McKinsey~25% of spend
Administration is the value pool
Administrative activity accounts for roughly 25% of total healthcare spending — the single largest addressable inefficiency, and reachable without clinical risk.
Program 05 · Domain specialisation
SCIKIQ Certified Data and AI EngineerHealthcare
Healthcare AI fails for boring reasons: PHI scope, EHR integration timelines and a clinician who will not use a tool that adds a click. Administrative work is roughly a quarter of healthcare spending — that is the value pool, and it is reachable without touching clinical decision-making.
- 80 hrstaught
- 40hands-on AI labs
- 8assessment tests
- +20 hrs2 E2E capstones
~25%Of total healthcare spend consumed by administrative activityMcKinsey
30-60%Potential reduction in cost to collect from AI-enabled revenue cycleMcKinsey
up to 40%Reduction in payer administrative costBCG
60-90 minClinician time returned per day by ambient documentation
01
Part 1 · The business
The value chain, the people, the trends
Healthcare has two intertwined chains: the care a patient receives and the revenue that pays for it. Almost all deployable AI value in 2026 sits in the second one and the administration around the first.
The care and revenue value chain
Six stages from access to compliance. Select one to see the decisions and the pain.
01
Access & intake
Scheduling, registration, eligibility and patient intake.
Decisions made here
- When and where can this patient be seen?
- Is this patient eligible?
- What information do we still need?
Where it breaks today
- Phone queues and no-shows
- Eligibility checked manually
- Intake data re-entered by staff
What is measured
- Time to third-next-available
- No-show rate
- Registration errors
Where AI landsConversational scheduling, automated eligibility checks and structured intake capture.
Time to third-next-availableNo-show rateRegistration errors
02
Authorise & verify
Prior authorisation, medical necessity documentation and payer communication.
Decisions made here
- Does this require authorisation?
- What evidence supports necessity?
- Is this approvable as submitted?
Where it breaks today
- Authorisation cycles measured in weeks
- Requirements differing by payer and plan
- Denials for missing documentation that existed all along
What is measured
- Authorisation turnaround
- Denial rate
- Peer-to-peer volume
Where AI landsRequirement retrieval per payer, evidence assembly from the chart and drafted submissions.
Authorisation turnaroundDenial ratePeer-to-peer volume
03
Document & code
Clinical documentation, coding, CDI and chart completion.
Decisions made here
- What happened in this encounter?
- What codes are supported?
- What documentation is missing?
Where it breaks today
- Documentation done after hours
- Coding queries that interrupt clinicians
- Under- and over-coding risk in both directions
What is measured
- Documentation time per encounter
- Query rate
- Coding accuracy
Where AI landsAmbient documentation, code suggestion with evidence and gap detection before submission.
Documentation time per encounterQuery rateCoding accuracy
04
Deliver care
Care coordination, referrals, discharge and follow-up.
Decisions made here
- What is the plan?
- Who needs to act next?
- Is this patient at risk of readmission?
Where it breaks today
- Discharge summaries delayed
- Referral loops that never close
- Care gaps identified retrospectively
What is measured
- Length of stay
- Readmission rate
- Referral closure
Where AI landsSummarisation, care-gap surfacing and coordination agents — administrative, not diagnostic.
Length of stayReadmission rateReferral closure
05
Bill & collect
Claim submission, denial management, appeals and patient collections.
Decisions made here
- Is this claim clean?
- Is this denial worth appealing?
- What will actually collect?
Where it breaks today
- Denials worked manually at volume
- Appeal letters rewritten from scratch
- Cost to collect rising faster than collections
What is measured
- Cost to collect
- Denial rate
- Days in AR
- Clean claim rate
Where AI landsDenial classification, appeal drafting with chart evidence and touchless claim workflows.
Cost to collectDenial rateDays in ARClean claim rate
06
Quality, risk & compliance
Quality reporting, risk adjustment, audit and privacy compliance.
Decisions made here
- Is this documented to the standard?
- What is our risk exposure?
- Can we evidence this?
Where it breaks today
- Abstraction done by hand for quality measures
- Audit evidence assembled reactively
- PHI access impossible to attest for
What is measured
- Measure performance
- Audit findings
- Access attestation coverage
Where AI landsChart abstraction, evidence assembly and access-log reasoning — all under minimum necessary.
Measure performanceAudit findingsAccess attestation coverage
Who you are building for
Clinicians will abandon anything that adds a click. Build for the workflow they already have. Select one to light the stages they own.
What the strategy houses are seeing
The administrative value pool is the consensus 2026 answer.
McKinsey30-60% cost to collect
The touchless revenue cycle is the race
AI-enabled revenue cycle could cut cost to collect by 30-60% with faster cash realisation and staff redeployed to patient-facing work.
BCGup to 40% payer admin
Payers are going AI-first on admin
Payer administrative costs could fall by up to 40%, though cost reduction alone will not protect incumbents from AI-native entrants.
Industry60-90 min/clinician/day
Ambient documentation is the proven ROI
Ambient clinical documentation returns 60-90 minutes of clinician time per day — the fastest measurable win in most provider settings.
