Experience Data Preparer
L3 · Act within limits
Assembles exposure, premium and claims experience by segment from policy and claims systems, flags gaps and reconciles totals before the actuarial team starts.
Agents that do the routine work across the insurance value chain — reading submissions, triaging claims and complaints, preparing fraud referrals and reconciling the books — each with a narrow remit, an autonomy level and a named person who owns the outcome. Underwriting, claims and investigation decisions stay with people.
These are agent designs from our AI & Agentic Engineering practice, built on the same controls as the live banking demo. The live demo runs on banking sample data.
Agents that keep pricing and product teams supplied with clean experience data and early signals, so actuaries spend their time on judgement rather than data preparation.
Assembles exposure, premium and claims experience by segment from policy and claims systems, flags gaps and reconciles totals before the actuarial team starts.
Tracks how filed rate changes land: retention, mix and loss-ratio drift by segment, and drafts the monitoring note for the pricing committee.
Links customer verbatims and complaints to product features and exclusions, so product owners see which terms cause confusion or disputes.
Agents that support brokers, tied agents and partner channels with faster answers and clean commission data, while the relationship stays human.
Answers broker questions on appetite, documents and policy status from governed sources, and drafts replies for the account handler.
Matches commission statements to bound policies and partner agreements, explains differences and prepares the adjustment file.
Profiles each partner channel on volume, conversion, loss ratio and complaints, and flags channels drifting from appetite or conduct expectations.
Agents that remove the re-keying and evidence-chasing from underwriting. Underwriters keep their authority; agents prepare the file.
Reads broker emails, applications and schedules with document AI, extracts the risk data into the underwriting workbench and lists what is missing.
Checks each submission against underwriting guidelines and appetite, and shows the rule and evidence behind every flag.
Builds the referral pack for risks above authority: summary, evidence, comparable risks and a draft recommendation with reasons.
Agents and bots that work the servicing queues across legacy admin systems, so changes, renewals and documents are right the first time.
Classifies incoming service requests by intent and urgency and routes them to the right queue with the policy context attached.
Prepares mid-term changes and renewal documents, checks them against the policy record and highlights differences for sign-off.
Compares billing, collections and policy terms to find under-charged premium, missed instalments and incorrect cancellations.
The voice-of-the-customer squad: every call, survey and complaint analysed, with personal data masked first and people deciding the critical cases.
Classifies each call by category, sub-category and reason, scores sentiment and agent quality, detects claims mentioned and writes the call summary.
Detects frustration, classifies scope and assigns an impact level from L0 critical to L3 low, so critical complaints are escalated the same day.
Tags survey verbatims with enterprise, programme and AI-discovered themes across platforms and languages, including regional Spanish and Brazilian Portuguese.
Detects and masks names, account numbers, phone numbers, emails and addresses in text before any model or analyst sees it.
Agents that assemble the claim file and watch the customer voice for delays. Coverage, settlement and payment decisions stay with adjusters.
Captures first notice of loss from any channel, structures the details, checks the policy is in force and proposes a triage route: simple, standard or complex.
Collects estimates, invoices, reports and notes into one file, extracts amounts with document AI and highlights inconsistencies for the adjuster.
Spots approved claims that are not paid on time from calls and complaints, and raises a case before the customer escalates.
Investigator agents that gather evidence and draw the network, so SIU spends its time deciding. Agents never accuse, deny or refer on their own.
Scores claims at FNOL and before payment and explains the indicators behind each score.
Links claimants, policies, repairers, providers and attorneys to surface shared identities and repeated patterns.
Builds the investigation pack — timeline, evidence, links and open questions — for a referral the investigator decides on.
Agents that take the manual matching and assembly out of the close, reserving and regulatory reporting, with lineage from policy and claim to the reported number.
Matches premium, claims and cash across policy, claims, bank and ledger, and explains the breaks it cannot clear.
Identifies claims that trigger treaty or facultative recoveries, calculates the expected recovery and prepares the bordereau lines.
Checks contract-group, cash-flow and assumption data feeding IFRS 17 and Solvency II reporting for completeness, consistency and lineage.
Drafts the variance and results commentary for the close and board pack from governed numbers.
Policy decides what goes straight through; people approve everything else; a kill switch halts all agents. The live demo runs on banking sample data — the same controls apply to every Insurance squad.