traceable through a detailed AI scoring breakdown.

⚙️ How It Works

FlowPilot uses a multi-agent orchestration model:

Email Input ↓ Parsing Agent ↓ Priority Scoring Agent ↓ Conflict Detection Agent ↓ Approval Agent (HITL if needed) ↓ Execution Agent (Calendar / Slack) 📊 Priority Scoring Model

Priority is calculated using weighted decision factors: Priority Score=(w1​⋅U)+(w2​⋅I)+(w3​⋅D)+(w4​⋅S)

Where: U = Urgency score I = Importance score D = Deadline proximity S = Sender role impact 𝑤n= Configurable weight values The system outputs:

Final score (0–100) Priority level (Low / Medium / High) Transparent reasoning breakdown Confidence indicator

This ensures explainability — a key requirement for enterprise AI systems.

🏗️ How We Built It

FlowPilot was developed using:

Airia for agent orchestration and publishing

A modular backend for decision logic

API integrations for Slack and calendar systems

A React-based dashboard for real-time visibility

Airia acts as the orchestrator layer, managing the agent workflow and tool invocation, while the execution modules handle decision scoring and automation tasks.

The dashboard provides: Decision explainability panel Automation efficiency metrics Time saved calculations ROI estimation System uptime tracking

📈 Measurable Impact

FlowPilot does more than automate — it quantifies efficiency.

It tracks: Emails processed per day Tasks automated Automation rate Time saved Estimated operational value

For example: Estimated ROI=Time Saved×Avg Hourly Coordination Cost This allows teams to see tangible productivity gains from AI automation.

⚡ Challenges We Faced 1️⃣ Balancing Autonomy with Control

Fully autonomous systems can be risky. We implemented Human-in-the-Loop safeguards for high-impact decisions to maintain oversight.

2️⃣ Ensuring Explainability

Many AI systems operate as black boxes. We prioritized transparent scoring to build trust and enterprise readiness.

3️⃣ Multi-System Coordination

Synchronizing actions across email, calendar, and messaging systems required careful workflow orchestration and error handling.

4️⃣ Avoiding Over-Engineering

Hackathon time constraints forced us to focus on core intelligence rather than feature overload. We prioritized clarity, robustness, and measurable impact.

📚 What We Learned

Multi-agent systems are far more powerful than single-response AI assistants.

Explainability is critical for real-world AI adoption.

Automation must be measurable to demonstrate value.

Good architecture matters as much as model capability.

Most importantly, we learned that intelligent orchestration — not just text generation — is the future of enterprise AI.

🚀 Future Vision

FlowPilot can evolve into:

Adaptive weight learning from user behavior

Cross-department workflow expansion

Predictive scheduling optimization

Enterprise-grade analytics dashboards

Fully autonomous coordination across 10+ systems

Our long-term vision is simple:

AI agents that don’t just assist — they operate.

Built With

  • and-executes-tasks-automatically-across-tools-like-google-calendar-and-slack.-modern-teams-waste-hours-triaging-emails
  • decides
  • fastapi
  • pydantic
  • python
  • react
  • scheduling-meetings
  • typescript
  • vite
Share this project:

Updates