Sources and methodology
Where the intelligence comes from.
KAIRO reads across five source categories and synthesizes through the agent workforce. Every output ships with the provenance, the confidence, and the freshness shown.
Public web signals
News, blog posts, social activity, press releases, podcast transcripts, public investor communications. Captured continuously and indexed for the agent workforce.
- Funding announcements and M&A activity
- Leadership changes and executive moves
- Product launches and category shifts
- Public commentary by buyers in your category
Public records
Job postings, regulatory filings, court records, property records, business registries. Structured into account-level signals.
- Hiring posts that show role expansion or restructuring
- Trademark and patent filings
- Property transfers and ownership changes (real estate)
- State and federal regulatory disclosures
Behavioral signals
Account-level intent inferred from public engagement patterns across review sites, community forums, comparison searches, and content interactions.
- Category research that suggests a buyer is in market
- Competitor evaluations on public review sites
- Topic clusters that signal an upcoming RFP
- Engagement spikes that flag the right moment to reach in
Verified contact and signal data
Licensed third-party datasets that add verified contact lines, technographic install events, and signal-grade triggers to public-data lookups. Data is licensed under contracts that respect privacy law in the US and EU.
- Verified email addresses and direct phone lines
- Tech stack install and removal events
- Real-time job change and tenure data
- Funding round and corporate event triggers
Agent-generated synthesis
The agent workforce reads across the sources above and writes the brief, the recommendation, the sequence, or the score. Every synthesis runs through the 5-agent challenge loop before delivery.
- Daily briefings synthesized from twenty signals
- CMAs and seller pitches written from comparable data
- Strategic audits that aggregate competitor moves
- Risk and forecast outputs assembled from pipeline signals
The verification rules
Source provenance is shown
Every claim that depends on a specific source surfaces that source in the output. You can trace any output back to the public signal or licensed dataset that produced it.
Confidence is scored
Outputs ship with a confidence number and an explicit gap callout. The platform tells you what it knows, what it is inferring, and what it does not know.
Freshness is recorded
Every source is tagged with the date it was captured. Outputs declare the freshest signal that drove the synthesis. Stale data is flagged.
Privacy and licensing
Licensed datasets are used only under the contract that allows them. Public web data is captured within the technical and legal limits of public access. Personal data is handled under the policy at /policy.
Hallucination is gated
Agent outputs that cannot be grounded in a captured source are flagged. The 5-agent challenge loop has a fact-check round dedicated to surfacing claims without a source.
Build with the receipts.
Every output ships with confidence, provenance, and freshness. If a source rule does not hold for a specific tool, the platform tells you up front.