CoffeechAI

n8n workflow
Home Page
Mentor match results
Auto message creation
FAQ auto answer

Inspiration

The idea for CoffeechAI came from a real personal experience. As an intern thrown into a data analytics role, I had no clear understanding of how the company’s databases were structured or who to ask for help. Documentation was scattered, Slack felt intimidating, and I didn’t want to bother the “wrong” people. That friction slowed learning and made onboarding far harder than it needed to be. CoffeechAI was born to make asking for help easier, more human, and more discoverable.

What it does

CoffeechAI helps employees get unstuck by connecting them to the right knowledge or the right person—fast.
When a user asks a question:

CoffeechAI first checks if it can be confidently answered using existing internal FAQs.
If not, it intelligently recommends coworkers who are best suited to help, based on skills, experience, timezone, and availability.
The focus is on people connection—making it easy to meet and learn from others within the company, not just get an answer.

How we built it

We built CoffeechAI as a hybrid AI system with a strong workflow backbone:

n8n orchestrates the entire flow, handling routing, branching, and integrations.
A keyword-based scoring layer evaluates whether a question can be answered via FAQs with sufficient confidence.
When FAQs fall short, an AI ranking step evaluates mentor candidates using structured employee data such as skills, past projects, seniority, location, and availability.
Carefully designed AI prompts generate consistent, structured outputs and unique, personalized reasons for each mentor recommendation.
Webhook-based APIs keep the system modular and extensible.
Lovable was used to build a modern, sleek, and intuitive frontend that makes the experience feel approachable and frictionless.

The system prioritizes deterministic logic first, escalating to AI only when human help is genuinely needed.

We explicitly used tools like n8n, Keywords AI, and Lovable, as they allowed us to move fast while still building something production-minded.

Challenges we ran into

The most hackathon-worthy challenge was setting up and mastering n8n under time pressure. Specific pain points included:

Learning how n8n handles execution, item flow, and branching
Correctly formatting and passing data between nodes
Integrating Keywords AI via HTTP requests
Debugging duplicate executions and unexpected workflow behavior
Designing prompts that consistently returned valid, structured JSON

Balancing deterministic rules with AI flexibility required a lot of iteration and careful debugging.

Accomplishments that we're proud of

Built a fully working end-to-end system, not just a prototype
Achieved reliable FAQ confidence scoring before escalating to mentors
Generated unique, personalized mentor reasons instead of generic matches
Integrated real availability so mentors are actually reachable
Designed a system that scales across teams and companies
Delivered a sleek, intuitive UI that makes asking for help feel natural

We’re especially proud of the architectural cleanliness combined with a polished user experience.

What we learned

We learned that:

Hybrid systems (rules + AI) are far more reliable than AI alone
Clear data contracts and strict JSON outputs are essential for AI pipelines
Workflow observability matters—small execution quirks can cascade quickly
Prompts need to be treated like code: designed, tested, and iterated
Even the best AI systems fail without solid underlying logic

This project reinforced that great AI products require both strong engineering and thoughtful human-centered design.

What's next for CoffeechAI

Our biggest excitement is bringing CoffeechAI to real companies. Next steps include:

Deep Slack, Teams, and email integrations
Calendar invites and automated scheduling
Feedback loops to continuously improve mentor recommendations
Personalization based on past interactions
Admin dashboards to surface knowledge gaps and mentorship demand

Ultimately, we see CoffeechAI as a company-wide layer that helps people connect, learn faster, and feel supported—no matter where they’re starting from.