BCG~20% cost, 3-5% GWP
AI-first carriers are separating from the pack
Carriers redesigning core processes around AI report roughly 20% cost reduction and 3-5% GWP growth from share gains and productivity.
Program 04 · Domain specialisation
SCIKIQ Certified Data and AI EngineerInsurance
Insurance is the industry where agentic AI has moved fastest from pilot to production — and where the failure modes are most expensive. Quote in three minutes instead of three days is real; so is an unfair-discrimination finding. This specialisation covers both.
- 80 hrstaught
- 40hands-on AI labs
- 8assessment tests
- +20 hrs2 E2E capstones
~20%Cost reduction captured by AI-first insurersBCG
3-5%GWP growth from share gains and productivityBCG
70-90%Straight-through processing on simple claims, up from 10-15%
20+US jurisdictions that adopted the NAIC AI model bulletinNAIC
01
Part 1 · The business
The value chain, the people, the trends
Insurance is a chain of judgements about risk, made under time pressure with incomplete information. Part 1 maps where those judgements happen, who makes them, and what the strategy houses see changing.
The insurance value chain
From product design to reserving. Select a stage to see what breaks and who owns it.
01
Product & pricing
Product design, rating factors, filings and portfolio strategy.
Decisions made here
- What risk will we take?
- At what rate?
- What will the regulator accept?
Where it breaks today
- Rate filings that take quarters
- Actuarial insight trapped in spreadsheets
- Slow feedback from experience to price
What is measured
- Loss ratio
- Rate adequacy
- Filing cycle time
Where AI landsExperience analysis at speed, filing documentation drafting and scenario modelling.
Loss ratioRate adequacyFiling cycle time
02
Distribute & quote
Submission intake, appetite matching, triage and quotation.
Decisions made here
- Is this in appetite?
- Should we quote it?
- How fast can we respond?
Where it breaks today
- Submissions arriving as email plus spreadsheet plus PDF
- Broker questions answered inconsistently
- Good risks lost to slower response
What is measured
- Quote turnaround
- Hit ratio
- Submission-to-quote rate
Where AI landsSubmission ingestion and pre-fill, appetite matching and instant triage.
Quote turnaroundHit ratioSubmission-to-quote rate
03
Underwrite & bind
Risk assessment, pricing decision, terms, referral and binding.
Decisions made here
- Accept, decline or refer?
- What terms and exclusions?
- What price?
Where it breaks today
- Re-keying submission data into the rating engine
- Referral queues waiting on senior capacity
- Inconsistent decisions across teams
What is measured
- Time to quote
- Referral rate
- Loss ratio by segment
Where AI landsPre-fill, risk-signal enrichment, guideline checking and drafted rationale for referral.
Time to quoteReferral rateLoss ratio by segment
04
Service & endorse
Policy administration, endorsements, renewals and customer service.
Decisions made here
- What is the customer entitled to?
- Does this endorsement change the risk?
- Should we renew and at what terms?
Where it breaks today
- Policy wording questions escalated unnecessarily
- Renewal prep done manually at volume
- Service data disconnected from underwriting
What is measured
- Retention
- Service cost per policy
- Endorsement cycle time
Where AI landsGrounded policy-wording answers, renewal preparation and endorsement impact assessment.
RetentionService cost per policyEndorsement cycle time
05
Claims: FNOL to settlement
First notice, triage, coverage check, investigation, reserve and settlement.
Decisions made here
- Is it covered?
- What is it worth?
- Is it fraudulent?
- Fast-track or investigate?
Where it breaks today
- Coverage checked by reading the wording each time
- Reserves set late and revised often
- Fraud signals spotted after payment
What is measured
- Cycle time
- Leakage
- STP rate
- Loss adjustment expense
Where AI landsAutomated FNOL capture, coverage determination with citations, damage assessment and fraud signals.
Cycle timeLeakageSTP rateLoss adjustment expense
06
Reserve, reinsure & report
Reserving, reinsurance cessions, regulatory reporting and portfolio review.
Decisions made here
- Are reserves adequate?
- What do we cede?
- What do we disclose?
Where it breaks today
- Reserving cycles constrained by data preparation
- Cession errors found in audit
- Portfolio insight arriving a quarter late
What is measured
- Reserve adequacy
- Cession accuracy
- Reporting cycle time
Where AI landsData preparation, anomaly detection on cessions and drafted commentary with traceable numbers.
Reserve adequacyCession accuracyReporting cycle time
Who you are building for
Five roles whose decisions the agent will either sharpen or undermine. Select one to light the stages they own.
What the strategy houses are seeing
Insurance is further ahead than most industries — and further exposed.
BCGThree core processes
Claims, underwriting and servicing first
Agentic effort concentrates in claims, underwriting and servicing; distribution and customer service follow rather than lead.
BCGOversight, not operation
Agent-led execution, human-held accountability
Work shifts from human coordination to agent-led execution, with human judgement reserved for complex risk and accountability.
IndustrySTP 10-15% → 70-90%
Straight-through processing is the metric that moved
STP on simple claims has gone from roughly 10-15% to 70-90% at leading carriers, and time-to-quote on standard SME risk from days to minutes.
NAIC20+ jurisdictions
Regulation arrived faster than in most sectors
More than 20 US jurisdictions have adopted the NAIC model bulletin on insurer use of AI, which expects governance, testing and documentation.
HBRTest during build
Fairness is a design constraint, not a review step
Proxy discrimination is the reputational risk that ends programmes. It has to be tested for during the build, not discovered in audit.
Where these numbers come from — 4 sources
Consulted when this program was written, in July 2026. Each entry names the claim on this page that it backs, so the pairing stays checkable if the copy is later edited.
