Invisible Feedback Engine

About the Project

Surveys are broken.

Despite the value of user feedback, traditional surveys interrupt users, demand effort, and create a mental barrier. SurveyMonkey’s challenge highlights this exact problem: people don’t enjoy stopping what they’re doing to answer questions, which results in low response rates and less authentic insights.

We asked ourselves a simple question:

What if feedback didn’t feel like feedback at all?

That question inspired Invisible Feedback Engine — a system that captures user sentiment passively and continuously by observing natural digital behavior instead of asking explicit survey questions.

Inspiration

Our inspiration came from two places:

Survey fatigue — We’ve all ignored surveys, rushed through them, or answered dishonestly just to finish. That friction fundamentally limits the quality of feedback companies receive.
Event-driven AI systems — With the rise of real-time analytics and agentic AI, we saw an opportunity to rethink feedback collection as a stream of behavioral signals rather than a form to be filled out.

By combining these ideas, we envisioned a future where feedback is invisible, real-time, and adaptive, rather than explicit and disruptive.

What the Project Does

Invisible Feedback Engine replaces surveys with behavioral observation.

Instead of asking users questions, we analyze:

What content they engage with
How long they spend on certain pages
Where they hesitate, scroll repeatedly, or show frustration
Which features attract attention vs. confusion

These interactions are converted into events and processed by a multi-agent AI system that continuously generates insights about user sentiment, preferences, and UX issues.

The output is a live, dashboard-ready stream of insights — without ever interrupting the user.

How We Built It

Architecture Overview

At the core of our system is Solace Agent Mesh, which enables an event-driven, distributed AI architecture.

Each user interaction is published as an event into Solace, and multiple independent agents consume and process those events:

Emotion Agent
Detects emotional patterns such as frustration, confusion, or disengagement based on interaction behavior.
Context Agent
Interprets where and when interactions happen to understand situational meaning.
UX Friction Agent
Identifies problematic patterns like repeated clicks, stalled navigation, or abandonment signals.
Insight Agent
Aggregates agent outputs into human-readable, dashboard-ready insights.

All agents communicate only through events, not direct calls, which allows the system to scale naturally and remain resilient.

Why Solace Agent Mesh

Solace is not just a message broker in our project — it’s the foundation of the intelligence.

Using Solace Agent Mesh allowed us to:

Build a true multi-agent AI system
Decouple agents so they evolve independently
Process feedback in real time
Scale effortlessly as more sites, users, or agents are added

This architecture reflects the future of AI systems: collaborative, event-driven, and distributed.

What We Learned

Through this project, we learned:

How event-driven architectures enable real-time intelligence
How to design agentic AI systems where responsibility is distributed
The importance of decoupling logic to avoid brittle pipelines
How real-world feedback problems require both technical and UX innovation

Most importantly, we learned that meaningful feedback doesn’t have to be asked for — it can be observed.

Challenges We Faced

Real-time event debugging — Working with live event streams required careful logging and tracing.
Agent coordination — Designing agents that remain independent yet produce cohesive insights was non-trivial.
Time constraints — Building a distributed backend under hackathon pressure required strict prioritization.
Avoiding assumptions — We had to be careful not to oversimplify user behavior and instead leave room for future interpretation layers.

Each challenge pushed us to design a cleaner, more scalable system.

The Future

Invisible Feedback Engine opens the door to a new kind of feedback platform:

Deeper sentiment analysis powered by LLMs
Long-term behavioral trend detection
Cross-site analytics for companies with multiple products
Seamless integration with SurveyMonkey’s analytics ecosystem

Our long-term vision is a world where feedback is continuous, invisible, and honest — powered by intelligent agents communicating through events.

Final Thoughts

Invisible Feedback Engine represents a shift away from forms and toward understanding.

By combining SurveyMonkey’s vision for frictionless feedback with Solace’s Agent Mesh, we demonstrate how real-world problems can be solved using event-driven, multi-agent AI systems.

Feedback doesn’t need to be asked.

It just needs to be listened to.