Now booking software engineering engagements for Q3 — Q4 2026

Claims, underwriting,
and the document work in between.

Claims FNOL and triage automation, underwriting workflow platforms, document-heavy intake at scale, and the subrogation and recovery tooling that recovers what the policy says it should. Built so the audit trail and the policy contract agree.

The landscape

What insurance software looks like in practice.

Carriers, MGAs, and TPAs sit on top of policy-administration and claims systems that are usually one or two generations behind what the front office now needs. The work that fills the day isn't the policy system itself. It's the FNOL intake, the document review, the adjuster routing, the subrogation queue, the regulatory reporting, and the eight integrations between underwriting and the rate filings.

The pressure isn't to replace the policy admin or claims platform. Those replacements take years and burn capital. The pressure is on the surrounding work: claims FNOL that still arrives by phone and email, underwriting decisioning that still routes by spreadsheet, subrogation files that age while nobody looks at them, customer-facing workflows that haven't kept pace. AI is starting to land here in real ways, but every output that touches a claim or a policy decision needs the audit trail to back it up.

Apollo builds the workflow, data, and AI work that sits around the core insurance platforms. State-regulator-aware by default, audit-recorded everywhere a decision gets made, and integrated with the policy admin, claims, and reporting systems you already run.

The regulatory reality

Fifty state insurance regulators, the NAIC model laws behind them, market-conduct examinations, rate-filing requirements, and a privacy stack that picks up GLBA, HIPAA for health-adjacent products, and state privacy laws. Add fair-claims-handling rules and unfair-trade-practices statutes. Every state has a slightly different version of all of it.

Where systems break

The FNOL intake that splits across phone, fax, email, and a portal that doesn't talk to claims. The underwriting workflow that lives partly in a vendor product and partly in a spreadsheet a senior underwriter maintains. The subrogation queue that ages because the recovery system can't read the claim file. The reporting submission that gets stitched together at the end of every quarter.

Where AI lands

FNOL triage and severity scoring with explicit reasoning trails. Document extraction for loss-run files, medical records, police reports, and underwriting submissions. Adjuster copilots that pull the claim history and the policy language at the same time. Subrogation identification on claims the human queue would have missed. Every output reviewable, every action attributable.

Where we work

Four engagement patterns in insurance.

Most engagements with carriers, MGAs, and TPAs land in one of these shapes. Each gets built audit-recorded by default, integrated with the policy admin and claims systems you already run, and ready for the market-conduct examiner who will eventually ask.

Claims FNOL & Triage Automation

First-notice-of-loss intake across phone, email, portal, and broker channels, normalized into a single claim file. Severity and complexity scoring with explicit reasoning trails. Adjuster routing that respects licensing, line-of-business, and capacity at the same time. Every decision attributable to the model version that made it.

FNOL intakeSeverity scoringRouting logicAttributable decisions

Underwriting Workflow & Data

Submission intake, document extraction, third-party-data enrichment, and the workflow tooling underwriters actually want. AI assistance on the unstructured parts (loss-run review, exposure summarization), with the model documentation and rate-filing-aware boundaries that compliance will check.

Submission intakeData enrichmentAI assistRate-filing-aware

Document-Heavy Intake

Loss-run files, medical records, police reports, repair estimates, certificates of insurance, and the submission packets underwriters receive in 47-page PDFs. Extracted, classified, and routed into the workflow the human side actually uses. Confidence thresholds route the clean cases through and push the rest into reviewer queues.

OCR + extractionDoc classificationConfidence scoringReviewer queues

Subrogation & Recovery Workflows

The recovery work that gets left on the table when nobody has time to look. Subrogation-opportunity identification on closed and in-flight claims. Recovery-queue prioritization. Workflow tooling for the recovery specialists who run the queue. Built to integrate with the claims system you already have, not to replace it.

Recovery IDQueue prioritizationWorkflow toolingClaims integration
The systems landscape

Where our work sits.

Insurance operators run a small set of core platforms with a lot of document and channel traffic around them. Apollo builds the workflow, data, and AI work that sits around this landscape, not on top of it. The map below is simplified, but representative of what we walk into.