- BCGExecutive Perspectives: AI-first companies win the future — P&C insurance~20% cost reduction and 3-5% GWP growth
- BCGThe AI-First Life Insurance CompanyClaims, underwriting and servicing lead adoption
- BCGCompetitive Advantage in the Agentic AI EraAgent-led execution with human accountability
- IndustryAI in Insurance: Underwriting, Claims & Fraud Detection 2026STP on simple claims rising from 10-15% to 70-90%
02
Part 2 · Data and meaning
Domains, ontology, taxonomy and governed terms
Insurance data is unusually document-heavy and unusually regulated. The architecture that works is one where every coverage determination can be traced to a clause and every rating input to a source.
The data landscape
Six families, and the wording corpus is the one that decides your accuracy.
Policy administration
Policies, endorsements, coverage terms and the transaction history.
Guidewire PolicyCenter, Duck CreekClaims systems
FNOL, claim notes, payments, reserves and adjuster narrative.
Guidewire ClaimCenter, SapiensWordings & forms corpus
Policy wordings, endorsements, ISO forms. Where coverage truth lives.
Forms library, ECMSubmission intake
Broker email, ACORD forms, schedules and loss runs — the messiest input in insurance.
ACORD, email, spreadsheetsThird-party risk data
Property, geospatial, catastrophe, credit and telematics enrichment.
Cat models, geospatial, MVRActuarial & finance
Experience data, triangles, reserves and cession records.
Actuarial marts, GLThe semantic layer
Coverage is contractual, so meaning is contractual too. If the agent’s idea of "policy", "claim" and "exposure" does not match the carrier’s, every determination it makes is unsafe.
Data domains
Six subject areas, each with the entities it holds, its critical data elements and the function accountable for it.
Party & role
Everyone attached to a policy and the role each one plays on it.
PolicyholderInsuredBeneficiaryBrokerClaimant
- Critical data elements
- Party identifier; Role type; Relationship to insured
- Owned by
- Distribution & underwriting
Policy & coverage
The contract, its coverages, limits, exclusions and endorsements.
PolicyCoverageEndorsementWording version
- Critical data elements
- Policy number; Effective and expiry dates; Limit and deductible
- Owned by
- Underwriting
Risk & exposure
What is insured, where it sits, and what it is worth.
Insured objectLocationPerilSum insured
- Critical data elements
- Total insured value; Occupancy class; Geocode
- Owned by
- Underwriting & catastrophe
Claim
Everything from first notice to settlement, including reserves and payments.
ClaimReservePaymentRecovery
- Critical data elements
- Claim number; Loss date; Incurred amount
- Owned by
- Claims
Financial & actuarial
Premium, reserving, reinsurance cessions and experience.
PremiumCessionTriangleLoss ratio
- Critical data elements
- Written premium; Earned premium; Ceded amount
- Owned by
- Actuarial & finance
Product & rating
What is sold, the rating factors, and the filings behind them.
ProductRating factorRate tableFiling
- Critical data elements
- Product code; Rating factor value; Filing reference
- Owned by
- Product & pricing
Ontology & knowledge graph
The entities and the typed relationships between them — what the agent traverses instead of guessing joins. Select any entity.
Core entityEventReference / classification
Core entity
Party
Any person or organisation attached to a policy or claim.
- Party claims under Policy
- Policyholder is a Party
Core entity
Policyholder
The party that contracts with the carrier.
- Policyholder is a Party
- Policyholder holds Policy
Core entity
Policy
The insurance contract for a period of cover.
- Policy grants Coverage
- Policy governed by Wording version
- Policy amended by Endorsement
- Policy instantiates Product
- Policyholder holds Policy
- Claim made against Policy
- Party claims under Policy
Core entity
Coverage
A specific protection granted within a policy.
- Coverage protects Insured object
- Coverage responds to Peril
- Policy grants Coverage
Reference / classification
Wording version
The exact contract text edition attached to the policy.
- Policy governed by Wording version
Event
Endorsement
A change to the contract mid-term.
- Policy amended by Endorsement
Core entity
Insured object
The property, vehicle, life or liability covered.
- Coverage protects Insured object
- Claim affects Insured object
Reference / classification
Peril
A cause of loss the coverage responds to or excludes.
- Coverage responds to Peril
- Claim caused by Peril
Event
Claim
A notified loss made against a policy.
- Claim made against Policy
- Claim caused by Peril
- Claim affects Insured object
- Claim carries Reserve
- Claim settled by Payment
Event
Reserve
The estimated ultimate cost of a claim.
- Claim carries Reserve
Event
Payment
Money paid on a claim.
- Claim settled by Payment
Reference / classification
Product
The rated offering the policy instantiates.
- Policy instantiates Product
Taxonomies
The classification hierarchies that make records comparable across systems.
Product & line of business
- Segment
- Line of business
- Product
- Coverage
Commercial → Property → Package → Business interruption
Cause of loss
- Category
- Peril
- Sub-peril
Natural → Windstorm → Named storm
Occupancy / risk class
- Class
- Sub-class
- Occupancy code
Manufacturing → Food processing → Bakery
Governed business terms
The definitions an agent must use rather than invent. Most wrong answers in this industry are a term used loosely.
- Policy
- The contract of insurance for a defined period. A renewal is a new policy, not the same one.
- Coverage
- A specific protection within a policy, with its own limit, deductible and exclusions.
- Wording version
- The exact contract text edition attached to a policy — the source of any coverage answer.
- Insured
- The party whose interest is protected, which may differ from the policyholder who pays.
- Incurred
- Paid amount plus outstanding reserve on a claim. Not the same as paid.
- Leakage
- Money paid on a claim that, correctly handled, should not have been paid.
