Insurance

The most paper-heavy sector in Mexico

Insurers in Mexico paid 522 billion pesos in claims during 2024, growing over 10%, driven by medical expenses, life and auto. Behind every peso paid is a file: a claim report, a policy, an expert appraisal, a medical invoice, a report. And behind every file, historically, someone reading, classifying, transcribing, and moving documents from one system to another.

That's the problem AI is solving in insurance — not just at one point, but end to end: from the moment the customer reports the claim until the payment is released.

The document cycle AI is transforming

1. Reporting and initial capture. The customer notifies the claim via app, portal, WhatsApp or call. Systems combining a chatbot for initial data collection, OCR for automatic document data extraction, and integration with the internal management system, cut file-opening time from days to under 48 hours. The customer notifies autonomously, information is classified, and it's sent to the adjuster instantly.

2. Document classification and extraction. A claim's file can contain dozens of documents in different formats: scanned PDFs, photos, invoices, medical reports. IDP (Intelligent Document Processing) solutions automatically capture and classify claim reports, policies, expert reports and medical documentation, extracting the relevant data — policy number, insured, dates, coverages, amounts, diagnoses — without manual intervention and with rules that adapt to each document type.

3. Coverage verification. Generative AI systems process the claim report, analyze attached documentation, verify coverage per policy terms, and issue a resolution — approval, denial, or request for additional information — in minutes. For low-amount, high-frequency claims (minor car collisions, home claims), this automated process resolves in hours what used to take weeks.

4. Preventive fraud detection. AI is already used in the sector for fraud detection with models comparing the reported claim's pattern against the insured's history, geolocation, driving behavior (telematics), and known fraud patterns — before releasing payment, not after. Fraud that used to be caught in later audits now gets filtered in the authorization flow.

5. Underwriting and quoting. Generative AI impacts underwriting by enabling hyper-personalized quotes based on behavioral data. Several insurers in Mexico use predictive models and telematics — especially in auto — to calculate premiums based on driving behavior, claims history and contextual variables. The result is a more accurate premium for the insured and lower loss ratio for the insurer.

The measured impact

The clearest ROI cases for AI in Mexico's insurance sector concentrate in three areas: 20-40% reduction in manual tasks when processes are correctly redesigned; use of intelligent triage, automated document validation and integrated workflows; and copilots for agents and quoting/issuance automation, according to Tech and Solve (El Asegurador México).

The impact pattern is consistent: every process that used to require a human to read a document, extract data, and capture it in another system is an automation candidate with direct return — the sector's volume (millions of policies, hundreds of thousands of claims a year) turns marginal savings per file into millions a year.

The "decision factory" as an operating model

The most relevant 2026 trend is agentic AI in insurance: unlike traditional automation, this approach coordinates complete processes and lets different systems collaborate to meet broader goals, connecting areas like underwriting, document management, claims, customer service, fraud and back office.

McKinsey argues that AI's real value in insurance shows up when it's integrated systematically across different business processes. In practice, this looks like "building a decision factory" with automated processes and human involvement only where risk or regulation demands it.

In 2025, many AI initiatives in insurance were pilots. The relevant 2026 shift is industrialization: AI in production, integrated into the process, with governance. Generative AI grows as a copilot for sales, operations and support — with more control than enthusiasm.

The limit that can't be ignored

In Mexico, the CNSF still has no specific provisions for AI, though the LISF, LSCS and LFPDPPP generate applicable obligations by analogy. The insurer's liability doesn't transfer to the algorithm: any automated decision breaching legal or contractual obligations can be challenged.

The principle guiding any responsible implementation in the sector: without good document management there's no control. Scattered papers, poorly classified PDFs, or duplicated information generate delays, errors and unreliable decisions. AI needs structured data and well-defined processes before adding intelligence — automating chaos produces automated chaos, faster.

The underlying question

It's not whether AI can automate the claims cycle — it already is at several insurers in Mexico. The question is whether the starting point (document management, data quality, process definition) is in shape to support automation, and whether the insurer is ready to operate with human oversight where regulation requires it, without slowing flows AI can resolve completely.