What intelligent document reading with an AI layer means
Intelligent document reading for insurance submissions is Machine Learning based extraction that reads a submissão (submission) wherever it arrives, an e-mail body, a PDF proposal form, a scanned certificate, a spreadsheet, or a broker cover note, and returns the fields an underwriter needs as clean, validated, structured data. It is the first stage of the automated quotation and underwriting (subscrição) journey, and it belongs to any P&C (Seguros e Danos) insurer whose intake still depends on people reading and rekeying documents by hand. The reader who should care most is the underwriting lead or innovation head tired of slow, inconsistent intake that burns underwriting capacity on data entry.
The problem is concrete. Before anyone can price a risk, a person has to find the relevant values across heterogeneous documents, key them into the core or the rating engine, and reconcile contradictions between files. Gartner has estimated that corporate teams lose 20 to 30 percent of their time organizing unstructured data rather than doing analytical work, and intake is exactly where that loss lands in insurance. WIR is an external AI layer that automates this reading on top of the systems the insurer already runs, so the submission reaches the underwriting desk already structured and triaged. The intelligence is calibrated to the insurer's own risk appetite and underwriting manual, never a generic black box.
How automatic data extraction and validation works
Most insurers have tried to fix intake before, usually with traditional OCR and RPA, and both under-delivered for the same reason. Classic OCR turns pixels into characters but does not understand the document. It is template bound, so it works only when a field sits in the same place on the same form every time, which is the opposite of P&C reality where every broker, every line of business (ramo), and every insured sends a different layout. RPA scripts the clicks a human would make, so it breaks the moment a portal, a form, or an upstream field changes. Both treated reading as mechanical transcription when the real task is interpretation.
Intelligent document reading asks what a value means rather than which character sits in a pixel region. It identifies the insured's CNPJ, the sum insured (importância segurada), the requested coverage, the vigência, and the risk address wherever they appear, across formats it has never seen in that exact layout. Three properties separate it from the older tools. First, layout independence, since models trained on insurance documents generalize across forms, e-mails, certificates, and spreadsheets, so a new broker template does not need a new rule. Second, a confidence score per field, so high-confidence values flow straight through while low-confidence ones route to a person for a quick check. Third, validation, where each value is checked for completeness against the requirements of that ramo, for format on fields like CNPJ, CEP, dates, and currency, and for consistency across documents.
The intake is also multichannel, because a real insurer receives submissions by e-mail, broker portal upload, and partner API at the same time. WIR reads the message and every attachment as one submission, validates the extracted data, and flags missing or conflicting items for broker enrichment before pricing. Only the genuinely ambiguous fields reach a human, which inverts the OCR model where people checked everything. The output is not raw text. It is a structured, validated submission object that the rest of the journey can consume, and clean intake is what makes the downstream stages, broker enrichment, risk and fraud scoring, dynamic pricing, and the final decision, fast and consistent.
How to deploy intelligent reading as an external layer
Deployment does not require a core migration, and that is the central architectural point for a Brazilian insurer. The AI layer sits on top of existing systems. It ingests submissions from the channels above, produces structured validated data, and pushes that data into the policy core, the underwriting workbench, or the rating engine through APIs or files. The system of record stays intact. This matters because IT limitations are one of the most cited blockers to insurance innovation. BCG has found that 70% of insurers do not execute innovation because of IT constraints, and an overlay model lets an insurer modernize intake without betting the company on a multi-year core program.
The implementation runs as a defined project, not an open-ended IT effort the insurer's team has to staff. Setup is a one-time phase of 3 to 12 months that covers scope, integration with the existing core, calibration to the underwriting manual and risk appetite, testing, and go-live adjustments, with a fixed price, a clear scope, and KPIs agreed before the work starts. After go-live the layer moves to continuous operation in production, with a billing model adjusted per client. Throughout, WIR remains 100% external, with no load on the insurer's IT and no replacement of the core. The reason to automate intake first is commercial as much as technical. Capgemini reports that 60%+ of brokers (corretores) choose an insurer by response speed, and slow manual intake directly costs conversion and shelf space with distribution.