- Earned premium
- The portion of written premium relating to cover already provided.
- Total insured value
- The full declared value at risk at a location, driving accumulation limits.
03
Part 2 · The architecture
How it is actually built
The layers a deployment needs, what each one holds, and what the engineer owns there.
The reference architecture
Seven layers. The clause-citation path is the one that gets audited.
01/07
Sources & intake
Policy admin, claims, forms library, broker submissions and third-party enrichment.
Takes in
Nothing. This is the upstream edge.
Hands on
Policy, claim and wording feeds, each carrying the edition and effective dates it was valid under.
SystemWhat it doesKey business entities
Policy administration
Issues and services the contract. Holds the wording edition attached to each policy, which is the fact most coverage answers turn on.
PolicyCoverageEndorsementWording edition
Claims management
FNOL through to settlement, including reserves movements. The reserve history, not the final figure, is what tells you how a claim actually developed.
ClaimReservePaymentRecovery
Underwriting and rating
Risk capture, rating engines and referral rules. Where a decision was made is recorded separately from why, and joining the two is a routine piece of work.
SubmissionRiskRateReferral
Billing and collections
Premium, instalments and lapse. Drives the retention questions, and disagrees with policy admin more often than anyone expects.
PremiumInstalmentLapse
Reinsurance and bordereaux
Treaty and facultative cessions, and the bordereaux exchanged with partners. Spreadsheet-shaped in practice, however the contract describes it.
TreatyCessionBordereau
Document and correspondence stores
Wordings across editions, schedules, adjuster reports and medical evidence. Multiple editions of one wording live here simultaneously.
WordingScheduleAdjuster report
What the engineer owns here
Turning unstructured broker submissions into something a pipeline can trust.
Where this layer fails
Treating the wording corpus as one document set. Four editions of one wording differ in a single exclusion, and an agent that answers from the newest one is confidently wrong.
02/07
Extraction & normalisation
Document AI over ACORD forms, schedules and loss runs, with confidence-based routing.
Takes in
Read-mostly access paths, change feeds and document streams from the systems of record.
Hands on
Replayable, resolved, quality-checked records with their lineage back to the source row.
SystemWhat it doesKey business entities
Change data capture
Reads the source’s own change log rather than polling it, so the platform sees every state a record passed through instead of only where it ended up.
Change eventOperation typeCommit timestamp
Landing and raw zone
Every extract kept as it arrived, immutable and timestamped. The thing you replay from when a downstream definition turns out to have been wrong.
Raw extractBatchWatermark
Document processing
Layout-aware extraction over the unstructured half of the estate: sectioning, tables, signatures, and the page reference every later citation depends on.
DocumentSectionPage reference
Entity resolution
Decides that two records are the same real-world thing, with a survivorship rule and a confidence, so downstream joins are a decision rather than an assumption.
Match candidateSurvivorship ruleCluster
Data quality and contracts
Schema, freshness and volume expectations asserted at the boundary, so a bad load fails loudly here rather than quietly two layers later.
ContractExpectationViolation
What the engineer owns here
The confidence threshold that decides what a human still sees.
Where this layer fails
Snapshot loads instead of change capture. It looks identical in a demo and it silently loses every intermediate state, which is exactly what an audit later asks you for.
03/07
Governance & semantics
Coverage ontology, governed metrics, entitlements and the wording-to-clause map.
Takes in
Replayable, resolved records with lineage from ingestion.
Hands on
Typed, classified, defined data an agent can be grounded on without inferring meaning.
SystemWhat it doesKey business entities
Business glossary
One agreed definition per term, owned by a named person, so the number an agent quotes means what the business means by it.
TermDefinitionOwnerSteward
Ontology and knowledge graph
The typed entities and relationships the domain actually has, so an agent can traverse "which customers are exposed to this" rather than guess from adjacent text.
EntityRelationshipNodeEdge
Semantic and metrics layer
Metrics defined once, in one place, with their filters and grain. Removes the class of error where the agent computed something plausible and wrong.
MetricDimensionGrainFilter
Classification and policy
Sensitivity labels and the purpose each classification permits, applied at the field level and inherited by everything downstream.
ClassificationPurposePolicy
Lineage and catalogue
Where a value came from and what it touched on the way. The layer that makes an answer defensible rather than merely correct.
Lineage edgeDatasetColumn
What the engineer owns here
Making “covered” a resolvable reference rather than a model opinion.
Where this layer fails
Letting the model infer meaning from column names. It will, it will be plausible, and nobody will notice until the number reaches a regulator or a board pack.
04/07
Retrieval & grounding
Clause-level retrieval over wordings, precedent claims and guideline documents.
Takes in
Typed, classified, defined data from the governance layer.
Hands on
Ranked, permission-trimmed, citable evidence scoped to the caller and the moment.
SystemWhat it doesKey business entities
Hybrid index
Vector and lexical retrieval together, because exact identifiers, codes and clause numbers are the thing semantic search is worst at.
ChunkEmbeddingTermScore
Chunking and metadata
Splits on the document’s own structure and carries effective dates, version and source into every chunk, so a retrieved passage knows when it was true.
ChunkEffective dateVersionSource reference
Security trimming
Applies the caller’s permissions inside the query rather than filtering results afterwards, so the model never sees what the user may not.
PrincipalEntitlementFilter
Graph retrieval
Traverses the knowledge graph for questions that are joins rather than similarity — exposure, genealogy, ownership, causation.
PathHopTraversal
Reranking and citation
Reorders candidates on relevance and returns the passage identifier behind every sentence, so the answer can be checked rather than trusted.
