The market in one read
Unstructured data in insurance and AI meet at a single number: corporate teams lose 20-30% of their time locating, organizing, and re-entering unstructured data instead of acting on it, according to Gartner. In Brazilian Seguros e Danos (P&C), that lost time lands squarely on the underwriter. Commercial submissions arrive as e-mail bodies, scanned PDFs, spreadsheets in inconsistent formats, broker (corretor) cover notes, property schedules, and photographs, and someone has to turn that mess into structured fields before any risk can be assessed. Intelligent document reading attacks exactly this cost: it reads the raw inputs and extracts the underwriting-relevant fields automatically, with high precision.
State of the P&C insurance market
Brazil runs one of the largest insurance markets in Latin America, and Seguros e Danos (P&C) is among its fastest-growing lines. The segment grows double digits per year, while the company structure built to process it does not keep pace. That mismatch is the operational story underneath the headline growth.
The market is supervised by SUSEP (Superintendência de Seguros Privados), and aggregate sector figures are published by CNseg and its P&C federation FenSeg. Readers who need exact premium or loss-ratio numbers should treat the SUSEP statistics portal as the canonical source, with CNseg and FenSeg bulletins for sector-level context. What matters for this topic is not a single premium figure but the shape of the flow: rising volume of commercial and corporate risk, almost none of it arriving in clean, structured form. Property, automotive, rural, and corporate lines all feed submissions that an underwriter (subscrição, underwriting) must reconstruct before pricing the premium (prêmio).
What is pressuring underwriting
Start with the time itself. Deloitte found that underwriters spend 40% of their time on administrative tasks rather than risk judgment, and unstructured intake is a large part of that load. A single commercial risk can span an e-mail body, several PDF attachments, a property schedule in Excel, and broker annotations, with no standardized submission format across the corretor channel. The underwriter rebuilds a coherent risk picture by hand before assessing anything.
Manual rekeying compounds it. Because intake is unstructured, fields such as insured value, location, occupancy, construction type, deductible, claims history, and coverage requested are transcribed by hand into quotation and policy systems. Every transcription is a point where data quality degrades and where time leaks away from actual underwriting.
Distribution turns that delay into lost business. Capgemini reports that 60%+ of brokers choose an insurer by response speed, so a slow quote is a lost quote. When underwriters are buried in organizing submissions, time to quote rises and the insurer cedes business at the distribution layer.
There is also a structural reason this rarely gets fixed at the source. BCG found that 70% of insurers do not execute innovation because of IT limitations. Many run legacy core and policy systems that are expensive and risky to modify, so the appetite is to automate the front of the funnel, intake, reading, structuring, and scoring, without a core migration. That is precisely the demand an external intelligence layer answers.
Risk, fraud, and the AI shift
The shift underway is from manual organization of submissions to AI-driven intelligent document reading that converts raw inputs into structured fields automatically. Machine Learning and document-understanding models read typed and scanned PDFs, tables, free text, and broker notes, then extract the underwriting-relevant data with high precision. Each extraction carries a confidence score, so low-confidence fields are flagged for human review rather than silently passed through.
Once data is structured, risk scoring can run against the insurer's risk appetite and underwriting manual. Models flag risks inside appetite for fast-track and route out-of-appetite or complex cases to a human underwriter, applying the same policy to every submission instead of leaning on each underwriter's interpretation under time pressure. Structured inputs are also the precondition for any data-driven pricing of the premium, because dynamic pricing fed on noisy data only produces noisy prices.
Fraud and quality dynamics change too. Manual rekeying introduces transcription errors that propagate into pricing and reserving, while automated extraction with confidence scoring reduces silent errors and creates a consistent, auditable trail from source document to structured field. Structuring submissions also makes anomalies detectable at scale: inconsistent values across documents, mismatched locations, duplicated or altered loss histories, and document-level signs of tampering that are impractical to catch by hand.
This is where the regulatory frame bites. Insurance submissions carry personal and sensitive data, so the LGPD (Lei Geral de Proteção de Dados, supervised by the ANPD) governs how that data is processed, including automated processing. For automated underwriting decisions, the expectation is auditability and explainability: the insurer should be able to show how a decision was reached and which data drove it. An AI layer that logs source document, extracted field, confidence, and decision rationale supports both LGPD accountability and SUSEP-aligned supervision.