Governance, explainability, and LGPD
Automating intake touches personal and corporate data, so governance is part of the design, not an afterthought. Under the LGPD (Lei Geral de Proteção de Dados, Lei 13.709/2018), submission data often contains personal data such as names, CPF, and addresses, so the layer must process it on a lawful basis, with data minimization and security, and must keep a human in the loop where an automated decision affects a person. Data is encrypted at every step and processed in line with LGPD, and the ANPD is the supervisory authority. SUSEP supervises the P&C market, so automated reading and downstream underwriting must stay consistent with the registered product terms and the underwriting manual.
Explainability is where the confidence score stops being only an accuracy feature and becomes a governance mechanism. Every extraction, every confidence level, every validation outcome, and every human override is logged, so the insurer can prove which values were machine-extracted with high precision and which were checked by a person. Decisions are explainable and return a full audit trail, never presented as infallible. That provenance is what makes automated intake defensible to internal audit and to the regulator, because each field's certainty and origin are recorded rather than hidden. For the wider market backdrop the insurer can review the WIR insurance market intelligence coverage of where automation moves the needle in Brazilian P&C.
How WIR automates submission reading
WIR is the AI layer for insurance. On top of the systems the insurer already runs, never in their place. It automates the quotation and underwriting journey according to the insurer's own risk-acceptance policy, with Machine Learning calibrated to the risk appetite and the underwriting manual. Intelligent document reading is the second stage of a six-stage flow that begins with multichannel intake and automatic validation, then reads the documents, then enriches with broker context and scoring, then runs a multi-factor risk and fraud engine, then prices the risk, and finally returns a decision, a quote, an automatic decline, or escalation to a human, always with an explanation written back to the policy core alongside the audit trail.
Two modules carry the work. Underwriter Intelligence automates the quotation journey per the insurer's risk policy, with real-time ML scoring, automatic routing by appetite and exposure, and predictive conversion analysis by product, risk, and broker, so underwriters spend their time on risk judgment and business development rather than rekeying. Smart Sales maps the portfolio across client and product, scores upsell and next-best-action, and runs multi-channel campaigns with an attribution trail. Real-time dashboards give a proactive view of in-flight deals and pipeline. WIR Innovation was founded in 2025 from accumulated operational experience, built with Mahway, a Venture Builder in California, and Avante, a Venture Studio in Brazil. Its first POC is in execution with a global insurer in the Transport line. To see how submission reading would map to a specific insurer, talk to WIR.
Frequently asked questions
Does intelligent reading parse quote e-mails and attachments?
Yes. Intelligent document reading treats the e-mail body and every attachment as one submission, extracting fields wherever they arrive. It reads PDF proposal forms, scanned certificates, spreadsheets, and broker cover notes across layouts it has never seen in that exact form. WIR ingests submissions from e-mail, broker portal upload, and partner API at the same time, then validates the data and flags missing or conflicting items for broker enrichment before pricing.
How does this AI layer differ from traditional OCR?
It asks what a value means, not which character sits in a pixel region. Traditional OCR is template bound and breaks when a layout changes, which is the opposite of P&C reality. WIR's Machine Learning generalizes across forms, certificates, and spreadsheets, attaches a confidence score per field, and validates each value. High-confidence values flow straight through, while low-confidence ones route to a person for a quick check.
Is extracted data validated before it moves through the journey?
Yes. Every value is checked for completeness against that ramo's requirements, for format on fields like CNPJ, CEP, dates, and currency, and for consistency across documents. Only genuinely ambiguous fields reach a human, which inverts the OCR model where people checked everything. The output is a structured, validated submission object that downstream stages, broker enrichment, risk and fraud scoring, dynamic pricing, and the final decision, can consume cleanly.
Is automatic extraction LGPD compliant?
Yes. Submission data is processed in line with the LGPD, encrypted at every step, with data minimization and a human in the loop where an automated decision affects a person. Every extraction, confidence level, validation outcome, and human override is logged, so the insurer can prove which values were machine-extracted with high precision and which a person checked. Decisions are explainable and return a full audit trail, never presented as infallible.
Do we need to replace the core to read submissions with AI?
No. WIR is an external AI layer that sits on top of existing systems, with no core migration. It ingests submissions, produces structured validated data, and pushes that into the policy core, underwriting workbench, or rating engine through APIs or files. The system of record stays intact. BCG found that 70% of insurers do not execute innovation because of IT constraints, and an overlay model removes that blocker.