CandidateRankCitation
Edition-bound retrievalthis domain adds
Constrains the corpus to the wording edition attached to the policy before ranking, so the newest edition can never win by relevance.
Wording editionEffective dateClause
What the engineer owns here
Clause-level, not document-level, citation — the difference an auditor cares about.
Where this layer fails
Ranking across editions and trusting the top hit. One cross-edition leak is a coverage answer that is wrong in a way the customer can enforce.
05/07
Agent & orchestration
Triage, coverage determination, fraud signal and settlement recommendation under approval gates.
Takes in
Ranked, permission-trimmed, citable evidence from retrieval.
Hands on
Actions taken or proposed, each with the evidence, the identity and the trace behind it.
SystemWhat it doesKey business entities
Agent runtime
Plans, calls tools and holds the loop. Where the step budget, the timeout and the stopping condition are enforced rather than hoped for.
TaskPlanStepTrace
Tool and MCP servers
Typed, permissioned tools with declared schemas. An agent’s real capability surface is this list, which is why the list is a security artefact.
ToolSchemaScopeInvocation
State and memory
Durable task state with checkpoints, so a run that dies mid-way resumes instead of restarting and re-doing side effects.
CheckpointSessionMemory record
Human-in-the-loop gates
Approval steps on the actions that need one, carrying enough context for the approver to actually decide rather than rubber-stamp.
ApprovalThresholdDecision
Workflow integration
Writes back into the systems people already work in, under a service identity with its own audit trail.
Work itemWrite-backActor
What the engineer owns here
The fast-track boundary: what settles automatically and what always sees an adjuster.
Where this layer fails
An agent that acts under a service account rather than on behalf of the user. It will eventually do something the requesting user had no right to do, and the log will not show it.
06/07
Fairness, guardrails & compliance
Proxy-discrimination testing, NAIC-aligned governance, documentation and action allow-lists.
Takes in
Proposed actions and drafted responses from the agent layer.
Hands on
Permitted, grounded, logged output — or a refusal with a stated reason.
SystemWhat it doesKey business entities
Policy enforcement
The rules that decide whether an action is permitted at all, evaluated before the action and independently of the model that proposed it.
PolicyDecisionObligation
Input and output screening
Injection detection on the way in, and on the way out the checks for leakage, unsupported claims and content the domain forbids.
PromptResponseViolation
Groundedness checking
Verifies each assertion resolves to retrieved evidence, and fails the response rather than shipping the sentence that does not.
AssertionSupportVerdict
Model risk management
Model inventory, intended use, validation evidence and the sign-off that lets a model be used for a purpose. Not optional in a regulated estate.
ModelIntended useValidationApproval
Audit log
Immutable record of what was asked, retrieved, decided and done — the artefact a reviewer reads when they do not take your word for it.
Audit entryActorEvidence reference
What the engineer owns here
Evidence that the model was tested for unfair discrimination before it priced anything.
Where this layer fails
Guardrails that only run on the prompt. Most real failures are on the output side: the unsupported sentence, the leaked field, the claim nobody approved.
07/07
Evaluation & observability
Coverage-accuracy evals, leakage tracking, drift monitoring and cost per claim.
Takes in
Permitted, grounded, logged output from the guardrail layer.
Hands on
Measured quality, cost and latency — and the evidence to change any of the three.
SystemWhat it doesKey business entities
Evaluation harness
Held-out sets and graded runs on every change, so a prompt edit is a measured change rather than a hopeful one.
Eval setCaseScoreBaseline
Regression gate
Blocks a release when quality drops, in CI, on the same evidence for everyone. The difference between a system and a demo.
ThresholdRunGate decision
Tracing
End-to-end spans across retrieval, model and tool calls, so a bad answer can be opened and read rather than argued about.
TraceSpanLatencyToken count
Cost and capacity
Cost per task and per tenant, against throughput and quota. The number that decides whether the pilot can become the rollout.
Cost per taskQuotaThroughput
Drift and feedback
Watches quality against production traffic rather than the test set, and routes real corrections back into the eval suite.
Drift signalFeedbackIncident
What the engineer owns here
Proving the agent reduced leakage rather than merely reduced touch time.
Where this layer fails
Evaluating once, before launch. Quality moves with the data, the model and the traffic, and a system with no live measurement has no idea which of the three moved.
Typical tech stack
Models
- Frontier models for reasoning over wordings
- Document AI for ACORD and schedules
- Vision models for damage assessment
Grounding
- Clause-level vector index
- Precedent claim retrieval
- Coverage ontology
Agents
- Triage and determination agents
- MCP over policy and claims systems
- Referral and approval gates
Data
- Extraction with confidence routing
- Third-party enrichment
- Governed actuarial marts
Assurance
- Coverage-accuracy eval sets
- Fairness and proxy testing
- NAIC-aligned documentation
Runtime
- Carrier tenancy
- Least-privilege identity
- Full decision retention
Constraints that shape the design
Unfair discrimination
Rating and claims outcomes must be testable for proxy discrimination. Design the test set with the actuary, early.
NAIC model bulletin
Adopted in 20+ jurisdictions: expects an AI governance programme, testing and documentation you can produce on request.
Coverage is contractual
A coverage answer without a clause citation is worthless. Retrieval must resolve to clause level.
Event surge
Catastrophe events multiply volume overnight. Capacity and cost per claim must hold under a 10× spike.
04
Part 3 · Getting it done
Where the business and the technology meet
Twelve use cases, each tied to a stage of the chain and the architecture layer that carries it. The labs build the hard ones; the capstones prove you can run one end to end.
Use cases, mapped to the chain and the architecture
Filter by stage or by earned autonomy. Selecting a use case jumps the value chain and the architecture to the stage and layer it depends on.