CMSFHIR PA by Jan 2027
FHIR prior authorisation is now a deadline
CMS-0057-F requires FHIR-based prior authorisation APIs by 1 January 2027, and any AI in that workflow must be transparent, auditable and documented.
HBRWorkflow > accuracy
Clinician adoption is a workflow problem
Tools that add a step get abandoned regardless of accuracy. Integration into the existing workflow predicts adoption better than model quality.
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.
- McKinseyWhat to expect in US healthcare in 2026 and beyondAdministrative activity as ~25% of healthcare spend
- McKinseyAgentic AI and the race to a touchless revenue cycle30-60% reduction in cost to collect
- BCGAI for Healthcare Payers: The AI-First StrategyUp to 40% reduction in payer administrative cost
- IndustryHIPAA Compliance for AI Agents: PHI Rules and ControlsMinimum necessary and BAA obligations for agents
02
Part 2 · Data and meaning
Domains, ontology, taxonomy and governed terms
Healthcare architecture is defined by two things: FHIR as the integration substrate, and the fact that every byte you touch may be PHI. Minimum necessary is an architecture constraint, not a policy statement.
The data landscape
Six families. The chart is the source of truth and the hardest thing to query well.
EHR / clinical
Encounters, notes, orders, results and problem lists — via FHIR where you are lucky.
Epic, Oracle Health, athenaClaims & billing
837/835 transactions, remittances, denials and AR.
X12 837/835, clearinghousePayer rules
Coverage policies, medical necessity criteria and plan-specific requirements.
Payer portals, policy PDFsTerminology
The code systems that make clinical data comparable at all.
SNOMED CT, ICD-10, CPT, LOINC, RxNormScheduling & access
Slots, referrals, waitlists and patient communication history.
Scheduling module, CRMUnstructured clinical text
Notes, letters, faxes and scanned records. Where most of the evidence is.
Notes, document imagingThe semantic layer
Healthcare has the richest standard vocabularies in any industry and the worst record of using them consistently. Terminology binding is what separates an agent that can read a chart from one that merely appears to.
Data domains
Six subject areas, each with the entities it holds, its critical data elements and the function accountable for it.
Patient & coverage
Who the person is, and who pays for their care.
PatientCoveragePayerPlanGuarantor
- Critical data elements
- Medical record number; Member identifier; Plan identifier
- Owned by
- Patient access
Encounter & care
Every interaction with the health system and what happened in it.
EncounterEpisode of careCare teamLocation
- Critical data elements
- Encounter identifier; Admission and discharge dates; Disposition
- Owned by
- Clinical operations
Clinical
The clinical record: problems, observations, medications, procedures.
ConditionObservationMedicationProcedureAllergy
- Critical data elements
- SNOMED concept; LOINC code; RxNorm code
- Owned by
- Clinical informatics
Order & authorisation
What was requested, and whether the payer agreed to it.
OrderReferralPrior authorisation
- Critical data elements
- Authorisation number; Requested service code; Decision date
- Owned by
- Utilisation management
Revenue cycle
Charges, claims, remittances, denials and appeals.
ChargeClaimRemittanceDenialAppeal
- Critical data elements
- Claim number; CPT/HCPCS code; Denial reason code
- Owned by
- Revenue cycle
Quality & risk
Measures, risk adjustment and the documentation supporting them.
Quality measureRisk scoreCare gap
- Critical data elements
- Measure identifier; HCC code; Numerator/denominator flags
- Owned by
- Quality & population health
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
Patient
The person receiving care.
- Patient has Coverage
- Patient participates in Encounter
Core entity
Coverage
The insurance arrangement paying for care.
- Coverage issued by Payer
- Patient has Coverage
Core entity
Payer
The organisation financially responsible.
- Coverage issued by Payer
- Prior authorisation decided by Payer
- Claim submitted to Payer
Event
Encounter
A single interaction between patient and provider.
- Encounter documents Condition
- Encounter includes Procedure
- Encounter records Observation
- Encounter generates Order
- Encounter billed as Claim
- Patient participates in Encounter
- Provider delivers Encounter
Core entity
Provider
The clinician or organisation delivering care.
- Provider delivers Encounter
Reference / classification
Condition
A diagnosed problem, coded to a terminology.
- Encounter documents Condition
- Procedure justified by Condition
Event
Procedure
An intervention performed.
- Procedure justified by Condition
- Encounter includes Procedure
Event
Observation
A measurement or finding.
- Encounter records Observation
Event
Order
A request for a service or medication.
- Order requires Prior authorisation
- Encounter generates Order
Event
Prior authorisation
Payer approval required before delivery.
- Prior authorisation decided by Payer
- Order requires Prior authorisation
Event
Claim
A billed request for reimbursement.
- Claim submitted to Payer
- Claim may receive Denial
- Encounter billed as Claim
Event
Denial
A payer refusal of all or part of a claim.
- Claim may receive Denial
Taxonomies
The classification hierarchies that make records comparable across systems.
