SaFA (Sales Flow Assistant)

Inspiration

We were inspired by a simple insight: sales opportunities are often lost not because customers aren’t interested, but because follow-ups are delayed, forgotten, or inconsistent. Today’s CRMs primarily act as data storage systems rather than intelligent partners that guide action. We built SaFA to transform the traditional CRM into an AI-powered assistant that remembers every interaction, understands customer intent, and recommends or executes the next best step. This allows sales teams to focus on building relationships and closing deals instead of manually tracking tasks.

What It Does

SaFA captures sales interactions, analyzes them using AI, and automatically updates the CRM. It evaluates customer sentiment and intent, determines the next best action, and generates personalized follow-ups such as emails or task reminders. By automating these processes and prioritizing high-value opportunities, SaFA ensures that no lead is forgotten and that every opportunity is moved forward efficiently.

Backend / How It Works

Our backend serves as the central orchestration layer of the system. It connects to a cloud-based Supabase database and handles all core API operations, including GET, POST, PUT, and DELETE requests. When data such as lead information, interaction history, or call transcripts is received, the backend processes and structures it before sending relevant context to the Gemini API. The AI then returns intelligent outputs such as recommended follow-ups, personalized cold emails, and next-best actions based on the stage of the sales journey or after a deal is closed. These results are stored in the database and sent back to the frontend as structured responses. The system follows an MVC architecture, ensuring a clean separation of concerns, and is containerized with Docker for flexible deployment across environments.

How We Built It

SaFA was built using a modern full-stack architecture. The frontend was developed with Next.js to create a fast and responsive user interface. The backend was built using C# and the .NET framework, which provided strong support for routing, API integration, and database communication. Supabase was used as the cloud PostgreSQL database to store CRM data, interactions, and AI outputs. The system integrates with the Gemini API for AI-driven recommendations and Calendly for scheduling automation. The application is hosted on Azure, with CI/CD enabled so that updates are automatically deployed whenever new changes are pushed.

Accomplishments We’re Proud Of

We were able to implement AI-driven features that generate personalized cold emails and recommend follow-up actions based on sales context. Additionally, we set up Docker containerization and automated deployment through Azure CI/CD, making the system scalable, portable, and production-ready.

Challenges We Ran Into

A major challenge was working with new technologies and ensuring smooth integration between multiple components, including the .NET backend, Gemini API, Supabase database, and the React-based frontend. Managing data flow between services and structuring prompts to produce reliable AI outputs required careful design and testing. However, overcoming these challenges helped us build a more robust and scalable system.

What’s Next for SaFA

Our next step is to expand SaFA beyond follow-up automation into an AI-powered coaching platform for new sales hires during their first year. After each call, the system will analyze the transcript to evaluate tone, confidence, objection handling, and overall conversation quality. Using ElevenLabs, SaFA will generate personalized voice-based feedback that sounds like a real coach, providing immediate and actionable guidance. The platform will also track each hire’s progress over a one-year timeline, showing performance trends and development milestones. This reduces the burden on managers who currently need to review multiple calls across multiple new hires, while helping teams scale faster and ensuring consistent, high-quality training.