Value chain stage
Autonomy earned
12 use cases
No use cases match that combination.
The delivery playbook
How a carrier deployment actually runs.
01
Week 1 — pick a claim segment
Choose one product, one claim type, one jurisdiction. Breadth is what kills insurance pilots.
02
Week 2 — get the wordings right
Version the wording corpus and build clause-level retrieval before anything else. Everything downstream depends on it.
03
Weeks 3-5 — label with the adjusters
Build the coverage eval set with the people who make the call today. Their disagreements are the signal.
04
Weeks 6-8 — fairness and fast-track
Run proxy-discrimination testing and set the fast-track boundary using measured accuracy, not appetite.
05
Weeks 9-12 — surge-test and hand over
Prove it holds at 10× volume, then hand over the runbook, evals and the NAIC-aligned documentation pack.
The curriculum
8 modules, 80 taught hours, 40 hands-on labs and 8 assessments — every lab provisioned and graded by the SCIKIQ Agentic AI Playground. Open a module to see its labs.
01The SCDAI delivery model in insuranceHow to land in a carrier, pick a segment and scope something that can actually ship.10 hrs · 5 labs
Delivery mandateSegment selectionValue sizing on leakage and cycle timeStakeholder mapThin-slice scopingRegulatory posture
01
Segment selectionScore four claim segments on volume, complexity and data availability.
02
Leakage sizingQuantify the cost of a coverage error and a cycle-time day.
03
Stakeholder mapMap underwriter, actuary, compliance and IT for one use case.
04
Thin sliceScope a two-week slice for a single product and jurisdiction.
05
Regulatory readSummarise the NAIC bulletin obligations that apply to your slice.
Assessment 01 — Operating model & scoping
02Insurance data foundationsPolicy admin, claims, wordings, submissions and third-party enrichment.10 hrs · 5 labs
Policy & claims modelsACORDWording versioningLoss runsEnrichment dataEntitlements
01
Policy modelModel policy, coverage, endorsement and version relationships.
02
ACORD extractionExtract a structured submission from an ACORD form and a schedule.
03
Wording versioningBuild a corpus where every policy resolves to its exact wording edition.
04
Loss run parsingNormalise three differently formatted loss runs into one structure.
05
Enrichment joinJoin third-party property data and surface provenance conflicts.
Assessment 02 — Insurance data landscape
03Context engineering for policy languageContract language is precise by design. The model must quote it, not summarise it.10 hrs · 5 labs
Clause-level promptingStructured determination outputsExclusion reasoningAmbiguity handlingRefusalContext budget
01
Clause groundingAnswer a coverage question quoting the operative clause verbatim.
02
Determination schemaEmit a structured determination with clause references and confidence.
03
Exclusion chainsHandle an exclusion with a write-back exception correctly.
04
AmbiguityMake the agent escalate genuine wording ambiguity instead of choosing.
05
Context budgetReduce context cost on determinations without losing accuracy.
Assessment 03 — Context engineering
04Grounding on wordings and precedentClause-level retrieval, precedent claims and the citation contract auditors will test.10 hrs · 5 labs
Clause chunkingVersion-aware retrievalPrecedent searchRerankingCitation contractsRetrieval evals
01
Clause chunkingChunk wordings at clause boundaries and measure the accuracy gain.
02
Version-aware retrievalGuarantee retrieval returns the edition attached to the policy.
03
Precedent searchRetrieve comparable settled claims to support a reserve recommendation.
04
RerankAdd reranking and quantify precision on coverage questions.
05
Retrieval evalsBuild a labelled coverage set and gate on recall@k.
Assessment 04 — Retrieval & grounding
05Agent design for claims and underwritingTriage, determination and the fast-track boundary — where autonomy is earned.10 hrs · 5 labs
Workflow vs agentFast-track designReferral gatesMulti-agent claims flowState and memoryContainment
01
Triage agentRoute claims by severity and complexity with stated reasons.
02
Determination agentDetermine coverage and produce a cited, structured decision.
03
Fast-track gateSet an STP boundary from measured accuracy and enforce it in code.
04
Claims pipelineCoordinate intake, determination and settlement agents under a supervisor.
05
ContainmentCap payment authority and prove the cap cannot be argued past.
Assessment 05 — Agent design & autonomy
06Integrating with carrier systemsPolicy admin and claims write-back, identity, and surviving an event surge.10 hrs · 5 labs
MCP over policy/claimsWrite-back safetyIdentityRate limitsSurge capacityError handling
01
MCP over claimsExpose claim read and note-write tools with typed schemas.
02
Safe write-backWrite a reserve recommendation with full audit metadata and idempotency.
03
IdentityScope the agent to the adjuster’s own authority, no more.
04
Surge testHold latency and cost under a 10× catastrophe-event spike.
05
Failure handlingDesign dead-letter and replay for a failing policy admin API.
Assessment 06 — Systems integration
07Fairness, compliance and securityProxy discrimination, NAIC-aligned governance and the attack surface of a claims pipeline.10 hrs · 5 labs
Proxy discrimination testingNAIC governanceDocumentationOWASP LLM Top 10Injection via submitted documentsRecourse design
01
Proxy testingTest a triage agent for disparate outcomes across protected proxies.
02
Governance packProduce NAIC-aligned governance documentation for your agent.
03
Injection defenceDefend against instructions embedded in a submitted claim document.
04
RecourseDesign the appeal path for an automated adverse determination.
05
Red teamRun an OWASP LLM Top 10 pass over the claims pipeline.
Assessment 07 — Fairness, compliance & security
08Production, leakage and costProving the agent reduced leakage — not just touch time — and keeping it that way.10 hrs · 5 labs
Eval suitesLeakage measurementDriftJudge calibrationCost per claimHandover
01
Eval gateGate deployment on coverage-accuracy regression.
