What next-best-action for insurers with an AI layer means
Next-best-action for insurers with AI is a distribution capability that reads an insurer's existing portfolio, scores which client should be offered which product next and through which corretor (broker), and routes that action to the channels the commercial team already uses. It runs as an external AI layer on top of the insurer's current CRM and policy systems, so it reads from them without replacing them. The reader who should care is the distribution or commercial lead who knows there is white space in the book but cannot see it at the account level.
The problem this solves is the invisible pipeline. Brazilian Seguros e Danos (P&C) distribution runs overwhelmingly through independent brokers, and the book an insurer holds is in practice a federation of broker relationships, legacy policy records, and CRM fields that were never built to answer one question: which account, with which broker, deserves which offer next, and why. Policy systems store contracts by ramo (line of business) and renewal date, not by household or commercial account, so a single insured may hold auto with one broker, property with another, and no life or liability cover at all, while nothing in the stack surfaces that gap. Without an explicit score on the next action, commercial teams fall back on renewal lists and gut feel, and the highest-propensity account gets the same generic outreach as a dead one. The pressure to fix this is real because the Seguros e Danos market grows double digits per year, which means even small penetration gains are material. WIR is the AI layer for insurance built for exactly this account-level intelligence.
How upsell and next-best-action scoring works
The work happens in four stages, layered on top of the systems the insurer already runs. First comes portfolio mapping across client and product. The layer ingests portfolio, policy, claims, and CRM data through APIs or scheduled extracts, resolves the same insured across ramos and across brokers, normalizes the product taxonomy, and builds a client by product matrix that exposes the white space. This is also where duplicate CPF or CNPJ records and inconsistent broker codes get reconciled, since the score is only as good as the map underneath it.
Second, Machine Learning models score each account for upsell propensity and rank the next-best-action. The score is calibrated to the insurer's risk appetite and underwriting manual, so the recommended action respects underwriting rules and product margin, not propensity alone. Next-best-action answers a sequenced question, not only whether an opportunity exists but which single action, for this account, through which broker, carries the highest expected value now. This is the same scoring discipline used on the automated underwriting journey, applied to the distribution side.
Third, scored actions become multi-channel campaigns. Each recommendation is routed to the right place, a broker task list, a commercial team queue, direct digital outreach, or a partner channel, and every step is logged. The attribution trail records which signal generated the recommendation, which broker or channel acted, which touch preceded the quotation, and which one closed the sale. Fourth, penetration and retention grow together rather than as separate programs. An account that holds more products with the insurer is structurally stickier, and acting before renewal or at a life or business event lifts both renewal rates and products per account. Penetration and retention are two readouts of the same account-level engine.
How to deploy distribution intelligence as an external layer
Deployment is an integration, not a migration. The layer leaves the system of record untouched, the CRM stays the CRM and the policy system stays the policy system, and it connects through APIs or scheduled extracts rather than a full platform rebuild. That stance matters in a market where roughly 70% of insurers do not execute innovation because of IT limitations, according to BCG. An external layer sidesteps that constraint because nothing in the core has to change for it to run, and if the layer is switched off the core continues to operate exactly as before.
The path runs in clear steps. The insurer first scopes the lines and books where white space is largest and data is cleanest, typically auto plus patrimonial, where account overlap is high, and fixes the success metric up front, whether that is products per account, cross-sell conversion, or retention on multi-product accounts. Integration comes next, read-only ingestion from the policy core and CRM, with recommendations flowing to the channels and to the CRM activity layer rather than back into the contract tables of the system of record. Then comes calibration, tuning the scoring to the underwriting manual, risk appetite, and margin per ramo so the recommended actions are sellable and underwritable. Testing follows, backtesting scores against historical conversions and running a champion challenger or holdout group so the attribution trail can prove incremental lift rather than correlation. Go-live releases prioritized lists and campaign triggers to a pilot set of brokers and commercial teams with humans kept in the loop, and continuous operation retrains the models on outcomes, monitors for drift, and expands to more ramos as confidence grows. With WIR, this setup runs 3 to 12 months as a fixed-scope engagement with KPIs agreed before it starts, followed by a continuous operation phase after go-live.
