🧠 Helix AI — Decision Intelligence for Teams

Inspiration

Every day, developers, students, founders, and teams face important decisions—choosing technologies, planning products, prioritizing features, or evaluating trade-offs. While AI chatbots can answer questions, they rarely provide structured decision analysis or remember previous decisions.

We wanted to build an AI assistant that doesn't just answer questions, but helps people make better decisions. This led to the idea of Helix AI, a Slack-native Decision Intelligence Assistant that analyzes decisions, presents structured recommendations, and keeps a history of previous analyses for future reference.

What it does

Helix AI is a Slack application that provides structured AI-powered decision analysis directly inside Slack.

Users can ask questions such as:

Should I build a mobile app first?
Should I use PostgreSQL or MongoDB?
Should I buy a MacBook Air M4?

Helix responds with a structured decision report containing:

✅ Recommendation
⭐ Confidence Score
👍 Pros
👎 Cons
⚠ Risks
💰 Estimated Cost
⏳ Estimated Timeline
➡ Actionable Next Steps

Every analysis is automatically stored, allowing users to review previous decisions using the history command.

How we built it

Helix AI was built using a modular Node.js architecture with Slack Bolt for Slack integration.

The application consists of:

Slack Bolt for event handling
OpenRouter API for AI-powered decision analysis
Prompt engineering to enforce structured JSON responses
A modular decision engine for processing AI output
Slack Block Kit for rich, user-friendly responses
Local JSON storage for persistent decision memory
Environment-based configuration using dotenv

The project follows a modular architecture by separating commands, services, prompts, utilities, and storage into dedicated modules, making the application easier to maintain and extend.

Challenges we ran into

One of the biggest challenges was ensuring reliable AI responses.

Initially, responses were inconsistent, making them difficult to parse programmatically. We improved prompt engineering, added response validation, and built fallback logic to handle malformed AI output gracefully.

Another challenge was selecting an AI provider. During development we migrated from Google's Gemini API to OpenRouter, which provided greater flexibility in model selection and simplified integration.

We also focused on improving the Slack user experience by replacing plain text responses with structured Slack Block Kit cards.

Accomplishments that we're proud of

We're proud that Helix evolved from a simple Slack bot into a complete Decision Intelligence Assistant.

Key accomplishments include:

Slack-native AI experience
Structured decision analysis
Persistent decision history
Modular project architecture
Professional Slack Block Kit interface
Robust AI response validation
Clean GitHub documentation and project structure

What we learned

Building Helix reinforced the importance of designing AI applications around reliability rather than simply generating text.

We learned how prompt engineering directly impacts application stability, how to build resilient parsing pipelines, and how modular software architecture makes AI-powered applications easier to extend and maintain.

We also gained valuable experience integrating Slack applications with external AI services while designing a user experience that feels native to the Slack ecosystem.