02
Leakage studyMeasure leakage before and after, and defend the methodology.
03
Drift watchDetect drift after a wording update and alert on it.
04
Cost per claimMeasure and reduce cost per claim by a third.
05
HandoverProduce the runbook, rollback and documentation pack for go-live.
Assessment 08 — Production & operations
Two capstones — 20 hours
End-to-end agent design, development, deployment and testing, 10 hours each. Reviewed by a Senior SCDAI Engineer against a published rubric.
Capstone A — Coverage determination agent
Build an agent that determines coverage for a claim against the correct wording edition, cites the operative clauses and exclusions, and escalates genuine ambiguity. Prove accuracy on a labelled set.
Capstone B — Submission-to-quote pipeline
Build a pipeline that ingests a broker submission, pre-fills the risk, checks appetite and guidelines, and produces a referral rationale — with proxy-discrimination testing evidence.
What you deliver, and how it is marked
- A running agent, deployed on the program's target stack, reachable by a reviewer.
- An architecture note: boundaries, tools, autonomy level and the trade-offs you took.
- A versioned eval set with a baseline and your final scores.
- A trace walkthrough of one success and one deliberate failure.
- Cost per task, measured — not estimated.
- A runbook and rollback path a client team could operate on Monday.
Problem fit15%Is this a decision worth automating, scoped to a defensible boundary?
Architecture20%Right pattern for the constraints; autonomy earned, not assumed.
Grounding & accuracy20%Measured groundedness against the eval set, with the failures named.
Safety & compliance20%Domain controls actually enforced, and demonstrably so.
Operability15%Traces, runbook, rollback, cost — could someone else run this?
Communication10%Can a client executive follow the reasoning in ten minutes?
One lab, start to finish
Every lab in this program follows the shape below. This is lab 04 in full — the brief you are given, the environment that is provisioned for you, the code you start from and the assertions that decide whether you passed.
Module 04 · Grounding on wordings and precedent
Version-aware clause retrieval
Guarantee a coverage answer resolves to the wording edition attached to that policy.
The brief
The corpus holds four editions of one wording, differing in a single exclusion. Given a claim and its policy, retrieve the operative clause from the correct edition and answer the coverage question. Answering from the newest edition is the failure this lab is built to catch.
Provisioned for you
- 4 wording editions, clause-chunked
- Policy register mapping policy to edition
- Vector + BM25 hybrid index
- 30 labelled coverage questions across editions
You start from this
def coverage_answer(policy_id, question, index, registry):
"""Answer grounded in the wording edition bound to this policy."""
edition = registry.edition_for(policy_id) # never assume 'latest'
# constrain retrieval to `edition` BEFORE ranking, not after
raise NotImplementedErrorWhat the grader asserts
- Never returns a clause from an edition other than the one bound to the policy
- Cites clause identifier and edition, not just the document
- Correctly applies the write-back exception in edition 3 that restores cover
- Escalates rather than answering when the question is genuinely ambiguous
- Recall@5 at or above 0.9 on the labelled set
05
Assessment
Assessed, not attended
Every module ends with a timed, randomised assessment delivered through the SCIKIQ Agentic AI Playground. The certificate requires a pass on all of them plus two reviewed capstones.
Assessment blueprint
What each test covers, how long it runs, and how many items are currently in the versioned bank behind it.
| # | Assessment & coverage | Items | Time | In bank |
|---|---|---|---|---|
| 01 | Operating model & scopingDelivery mandateSegment selectionValue sizingRegulatory posture | 25 | 35 min | 4 |
| 02 | Insurance data landscapePolicy and claims dataACORD and submissionsWording versioningEnrichment | 25 | 35 min | 4 |
| 03 | Context engineeringClause-level groundingStructured outputsAmbiguity and escalationCost | 25 | 35 min | 4 |
| 04 | Retrieval & groundingClause chunkingVersion fidelityPrecedent retrievalRetrieval metrics | 25 | 40 min | 4 |
| 05 | Agent design & autonomyTriage and determinationFast-track boundariesMulti-agent flowContainment | 25 | 40 min | 4 |
| 06 | Systems integrationMCP tool designWrite-back safetyIdentitySurge and resilience | 25 | 35 min | 4 |
| 07 | Fairness, compliance & securityProxy discriminationNAIC expectationsLLM securityRecourse | 30 | 45 min | 4 |
| 08 | Production & operationsEval gatingLeakage measurementDriftUnit economics and handover | 30 | 45 min | 4 |
Single best answerOne defensible option among plausible distractors drawn from real field mistakes.
Scenario judgementA client situation with constraints; you pick the action a Senior SCDAI Engineer would take.
Architecture selectionGiven non-functional requirements, choose the pattern and justify the trade-off.
Failure diagnosisA trace, an eval report or a cost curve — identify the root cause.
- Each module ends with a timed assessment; the program certificate requires a pass on every one.
- Items are drawn at random from a versioned bank, so no two learners sit an identical paper.
- Pass mark is 70%. Two retakes are included, each with a fresh draw.
- Scenario items carry double weight — they are the ones that predict field performance.
- Capstones are reviewed against a published rubric by a Senior SCDAI Engineer, not auto-graded.
Try four questions from the bank
Real items, drawn from 32 in this program's bank — weighted toward the scenario and diagnosis types, because those are the ones that predict field performance. Instant feedback, nothing saved.
Four sample items — one attempt each, then the reasoning is shown.
Q1Why is segment selection the first decision in a carrier deployment?
One product, one claim type, one jurisdiction. Every additional dimension multiplies the wording, data and regulatory surface.
