# Move from source evidence to engineer-confirmed findings\.

- Page ID: `ai-analysis`
- Canonical human URL: https://my-platform-site.kyosls.workers.dev/en/ai-analysis/
- Content revision: 2026-09-02
- Intended audience: Broker platform owners, operations leaders and MT4/MT5 engineers.
  - Source: https://my-platform-site.kyosls.workers.dev/en/#ownership

## Summary

The pipeline normalises and correlates events across agreed sources, then presents cited candidates for engineer review; it does not autonomously determine root cause.

Source: https://my-platform-site.kyosls.workers.dev/en/ai-analysis/#overview

## Evidence pipeline

Events are normalised to a common timezone and timestamp format while preserving source-clock metadata and detected clock skew.

### Collect

Ingest only the sources, time range and location agreed before collection.

### Classify

Parse source lines into versioned event schemas without discarding original evidence.

### Correlate

Align related events across journals, clients, gateways and infrastructure sources.

### Review

Present candidates and cited evidence to an engineer, who confirms or rejects each finding.

Source: https://my-platform-site.kyosls.workers.dev/en/ai-analysis/#pipeline

## Potential sources

The minimum necessary source set is agreed before collection.

- MT4/MT5 server journals and client logs
- Crash dumps, Windows fault events, gateway and bridge output
- Host, container and cluster events

Source: https://my-platform-site.kyosls.workers.dev/en/ai-analysis/#sources

## Feasibility sample

A representative sample is often sufficient for an initial feasibility review. We confirm the required time range and sources before collection.

Source: https://my-platform-site.kyosls.workers.dev/en/ai-analysis/#sample

## Deterministic evaluation

Evaluation records precision, recall, false-positive rate, false-negative review, evidence coverage, reviewer disagreement, parser version, prompt version and model version.

Source: https://my-platform-site.kyosls.workers.dev/en/ai-analysis/#evaluation

## Deployment follows the client's data policy

Deployment location and model or provider are selected against the client's data policy and operated through an agreed change process.

Source: https://my-platform-site.kyosls.workers.dev/en/ai-analysis/#deployment

## Operational boundaries

AI assists evidence review; it does not operate the trading platform.

- AI proposes or summarises evidence; an engineer confirms findings.
- No trading signals, strategy design or trading-logic design.
- No autonomous production change or autonomous root-cause determination.
- Sources, retention, deployment location and model/provider are agreed before collection.

Source: https://my-platform-site.kyosls.workers.dev/en/ai-analysis/#boundaries

## Verified facts

- [clock-normalisation] Events are normalised to a common timezone and timestamp format while preserving source-clock metadata and detected clock skew.
  - Source: https://my-platform-site.kyosls.workers.dev/en/ai-analysis/#pipeline
- [sample-qualified] A representative sample is often sufficient for an initial feasibility review; the required time range and sources are confirmed before collection.
  - Source: https://my-platform-site.kyosls.workers.dev/en/ai-analysis/#sample
- [human-confirmation] AI proposes or summarises evidence; an engineer confirms findings.
  - Source: https://my-platform-site.kyosls.workers.dev/en/ai-analysis/#boundaries
- [ai-exclusions] AI analysis provides no trading signals or strategy design and makes no autonomous production change or root-cause determination.
  - Source: https://my-platform-site.kyosls.workers.dev/en/ai-analysis/#boundaries

Contact route: https://my-platform-site.kyosls.workers.dev/en/contact/