Where WIR fits
WIR is the AI layer of insurance: an external AI intelligence layer that sits on top of the systems an insurer already runs. For unstructured data, the relevant capability is intelligent document reading. WIR ingests submissions in the format the insurer already receives, e-mail, attachments, and API, then reads the documents and extracts the underwriting-relevant fields automatically, with high precision. The extracted data is normalized into the insurer's own schema, with confidence scoring so weak extractions surface for review. The AI layer for insurance. On top of the systems the insurer already runs, never in their place.
From there, two modules carry the work forward. Underwriter Intelligence automates the quotation journey per the insurer's risk-acceptance policy, with real-time Machine Learning scoring calibrated to the risk appetite and underwriting manual, automatic routing by appetite and exposure, and predictive conversion analysis by product, risk, and broker. Smart Sales handles distribution intelligence, mapping the portfolio by client and product and scoring next-best-action so penetration and retention grow together. WIR is not an insurer, broker, or MGA, and it carries no risk. It does not replace the core. Every decision is explainable and returns a full audit trail, with data encrypted at every step and LGPD compliant.
On traction, WIR keeps the claim narrow. The one public fact is a POC in execution with a global insurer in the Transport line. The point of the architecture is that the intelligence runs as a layer on top of existing systems, which is why it can be adopted at the intake and scoring layer without a costly, high-risk core migration.
Outlook
Intelligent document reading is moving from experimental pilots toward a standard front-of-funnel capability. The value, recovered underwriter time, faster quotes, and cleaner data, is measurable, and the integration risk stays low when the intelligence sits on top of the core rather than inside it. As long as P&C volumes grow at double digits while underwriting headcount does not, the operational gap that automation closes will keep widening, which sustains demand for AI-driven structuring.
Distribution pressure points the same way. Broker expectations on response speed are not easing, so insurers that quote faster and more consistently will win share, and fast, accurate intake automation becomes a competitive lever rather than a back-office nicety. The dominant pattern will be layered, not a full core rebuild, because core migrations remain slow, costly, and risky. Governance maturity completes the picture: as automated underwriting scales, LGPD accountability and SUSEP-aligned auditability and explainability move from afterthought to procurement requirement, favoring solutions that capture a clean audit trail from source document to decision. None of this is certain for any single insurer, but the direction of the Brazilian P&C market is clear enough to plan against.
Frequently asked questions
Why does unstructured data stall underwriting?
Unstructured data stalls underwriting because the underwriter must manually rebuild a coherent risk picture before any risk can be assessed. A single commercial submission spans e-mail bodies, scanned PDFs, spreadsheets, and broker (corretor) notes with no standard format. Fields like insured value, occupancy, and claims history are rekeyed by hand, where data quality degrades and time leaks. Deloitte found underwriters spend 40% of their time on administrative tasks rather than risk judgment.
How much time do companies lose organizing unstructured data?
Corporate teams lose 20-30% of their time locating, organizing, and re-entering unstructured data instead of acting on it, according to Gartner. In Brazilian Seguros e Danos (P&C), that lost time lands on the underwriter, who reconstructs each submission before pricing the premium (prêmio). Slow intake also costs business: Capgemini reports 60%+ of brokers choose an insurer by response speed, so a slow quote is a lost quote.
How does AI turn quotation e-mails and PDFs into structured fields?
AI reads quotation e-mails and PDFs through intelligent document reading, extracting underwriting-relevant fields automatically with high precision. Machine Learning and document-understanding models process typed and scanned PDFs, tables, free text, and broker notes, then normalize the output into the insurer's own schema. Each extraction carries a confidence score, so low-confidence fields are flagged for human review rather than silently passed through into quotation and policy systems.
Is intelligent submission reading auditable?
Yes, intelligent submission reading is auditable. WIR's AI layer logs the source document, the extracted field, the confidence score, and the decision rationale, creating a consistent trail from raw input to structured field. Every decision is explainable and returns a full audit trail, with data encrypted at every step and LGPD compliant. This supports both LGPD accountability and SUSEP-aligned supervision of automated underwriting decisions.
Does WIR replace the core to structure the data?
No, WIR does not replace the core to structure the data. WIR is the external AI layer of insurance, 100% external, sitting on top of the systems the insurer already runs, with no core migration. It ingests submissions in the format the insurer already receives, reads and extracts fields, and writes structured data back to existing systems. WIR is not an insurer, broker, or MGA, and carries no risk. Its one public traction is a POC with a global insurer in the Transport line.