Q2Why is wording versioning critical in insurance retrieval?
Every policy attaches to a specific wording edition. Retrieval must resolve to that edition, not to the newest or most similar one.
Q3A wording is genuinely ambiguous. What should a well-designed agent do?
Resolving ambiguity is a human judgement with legal consequences. Surfacing it is the agent’s job; deciding it is not.
Q4Why chunk wordings at clause boundaries rather than fixed token windows?
Chunking should follow document structure. A clause split across chunks retrieves partially and cites incorrectly.
06
Career & outcomes
What this makes you, and where it goes next
The same ladder whichever specialisation you enter through — what changes is the domain you go deep in. Below: how the program is delivered, the skills it moves, the roles it leads to, and the specialisations closest to this one.
Available two ways
The same labs, assessments and capstones, delivered to an enterprise cohort or to individual professionals.
B2B
Direct to enterprise
Cohorts of 20 to 2,000+ on your own tenancy, with your data patterns and your cloud. Skill-gap baselining up front, per-team mastery reporting throughout, and capstones scoped against your real backlog so the output is deployable work.
- Private cohort on your tenancy
- Baseline skills assessment before kick-off
- Capstones scoped to your backlog
- Per-team mastery dashboards for L&D
- Optional Senior SCDAI Engineer architecture reviews
B2C
Direct to individual professionals
The same labs, assessments and capstones for individual engineers and analysts, run on shared infrastructure with a fixed cohort calendar. You leave with a graded portfolio, not a certificate of attendance.
- Fixed-date cohorts, self-paced within them
- Same graded labs and item bank
- Portfolio-ready capstones you own
- Certificate on a pass in every module
- Community and office-hours access
Skill matrix
Find your row and aim one column right. The Playground scores you against this after every module.
| Skill | Beginner | Intermediate | Advanced |
|---|---|---|---|
| Insurance domain fluency | Knows the products. | Maps the chain to decisions and owners. | Sizes leakage and designs the operating change. |
| Coverage grounding | Retrieves a document. | Retrieves the right clause and edition. | Designs version-safe, auditable citation. |
| Fairness engineering | Aware of bias risk. | Runs proxy-discrimination tests. | Designs testable fairness into the build. |
| Context engineering | Writes clear prompts. | Structures retrieval, tools and state deliberately. | Designs context strategy for reliability and cost at scale. |
| Retrieval & grounding | Builds basic vector search. | Tunes chunking, hybrid search and reranking. | Designs graph + vector grounding with measured recall. |
| Agent orchestration | Runs a single tool-calling agent. | Builds supervised multi-step and multi-agent flows. | Designs autonomy boundaries and failure containment. |
| Tool & system integration | Calls a documented API. | Writes an MCP server over a system of record. | Designs a least-privilege tool estate across systems. |
| Evaluation | Eyeballs outputs. | Builds labelled eval sets and regression gates. | Runs online evals with drift and judge calibration. |
| Observability & cost | Reads logs. | Traces runs, tracks tokens and latency. | Owns cost per task and capacity planning in production. |
| Security & guardrails | Adds output filters. | Mitigates the OWASP LLM Top 10 in a build. | Threat-models an agent estate and proves controls. |
| Client delivery | Takes notes in a workshop. | Runs discovery and scopes a thin slice. | Owns the account technically, from scope to handover. |
Where it takes you
The SCDAI ladder is the same whichever specialisation you enter through — what changes is the domain you go deep in.
01
Data and AI EngineerEntry
Python, LLM APIs, first RAG build, prompt and context basics.
02
Associate Data and AI EngineerPractitioner
Client-facing discovery, tool integration, agent build under supervision.
03
SCIKIQ Certified Data and AI EngineerSpecialist
Owns a deployment end to end — scoping, build, evals, handover.
04
Senior Data and AI Engineer / Domain LeadSenior
Multi-agent architecture, regulated deployments, reference patterns for the practice.
05
AI Architect / Practice LeadLeadership
Portfolio of accounts, delivery standards, hiring and capability strategy.
Terms worth knowing
- FNOL
- First Notice of Loss — the moment a claim enters the carrier.
- GWP
- Gross Written Premium — the top-line volume measure.
- Leakage
- Money paid on a claim that should not have been paid.
- STP
- Straight-through processing — a claim or quote completed with no human touch.
- ACORD
- The standard form and data family used in insurance submissions.
- Proxy discrimination
- Unfair outcomes produced by variables that stand in for protected characteristics.
- NAIC model bulletin
- The US regulatory expectation for governance of insurer AI use.
- SCDAI
- SCIKIQ Certified Data and AI Engineer — the credential this program awards.
- SCDAI engineer
- An engineer embedded with the customer who turns their constraints into a working system.
- Thin slice
- The narrowest end-to-end path through a workflow that still produces business value.
- Agentic RAG
- Retrieval where the agent decides when, what and how often to retrieve.
- MCP
- Model Context Protocol — open standard connecting an agent to tools and data.
- A2A
- Agent2Agent — protocol for agent-to-agent delegation across vendors.
- Eval set
- A labelled, versioned set of cases an agent must pass before release.
- LLM-as-judge
- Using a model to score outputs, calibrated periodically against human ratings.
- Guardrail
- An enforced constraint on an input, output or action — not a prompt request.
- Human-in-the-loop
- A required human approval step before a consequential action executes.
- Cost per task
- Total token, tool and compute cost of one completed unit of work — the engineer's unit economic.
- Golden path
- The reference architecture a practice standardises on so deployments stay reviewable.
- Handover
- The point at which the client can operate, evaluate and change the system without the engineer.
Related programs
Skill a team, or join a cohort
B2B cohorts run on your tenancy with capstones scoped to your backlog. B2C cohorts run on a fixed calendar.