CORE INSURANCE SYSTEMSDOCUMENT & CHANNEL SOURCESAPOLLO LAYERAPPLICATIONS + AIPolicy AdminIssuance · endorsements · renewalsClaims SystemsAdjudication · payments · reservesBilling & ReinsurancePremium · ceded · recoverablesInbound DocumentsLoss runs · medical · estimatesIntake ChannelsPhone · email · portal · brokerThird-Party DataMVR · MIB · CLUE · ISOIntegration LayerAPIs · events · file feedsApollo Platform LayerWorkflow · identity · auditDocument PipelineOCR · extraction · reviewFNOL & TriageIntake · scoring · routingUnderwriting ToolsSubmission · enrich · decideSubrogation & RecoveryIdentify · prioritize · pursueAdjuster CopilotsLogged · model-versioned
Cross-cutting concerns
Audit trail per claim eventModel versioning & reasoningFair-claims-handling boundariesState-by-state rule logicReviewer queues by default
Sits around the policy admin, not on topAudit-recorded by constructionExaminer-aware by defaultDocument intake is a first-class concern
The regulatory context

Compliance, designed in.

The frameworks that shape what we build, and what we ship.

Insurance is a state-regulated business in the US, which means fifty regulators, the NAIC behind them, and a stack of model laws each state adopts a slightly different version of. The right set of frameworks for any engagement depends on the lines of business, the states in scope, and whether health-adjacent or life-adjacent products bring in extra regimes.

Apollo treats these as design inputs, not as a checklist applied at the end. State-by-state rule boundaries, audit trails, model documentation, privacy controls, and reviewer-queue posture get specified before the first line of code and reviewed against the relevant frameworks at every iteration.

The panel on the right is the working set we encounter most often. Any given project picks a subset, and the proposal will be explicit about which ones apply and what we will and won't certify directly.

insurance / compliance-frameworks

State & market conduct

State DOIsFifty state departments of insurance with their own rules
NAIC model lawsModel acts adopted in variant forms across the states
Market-conduct examPeriodic state exams on claims handling and underwriting
Rate filingsFiled rates, rules, and forms each state holds carriers to

Privacy & data

GLBAPrivacy and safeguards for nonpublic personal information
HIPAAWhere health-adjacent products are in scope
State privacyCCPA, CPRA, and the growing patchwork of state privacy laws
Fair ClaimsFair-claims-handling and unfair-trade-practices acts

Industry frameworks & controls

SOC 2 Type IIControls attestation for the platforms we build
NIST CSFCybersecurity framework increasingly cited by regulators
ISO 27001Information-security management standard
How we engage on insurance work

Four phases. Built around the examiner and the policy.

Apollo's standard methodology, applied to insurance. Audit posture, state-by-state rule boundaries, and reviewer-queue design get specified before the build starts, so market-conduct readiness is a property of the system rather than a phase at the end.

Discovery

Map the workflow. Map the rules.

The current process, the systems involved, the document and channel volumes, and the breakpoints your team already knows about. The lines of business, the states in scope, and the rule boundaries that come with them. Where AI helps, where rules suffice, and where a reviewer still has to look.

Design

Architecture, audit plan, reviewer-queue plan.

System architecture and integration design against the policy admin and claims platforms. Audit trail design that satisfies SOC 2 and market-conduct expectations. State-by-state rule logic that doesn't sprawl into spaghetti. Reviewer UI wireframes for the human side of every decision loop.

Build

Platform. Document pipeline. Reviewer UIs.

Platform deployed inside the security boundary, policy admin and claims integrations connected, document extraction wired up with model versioning, reviewer UIs functional. Two-week iterations. Each shipped capability arrives with its tests, its monitoring, and its audit fields.

Operate

Measure. Tune. Expand.

Throughput, decision quality, AI accuracy, and reviewer-override rates in production dashboards. Threshold tuning based on what the data shows. Gradual rollout to adjacent lines of business or states once the first workflow is steady. Knowledge transfer to your team along the way.

Start a conversation

Tell us what you're trying to build.

Send a paragraph about the project: the lines of business, the systems involved, the volumes if you have them, the states in scope, and the parts that aren't working today. We'll reply within one business day, either with a 30-minute call or with an honest "this is not the right fit; here's who you should call instead."

A senior engineer reads every inquiry before anyone replies.
Thank you. Your message is in. We'll be in touch within one business day.