Want to dig deeper?
We'll walk you through the engine, the architecture, and what we're building next.
Get the technical briefingA real-time conversational intelligence system that listens, scores, predicts, and intervenes — purpose-built for the hardest conversations between two people.
Most conversation tools rely on a single signal — sentiment, tone, keywords. SKWKY runs on six at once, cross-referenced through four diagnostic layers, in real time, message by message. The mediator, the scoring, and the predictive metrics aren't separate features. They're one engine.
Six tracked dimensions, scored continuously across every exchange. None can be faked with politeness. All have to be earned.
How calmly each person is holding the moment.
Whether feelings are being named, not just acted out.
The sense that you're on the same side of the table.
Whether the conversation is actually going somewhere.
Whether both people are still in the room.
Actively mending after a misstep — what makes hard conversations actually work.
→ These six dimensions feed a single number — the Emotional Readiness Score (ERS) — between 0 and 1. Medium ERS (0.45–0.60) is the normal state of a real conversation. High ERS (≥0.80) has to be earned.
The score tells you what is happening. These four layers tell you what's coming, what kind of conversation you're really in, and what's at risk.
Short-term momentum.
Reads the energy of the last few exchanges to flag when the conversation is shifting — even before the score moves.
Conversation classification.
Identifies the kind of conversation you're in: Calm, Constructive, Repair, Withdrawal, Oscillation, Steady Breakdown — and adapts the engine's reading accordingly.
Predictive forecasting.
Estimates the probability of rupture, fragile repair, and disengagement in real time — so the mediator can step in before, not after.
Structural pattern recognition.
Detects deeper patterns — U-shapes, plateaus, oscillations — that reveal what's really happening over the arc of a conversation, not just at any single moment.
The AI mediator doesn't moderate every message. It watches the engine: score, candles, regime, and risk. It intervenes only when the conversation needs help. Each intervention is fresh, contextual, and directional, quietly steering the conversation toward where it needs to go.
↑ A real conversation flow. Two interventions in ten messages — placed where the engine detected they were needed.
Every score, every classification, every risk estimate is interpretable. We can tell you exactly why a conversation is rated where it is — and what would change it. No black box. No hallucinated emotions. No vibes-based scoring.
"Positive" and "negative" don't capture what happens in real conflict.
You can't fake your way to a high ERS by being nice. The engine knows the difference.
SKWKY is rules-first and deterministic. Every signal is interpretable. Nothing is hallucinated.
SKWKY is a structured conversational space — not a clinical tool, and not a replacement for one.
SKWKY's ERS is built on six tracked dimensions, weighted across structured signals, cross-referenced through four diagnostic layers. Strong moments shift the score meaningfully — but no single phrase distorts the trajectory. High scores have to be earned.
We'll walk you through the engine, the architecture, and what we're building next.
Get the technical briefing