Compound learning effect
The compound learning effect in organisational intelligence
A static dashboard can report what happened. A chat surface can answer a question. A self-improving organisational intelligence layer should do something more valuable: learn from every unknown, correction, recommendation, action, and measured result.
Key takeaways
- Unknowns are useful because they expose missing ownership, policy, documentation, or capability.
- Recommendations should carry expected outcomes and a way to measure whether they worked.
- The system becomes more valuable when measured results change future confidence and playbooks.
Unknowns should become signals
When a user asks who owns a decision and the system cannot answer with evidence, that should not be a dead end. It should create a gap signal: missing owner, weak documentation, unclear policy, or absent capability context.
Recommendations should be measured
A recommendation is only useful if it can be tested. Bridgly frames recommendations as improvement loops with owners, expected outcomes, evidence, and post-action measurement.
- Owner
- Expected outcome
- Evidence
- Post-action verdict
Learning compounds across cycles
As the organisation resolves gaps, accepts or rejects recommendations, and measures outcomes, the graph updates beliefs, confidence, playbooks, and future recommendations. That is the compound learning effect.