Docs/Developer API/Intelligence Scraper & Data Integrity

Developer API

Intelligence Scraper & Data Integrity

Vision™ ingests signals from your connected integrations and public sources to build your organizational intelligence layer. Here is how the system protects against false, stale, or conflicting information.

How the Intelligence Scraper Works

The scraper pulls structured signals from your connected integrations (Salesforce, Slack, QuickBooks, etc.) and from publicly available sources (company website, LinkedIn, news). Each signal is tagged with a source, confidence score, and timestamp before entering the ingestion pipeline.

Signal ingestion

Raw data arrives from integrations or scraped sources. Each signal records its origin, ingestion time, and raw payload.

Source credibility scoring

First-party integration data (OAuth-connected systems) scores highest. Public web data scores lower and is treated as supplementary context only.

Conflict detection

If a new signal diverges >20% from the existing value for the same metric, it is flagged as a conflict and held for review rather than auto-applied.

Provenance chain

Every data point in Vision™ carries a full provenance chain — you can trace any insight back to the exact signal that produced it.

Staleness expiry

Signals older than their configured TTL are marked stale and excluded from analysis until refreshed.

Protections Against False Information

First-party data priority

Data from your OAuth-connected systems (Salesforce, QuickBooks, etc.) always overrides scraped public data. Your own systems are the source of truth.

Conflict flagging (>20% divergence)

Any incoming signal that contradicts existing data by more than 20% is quarantined and surfaced in the Signals dashboard for human review before being applied.

Provenance tracing

Every insight links back to the exact signals that produced it. You can audit the full data lineage at any time via the API or the admin panel.

AI hallucination guard

The AI layer uses a HallucinationGuard that cross-checks generated insights against your actual org data before surfacing them. Claims without grounding are suppressed.

Staleness expiry

Signals past their TTL are excluded from analysis. Stale data cannot silently influence insights — it must be refreshed first.

Manual override

Any user with org access can flag an insight as inaccurate. Flagged insights are deprioritized and trigger a re-analysis of the underlying signals.

Signal Confidence Tiers

High ConfidenceFull weight in analysis

OAuth-connected integrations (Salesforce, QuickBooks, Slack, etc.)

Medium ConfidenceFull weight, flagged if conflicts arise

Manually entered data, CSV uploads, API-submitted signals

Low ConfidenceSupplementary context only — never overrides first-party data

Public web scraping, LinkedIn, news sources

Accessing Signals via API

You can inspect raw signals, conflicts, and provenance chains via the internal signals endpoint:

# List recent signals for your org

GET /api/companies/{orgId}/signals?type=recent

# View pending conflicts

GET /api/companies/{orgId}/signals?type=conflicts

# Trace provenance for a signal

GET /api/companies/{orgId}/signals?type=provenance&signalId={id}

API Overview Agentic Execution Guide

Newsletter

Stay ahead of the curve

Strategic insights, platform updates, and intelligence on digital transformation — delivered to your inbox.

Weekly insightsNo spam, ever

By subscribing, you agree to our privacy policy. Unsubscribe at any time.

Making the invisible, visible.

Docs/Developer API/Intelligence Scraper & Data Integrity

Developer API

Intelligence Scraper & Data Integrity

How the Intelligence Scraper Works

Signal ingestion

Raw data arrives from integrations or scraped sources. Each signal records its origin, ingestion time, and raw payload.

Source credibility scoring

First-party integration data (OAuth-connected systems) scores highest. Public web data scores lower and is treated as supplementary context only.

Conflict detection

If a new signal diverges >20% from the existing value for the same metric, it is flagged as a conflict and held for review rather than auto-applied.

Provenance chain

Every data point in Vision™ carries a full provenance chain — you can trace any insight back to the exact signal that produced it.

Staleness expiry

Signals older than their configured TTL are marked stale and excluded from analysis until refreshed.

Protections Against False Information

First-party data priority

Data from your OAuth-connected systems (Salesforce, QuickBooks, etc.) always overrides scraped public data. Your own systems are the source of truth.

Conflict flagging (>20% divergence)

Any incoming signal that contradicts existing data by more than 20% is quarantined and surfaced in the Signals dashboard for human review before being applied.

Provenance tracing

Every insight links back to the exact signals that produced it. You can audit the full data lineage at any time via the API or the admin panel.

AI hallucination guard

The AI layer uses a HallucinationGuard that cross-checks generated insights against your actual org data before surfacing them. Claims without grounding are suppressed.

Staleness expiry

Signals past their TTL are excluded from analysis. Stale data cannot silently influence insights — it must be refreshed first.

Manual override

Any user with org access can flag an insight as inaccurate. Flagged insights are deprioritized and trigger a re-analysis of the underlying signals.

Signal Confidence Tiers

High ConfidenceFull weight in analysis

OAuth-connected integrations (Salesforce, QuickBooks, Slack, etc.)

Medium ConfidenceFull weight, flagged if conflicts arise

Manually entered data, CSV uploads, API-submitted signals

Low ConfidenceSupplementary context only — never overrides first-party data

Public web scraping, LinkedIn, news sources

Accessing Signals via API

You can inspect raw signals, conflicts, and provenance chains via the internal signals endpoint:

# List recent signals for your org

GET /api/companies/{orgId}/signals?type=recent

# View pending conflicts

GET /api/companies/{orgId}/signals?type=conflicts

# Trace provenance for a signal

GET /api/companies/{orgId}/signals?type=provenance&signalId={id}

API Overview Agentic Execution Guide

Newsletter

Stay ahead of the curve

Strategic insights, platform updates, and intelligence on digital transformation — delivered to your inbox.

Weekly insightsNo spam, ever

By subscribing, you agree to our privacy policy. Unsubscribe at any time.

Making the invisible, visible.