IT HelpGenie

IT HelpGenie automates IT helpdesk requests in real-time, directly from employee emails: an AI-powered agent that understands, resolves, and responds to requests, reducing manual effort & saving time.

Comment

🧠 IT HelpGenie: Intelligent IT Helpdesk Automation Agent

💡 Inspiration

We’ve all faced the frustration of waiting days for simple IT requests — whether it’s access to a tool, password resets, or software installations.
During internal discussions with our IT support team, I realized how repetitive most of these requests are — yet they consume hours of manual effort daily.

That’s where HelpGenie was born — an intelligent, autonomous IT helpdesk agent that understands user intent from emails, performs the right actions using tools, and closes tickets automatically — with almost zero human intervention.

⚙️ What it does

HelpGenie acts as a virtual IT assistant that reads and understands employee emails and automatically performs end-to-end IT operations like:

  • 🔐 Handling password resets securely
  • 💻 Managing software access requests
  • 🧾 Automating license validation & procurement
  • 🖥️ Assisting with hardware support
  • 🎟️ Integrating directly with ServiceNow / JIRA ticketing systems

It identifies intent, validates policies (like approvals), performs actions using pre-configured tools (Lambda APIs), and replies to the employee with updates — all autonomously.

🧩 How we built it

  • 🧠 Core Framework: Built using Strands, an open framework for modular AI agents.
  • 🗣️ Model: Powered by Anthropic Claude 3 Sonnet (for reasoning and classification).
  • ☁️ Cloud Runtime: Deployed on AWS AgentCore, enabling stateful execution, memory, and observability.
  • 🛠️ Tools Integration: Implemented tools such as:
    • handle_software_request
    • handle_password_reset
    • get_ticket_status
    • handle_hardware_issue
    • get_user_info
  • 🔗 External Systems: Connected to JIRA and ServiceNow via API endpoints.
  • 🧱 Architecture: Serverless deployment with AWS Lambda, ECR for containerized agents, and S3 for persistent logs.

Mathematically, if we define:

[ T = N_r \times t_h ]

Where ( N_r ) = number of repetitive IT requests per day and ( t_h ) = average human handling time per request (in minutes),
HelpGenie reduces total time ( T ) by an estimated 85–90%, saving hundreds of hours monthly.

🚧 Challenges we ran into

  • 🧩 Integrating external APIs: Ensuring smooth connection to ServiceNow/JIRA with the MCP protocol.
  • 🔄 State management: Designing for continuity in conversations across multiple user sessions.
  • ⚙️ Tool orchestration: Ensuring tools trigger the right Lambda without conflict.
  • 💰 Cost optimization: Evaluating whether smaller models or asynchronous processing could reduce inference costs.
  • 🔒 Security and role-based access: Managing approvals and identity validation for software/hardware requests.

🏆 Accomplishments that we’re proud of

  • Built an end-to-end intelligent IT helpdesk agent powered by AWS Bedrock and Anthropic models.
  • Successfully simulated complete email-to-resolution automation for IT requests.
  • Created a modular, scalable, and deployable agent framework that integrates with enterprise APIs.
  • Demonstrated the potential to cut IT resolution time by 70% and save thousands of dollars annually.

📚 What we learned

  • How to build production-grade agents using AWS AgentCore and ECR deployments.
  • Best practices for managing serverless workflows with Lambda.
  • Real-world challenges in automating enterprise IT operations.
  • How small context windows can be optimized with memory-based workflows.

🚀 What's next for IT HelpGenie

  • 🔄 Integrate with AWS Kendra for contextual knowledge retrieval.
  • 🧠 Enable memory-based conversations with long-term context.
  • 💬 Add multi-channel support — Slack, Teams, and web chat.
  • 📊 Build analytics dashboards for IT performance tracking.
  • 💡 Explore multi-agent collaboration, where specialized agents handle procurement, access, and escalation separately.

Built With

Share this project:

Updates