Why some insurers reduce costs with automation while others stay stuck in pilots

Eighty percent of Brazilian insurers already use artificial intelligence in their operations. The data comes from a CNseg and EY research study conducted with 26 insurers representing nearly half of the national market. Despite near-universal adoption, only 23% reported significant business impact. The research also identified initiatives with significant operational gains, including time optimization between 30% and 50% in specific processes. Even so, these results remain concentrated in a limited portion of the sector's operations.

The advantage is no longer in the technology

Once virtually all insurers began testing AI, competitive advantage shifted away from the technology chosen and toward how it is applied in operations. The pattern observed in operational transformation projects at insurers that reach measurable impact is straightforward: they start by choosing the right process, integrate automation into the existing operational flow, and define success metrics before implementation. When any of these steps fails, the project tends to remain in pilot, joining the group that keeps testing without investment turning into concrete results.

In many cases, this process identification and prioritization step generates more impact on the final outcome than the technology used in the automation itself.

The right process outweighs the visible process

Many insurers prioritize the process most visible to the executive team — the one that appears in more presentations or has more stakeholders interested in it. The pattern that generates measurable impact is usually different: start with the most painful and easiest-to-measure process, even when it is operationally discrete.

In a project for triage of insurance quotes for reinsurance, highly specialized professionals consumed hours every day evaluating requests arriving from external brokers with no standard format. Most of that effort was spent discarding ineligible opportunities before technical analysis even began: economic activity outside the underwriting policy or insured value below the accepted minimum. This discarding work consumed the senior analyst's time — the most expensive resource in the operation — and stayed out of any board report. That is precisely where automation began: in the waste that consumed the team's most expensive resource daily.

Automation inside the real approval flow

The second decision is how automation enters operations. Tools tested in parallel to the real process generate internal demonstrations without impacting operational costs. In the quote triage case, AI began acting at the moment of submission, inside the same system brokers already used, extracting the critical eligibility fields and routing to the analyst only requests with a real chance of becoming business.

The result: more than 85% of the analytical effort previously consumed evaluating ineligible requests became unnecessary, freeing senior analysts for higher-value activities and directly reducing the operational cost of triage. The solution has remained in production for more than two years, without requiring reimplementation. What began as triage automation became a permanent part of operations, generating recurring gains rather than temporary results.

The success criterion comes before the contract

The third decision is what gets defined before technical implementation begins. Projects that become permanent pilots typically start without an agreed-upon success criterion. Projects that become concrete gains start with the metric and expected benchmark already defined between the parties, and with a leadership sponsor capable of unblocking barriers between departments. In the triage case, analysts helped design the eligibility criteria from the start, before the system went into production. That prior alignment — more than the technology — was what unlocked the project.

The pattern repeats beyond quote triage

The same pattern appears across different automation fronts. At an insurer processing quotes with data in free-text format, automatic extraction of fields such as tax ID, economic activity, and coverage freed underwriters from manual entry, with more than 98% accuracy and 80% reduction in review time. At another insurer operating in Central America, the business team began adjusting premium calculation rules without depending on IT, with an 80% reduction in time per change. Different products, same path to results. This pattern repeats itself in process automation projects in regulated sectors, regardless of the technology used.

When we compare to projects that stalled in pilot, the reasons tend to be surprisingly similar.

What typically sends a project to permanent pilot

In practice, projects that never leave pilot tend to share predictable characteristics:

• Process chosen without sufficient volume or recurrence to justify automation
• Automation running in parallel to the official flow, without replacing any step
• Success metrics defined only after implementation, when it is too late to course-correct
• No executive sponsor capable of unblocking barriers between departments
• Continuous dependence on IT for simple business rule adjustments

Checklist before moving forward

To evaluate whether an automation initiative is on the path that generates real operational cost reduction:

• The chosen process has measurable volume and recurrence, even if low-visibility to the executive team
• Automation replaces a step in the real approval flow, rather than running in parallel
• The success metric and expected benchmark are defined and agreed upon before implementation begins
• There is a leadership sponsor capable of unblocking barriers between departments
• The operational team participates in designing the criteria from the start

Insurers that answer yes to all five checklist questions above are in the group that has historically managed to transform automation initiatives into measurable operational cost reductions. When two or more answers are negative, the conversation about technology typically happens before the preparation necessary to capture results.

In AI initiatives applied to operations, implementation tends to receive all the attention. But in the insurers that effectively capture cost reduction, the most important decisions had already been made before the project even began.

Before investing in another AI initiative, it is worth identifying which processes concentrate the greatest operational waste and have real potential for value capture. CADS Digital supports insurers in identifying opportunities with the highest return potential and implementing automation directly into critical operational flows. Schedule a conversation to evaluate which opportunities can generate the greatest impact in your operation.

Sources:

1. CNseg and EY. Research on Artificial Intelligence adoption in the Brazilian insurance market, published in February 2026.
2. Mobile Time. 80% of Brazilian insurers already use AI in their operations. Published on February 24, 2026.