Governance, attribution trail, and LGPD
Automated scoring that influences commercial offers has to be explainable and auditable, and in Brazil that is a regulatory expectation, not a preference. Each next-best-action score should be traceable to its drivers, the features that pushed propensity up, so commercial and compliance teams can justify why an account was prioritized. Opaque scoring is both a governance risk and a trust risk, which is why the layer is built to explain its decisions rather than to hide them.
The attribution trail does double duty here. The same log that proves campaign effectiveness serves as the audit record of who was contacted, on what basis, through which channel, and with what result, which supports both performance review and regulatory defensibility. All of this sits under Brazil's Lei Geral de Proteção de Dados, the LGPD, which governs the personal data processed in portfolio mapping and scoring. The layer must operate on a valid legal basis, honor data-subject rights, minimize data to what the scoring needs, and log its processing, and the LGPD gives data subjects the right to request review of decisions made solely on automated processing, which is why human-in-the-loop and explainability are not optional for automated commercial scoring. The full text of the law is available from the official Brazilian government source. Every decision is explainable and returns a full audit trail, and data is encrypted at every step.
How WIR automates next-best-action with Smart Sales
WIR delivers this through Smart Sales, its distribution-intelligence module, and as an external AI layer it sits on top of the insurer's existing systems and never replaces the core. Smart Sales maps the portfolio across client and product, scores upsell and next-best-action with Machine Learning calibrated to the insurer's risk appetite and underwriting manual, and runs multi-channel campaigns with an attribution trail, so penetration and retention grow together. It prioritizes accounts and brokers rather than becoming the place where policies are issued or where the customer record lives, which is the distinction that keeps the program low-risk and reversible. The AI layer for insurance, on top of the systems the insurer already runs, never in their place.
Smart Sales runs alongside Underwriter Intelligence, which automates the quotation journey per the insurer's risk policy, with real-time ML scoring calibrated to appetite, automatic routing by appetite and exposure, and predictive conversion analysis by product, risk, and broker, plus real-time dashboards over the in-flight pipeline. WIR is an InsurTech and the AI layer of insurance in Brazil, not an insurer, broker, or MGA, and it does not carry risk. Its first traction is a POC in execution with a global insurer in the Transport line, and every figure here on the Brazilian market comes from named sources rather than from WIR's own book. To see where next-best-action would prioritize accounts and brokers in your portfolio, book a conversation with the WIR team at wirinnovation.ai.
Frequently asked questions
How is the next-best-action score calculated?
Machine Learning models score each account for upsell propensity and rank the single highest-value action, calibrated to the insurer's risk appetite and underwriting manual. Before scoring, the layer maps the portfolio across client and product, resolving the same insured across lines of business and brokers. The score respects underwriting rules and product margin, not propensity alone, and every recommendation is explainable and traceable to the drivers that pushed propensity up.
Does distribution intelligence replace the insurer's CRM?
No. Distribution intelligence runs as an external AI layer on top of the existing CRM and policy systems, reading from them without replacing them. WIR's Smart Sales prioritizes accounts and brokers rather than becoming the place where policies are issued or the customer record lives. Deployment is an integration through APIs or scheduled extracts, not a migration. If the layer is switched off, the core continues to operate exactly as before, which keeps the program low-risk and reversible.
Do the multi-channel campaigns have an attribution trail?
Yes. Each scored action routes to a broker task list, commercial queue, digital outreach, or partner channel, and every step is logged in an attribution trail. The trail records which signal generated the recommendation, which broker or channel acted, which touch preceded the quotation, and which one closed the sale. That same log doubles as the audit record of who was contacted and on what basis, supporting both performance review and regulatory defensibility under the LGPD.
How do penetration and retention grow together with this layer?
An account that holds more products with the insurer is structurally stickier, so penetration and retention are two readouts of the same account-level engine. Smart Sales acts before renewal or at a life or business event, which lifts both products per account and renewal rates rather than running them as separate programs. Because Brazil's Seguros e Danos market grows double digits per year, even small penetration gains on existing accounts are material.
How long until Smart Sales goes into production?
With WIR, setup runs 3 to 12 months as a fixed-scope engagement with KPIs agreed before it starts, followed by a continuous operation phase after go-live. The path covers scoping the lines with the largest white space, read-only integration with the core and CRM, calibration to the underwriting manual, backtesting against historical conversions, and a piloted go-live with humans kept in the loop. Models then retrain on outcomes and expand to more lines as confidence grows.