The Programmable
Reasoning Layer.
Turn unstructured narrative into structured data. Define any question. Our infrastructure scans millions of filings, news articles, and web pages to compute the answer—cited, grounded, and hallucination-free.
{
"question": "Is this company expanding into Asia?"
}
// Response
{
"verdict": "YES",
"confidence": 0.95,
"evidence": "2023 Annual Report, p.42..."▋
}Don't build the RAG. Just query the Context.
Building a specialized RAG pipeline for corporate data is painful. You have to scrape the web, parse PDFs, chunk the text, embed vectors, and manage context windows.
Zint has already done the heavy lifting. We have indexed the entire UK corporate corpus. Magic Insights allows you to execute 'reasoning jobs' on top of this data. You define the schema (e.g., 'Is this company expanding into Asia?'); we retrieve the relevant context, run the model, and return a binary verdict with evidence.
The "Reasoning" Pipeline
Define the Signal
You don't just ask a question; you define a Data Point you want to exist.
"esg_commitment",
"Does the company utilize renewable energy?"
)
The Context Hunt
Zint's engine automatically queries our Contextual Intelligence layer.
The Computed Verdict
The model synthesizes the findings into a structured JSON response.
"verdict": "YES",
"confidence": 0.95
}
Features & Capabilities
Infinite Schema
Stop waiting for data providers to add a column. If you can phrase it as a question, you can query it. Create your own proprietary risk scores or sales triggers instantly.
Citation, Not Hallucination
Every insight comes with a source_document_id and a text_snippet. If the answer is "Yes," we show you exactly where they said it. If the data doesn't exist, we return null, not a guess.
Batch Compute
Run your custom insights across 10,000 companies at once. "Find me every logistics company in the UK that mentions 'Brexit Supply Chain Issues' in their latest filing."
Extraction as Code.
Define insights programmatically and get structured, cited responses.
Request
{
"entity_id": "09876543",
"insight_definition": {
"key": "export_exposure_eu",
"question": "Does this company export goods to the EU?",
"required_evidence_strength": "HIGH",
"context_scope": [
"filings",
"news_last_12m"
]
}
}Response (200 OK)
{
"key": "export_exposure_eu",
"verdict": "YES",
"confidence_score": 0.98,
"summary": "The company explicitly mentions trade friction with EU partners in their strategic report.",
"citations": [
{
"source": "annual_report_2023.pdf",
"section": "Principal Risks",
"snippet": "...delays in customs clearance for goods moving to France and Germany..."
}
]
}What to Build
Risk & Compliance
Question:
"Is there any evidence of operations in sanctioned jurisdictions?"
Result:
Flags potential AML risks instantly.
Sales Intelligence
Question:
"Is the company hiring engineers for React or Python?"
Result:
Identifies tech stack migration before it's public knowledge.
Private Equity Sourcing
Question:
"Does the strategic report mention 'succession planning' or 'family ownership'?"
Result:
Spotting potential exit-ready founders.
Premium Compute.
Computationally Expensive
Magic Insights are computationally expensive operations that require advanced reasoning and context retrieval.
10 Credits per Insight
Each Magic Insight request costs 10 Credits.
Smart Caching
We cache results. If you ask the same question for the same company within 30 days, it costs only 1 Credit (Cache Hit).