07
Enrolling
What you need, what you get, and what it costs
Stated plainly enough to rule yourself in or out without a sales call: the prerequisites, how the program runs, exactly what the credential is worth, and the questions everyone asks.
Before you start
Stated plainly so you can rule yourself in or out without a sales call. Nothing here is a formal qualification — it is what the first lab assumes you can already do.
Technical
You should already be able to do these
- Comfortable in Python: functions, typing, virtual environments, reading a stack trace
- Have called an LLM API and handled a structured response
- Git, the command line, and reading someone else’s code without panic
Domain
What we assume, and what we teach
- Know the difference between a policy, a claim and a premium
- No underwriting or actuarial background needed — the chain is taught from the start
Time and access
What the program asks of your week
- A machine that can run a browser and a terminal — labs run in the Playground, not locally
- Roughly 8 hours a week for the taught weeks, plus 20 hours for the two capstones
How it runs
Cohort dates and pricing are confirmed on enquiry rather than printed here, because both move with the intake.
- Format
- Self-paced within a fixed cohort window, delivered in the SCIKIQ Agentic AI Playground
- Taught hours
- 80 hours across 8 modules
- Capstones
- 20 hours — two end-to-end builds, reviewed not auto-graded
- Typical duration
- 10–12 weeks at around 8 hours a week
- Level
- Intermediate → Advanced
- Built for
- Engineers and analysts deploying AI at carriers, brokers and MGAs
- Next cohorts
- — confirmed on enquiry
- Pricing
- — confirmed on enquiry
The credential
The credential is awarded per specialisation, so it names the domain or stack you were assessed in rather than claiming general competence. On this program the badge reads SCDAI — Insurance.
How it is earned
- A pass at 70% or above on every module assessment — not an average across them
- Both capstones reviewed against the published rubric by a Senior SCDAI Engineer
- All labs submitted and machine-graded green
- Awarded as
- A verifiable digital badge with a public credential record
- Valid for
- Two years from award
- Recertification
- One current-year assessment plus a refreshed capstone
- Retakes
- Two included per assessment, each with a fresh item draw
- Named specialisation
- The badge states the domain or stack, e.g. SCDAI — Healthcare
- Verification
- Third parties confirm a holder from the credential record, no login required
A certificate that cannot be checked is decoration. Every award resolves to a record showing the specialisation, the award date and the assessments passed.
The certificate you leave with
This is the credential the program awards, shown exactly as it is issued — with the specialisation named, the assessment record attached and a verification link anyone can check without an account.
SCIKIQ Data Academy
Certificate of Professional Certification
This is to certify that
Your name
has been assessed and certified as
SCIKIQ Certified Data and AI Engineer
Insurance
- Specialisation
- Insurance
- Assessments passed
- 8 of 8, each at 70% or above
- Labs graded
- 40 of 40 machine-graded green
- Capstones reviewed
- 2, against the published rubric
- Taught hours
- 80 across 8 modules, plus 20 capstone hours
AwardedOn completion
Valid untilTwo years from award
Credential IDSCDAI-INSURANCE-<issued>
Senior SCDAI Engineer, Certification Board
- 01The specialisation is on the certificate
Not “Data and AI Engineer” but the domain or stack you were actually assessed in. A general claim would be a weaker one.
- 02The assessment record travels with it
Modules passed, labs graded and both capstones reviewed — so the credential states what was measured rather than that you attended.
- 03Anyone can verify it, without a login
The credential ID resolves to a public record showing the specialisation, the award date and the assessments passed.
Share it on LinkedIn
Add it to your LinkedIn profile in one step. The link pre-fills the certification fields from the credential record, so the entry on your profile matches the record a reader can check.
Add to LinkedIn profile The button is live on your real certificate; here it opens LinkedIn pre-filled with this specialisation so you can see exactly what the profile entry will say.A certificate that cannot be checked is decoration. Every award resolves to a record showing the specialisation, the award date and the assessments passed.
Questions we get asked
The objections that come up in every conversation about this program, answered without the brochure voice.
Is this a course or a certification?
Both, and the second is the point. Eight timed assessments and two reviewed capstones stand between you and the credential, so a pass means someone measured the skill rather than recorded your attendance.
How is it self-paced without falling apart?
Every lab is provisioned, graded and unblocked by the Playground rather than by an instructor. That is what lets a cohort of 2,000 cost the same faculty time as a cohort of 20 — and why you are never waiting on someone to mark your work.
What if I fail an assessment?
Two retakes are included per assessment, each drawing a fresh item set from the bank, so a retake is a genuinely new paper rather than the same questions again.
Do I need my own cloud account?
For the domain specialisations, no — labs run in provisioned sandboxes. For the four tech-stack programs you will want access to that platform, since deploying into a real subscription is much of the point.
Can we run this privately for our team?
Yes. B2B cohorts run on your own tenancy with your data patterns, a skills baseline before kick-off, per-team mastery reporting, and capstones scoped against your actual backlog so the output is deployable work rather than an exercise.
Which specialisation should I take?
Take the domain you deploy into. If you move across industries, take a tech-stack program instead and pick up domain context on the engagement. The chooser on the programs page will narrow it.
How current is the material?
The trends, platform capabilities and regulatory positions are reviewed each quarter, and every external claim on these pages links to its source so you can check the date yourself.
What do I actually leave with?
A graded portfolio: forty machine-graded labs, two reviewed end-to-end agent builds with measured evaluation and cost per task, and a verifiable credential naming your specialisation.
Still deciding?
Tell us the systems you deploy into and we will say plainly whether this specialisation is the right one — or which of the 21 is.