logo
AgentLed

Scoring & Feedback Loop

A self-improving scoring system where historical outcomes feed back into future runs. Zero manual recalibration.

How It Works

The feedback loop has two paths: a write path that records decisions, and a read path that loads historical context before each scoring run. Together, they create compound learning — every execution makes the next one more accurate.

Write Path

After each scoring run, the workflow writes decision edges to the Knowledge Graph:

Edge TypeMeaningMetadata
SCOREDAI assigned a score to this entityscore, rationale, sector, timestamp
PROCEED_TO_ICHuman approved for investment committeereviewer, notes, timestamp
HOLDHuman decided to hold / not proceedreason, reviewer, timestamp
INVESTEDFinal outcome recordedamount, terms, date

Read Path

Before each scoring run, the AI step runs five parallel queries against the Knowledge Graph:

Scoring History

Previous scores for the same entity. The AI sees how scores changed over time and why.

Human Feedback

PROCEED_TO_IC and HOLD decisions from reviewers. The AI learns which scores led to approvals and which were rejected.

Calibration Data

Statistical benchmarks: average scores by sector, score-to-approval conversion rates, typical ranges.

Pattern Insights

AI-generated observations from previous runs: "fintech investors over-indexed on revenue growth","healthcare VCs prioritize regulatory approval status."

Outcome Benchmarks

Final outcomes (INVESTED, PASSED) compared to initial scores. The AI measures its own prediction accuracy.

Accuracy Over Time

Real results from an investor matching workflow processing 3,000+ profiles:

Run  1:  62% accuracy  (cold start, no history)
Run  3:  71% accuracy  (basic pattern learning)
Run  5:  78% accuracy  (incorporating IC feedback)
Run  8:  84% accuracy  (calibration adjustments)
Run 12:  89% accuracy  (compound learning, zero manual tuning)

Workspace-Scoped Memory

All feedback data is scoped to the workspace. This means:

  • •Institutional knowledge persists across team member changes
  • •A new analyst benefits from all previous scoring rounds immediately
  • •Different workflows sharing the same KG contribute to shared learning

Next Steps

  • KG API — Query scoring edges and history programmatically
  • AI Steps — Configure AI reasoning that uses the feedback loop
  • KG Overview — Full architecture of the Knowledge Graph