Clinical terminology
- Terminology
- Hierarchy
- Concept
SNOMED CT → Disorder of cardiovascular system → Atrial fibrillation
Service & billing codes
- Code set
- Section
- Code
CPT → Surgery → 33208 pacemaker insertion
Denial reason
- Category
- Group code
- Reason code
Administrative → CO → CO-197 authorisation absent
Governed business terms
The definitions an agent must use rather than invent. Most wrong answers in this industry are a term used loosely.
- Encounter
- One interaction between patient and provider. An admission spanning days is one encounter, not many.
- Episode of care
- A clinically related series of encounters treated as one unit for outcomes and payment.
- Coverage
- The insurance arrangement in force for a patient on a date — it can change mid-episode.
- Medical necessity
- The documented clinical justification a payer requires before it will pay.
- Clean claim
- A claim with no defect that would cause it to be denied or pended on submission.
- Care gap
- A recommended service the patient is eligible for and has not received.
- Minimum necessary
- The HIPAA limit restricting access to the PHI fields a task actually requires.
- Cost to collect
- Total revenue-cycle cost divided by cash collected — the CFO’s efficiency metric.
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, with a PHI boundary drawn explicitly through the middle.
01/07
Sources & FHIR
EHR access via FHIR APIs, claims transactions and payer policy sources.
Takes in
Nothing. This is the upstream edge.
Hands on
Field-scoped resource reads — only the elements a declared task needs, never the full resource.
SystemWhat it doesKey business entities
Electronic health record
The clinical system of record, exposed to you as FHIR resources rather than as its internal schema. Returns far more than any one task needs.
PatientEncounterConditionObservationMedication
Practice management and ADT
Registration, scheduling and the admit/discharge/transfer stream. ADT is the heartbeat that tells you a patient state changed at all.
EncounterLocationEpisode of care
Revenue cycle and claims
Charge capture, coding, claim submission, remittance and denials. Where administrative cost concentrates and where most automation value sits.
ChargeClaimRemittanceDenialAppeal
Payer and eligibility services
Coverage checks, benefits and prior authorisation, over X12 transactions and payer portals of varying quality.
CoveragePayerPlanPrior authorisation
Clinical documents and imaging
Notes, discharge summaries, referrals and reports. The narrative that structured fields do not carry, and the highest-risk data on the page.
Clinical noteReferralReport
Consent and identity
Patient matching, the master person index and the consent directives that decide whether a record may be used for a given purpose at all.
Patient identityConsentGuarantor
What the engineer owns here
Realistic EHR integration timelines — typically 8-16 weeks, and never on your critical path.
Where this layer fails
Reading whole FHIR resources because it is easier. Forty fields arrive, six are needed, and the other thirty-four are now in your logs and your traces.
02/07
Ingestion & de-identification
Landing PHI under BAA, with de-identification and tokenisation where the workload allows.
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
Deciding what genuinely needs PHI and stripping everything that does not.
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 & minimum necessary
Field-level entitlements, purpose-of-use, terminology mapping and access attestation.
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
Consent and purpose servicethis domain adds
Resolves whether this patient’s record may be used for this purpose right now, and is consulted before retrieval rather than after.
Consent directivePurpose of useRestriction
What the engineer owns here
Proving the agent saw only the fields its task required — the HIPAA question that decides deployment.
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
Chart retrieval scoped to the encounter, payer-policy retrieval and terminology-aware search.
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
What the engineer owns here
Scoping retrieval to the encounter rather than the patient’s whole life.
Where this layer fails
Filtering results after retrieval instead of constraining the query. The model has already seen the rows you removed, and it will use them.
05/07
Agent & orchestration
Administrative agents: intake, authorisation, coding support, denial and appeal workflows.
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
Keeping the agent on the administrative side of the clinical line, deliberately and visibly.
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
Privacy, safety & guardrails
PHI redaction, BAA-covered processing, clinical-scope refusal and audit trail.
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
Minimum-necessary enforcementthis domain adds
Field-level scoping tied to the declared task, dropping everything else before it can reach the model, the logs or the traces.
Declared taskAllowed fieldAttestation
What the engineer owns here
Hard refusal on clinical advice, and an audit trail that satisfies a privacy officer.
Where this layer fails
PHI reaching a trace or a log. The model never saw it, the observability stack did, and that is still a disclosure.
07/07
Evaluation & observability
Accuracy evals per workflow, denial-overturn tracking, drift and cost per encounter.
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
Tying the agent to an outcome metric a CFO already tracks.
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
- BAA-covered model endpoints only
- Ambient speech models for documentation
- Smaller models for classification
Grounding
- Encounter-scoped chart retrieval
- Payer policy index
- SNOMED/ICD terminology service
Agents
- Administrative workflow agents
- MCP over FHIR resources
- Human review gates
Data
- FHIR R4 integration
- X12 claims processing
- De-identification pipeline
Assurance
- Workflow accuracy evals
- Minimum-necessary attestation
- Immutable access audit
Runtime
- BAA-covered cloud
- Field-level entitlements
- PHI-safe logging
Constraints that shape the design
HIPAA minimum necessary
An agent may access only the PHI fields its task requires. Whole-chart retrieval is a finding waiting to happen.
BAAs everywhere
Every model endpoint, vector store and observability tool that touches PHI needs a BAA. This eliminates most default stacks.
The clinical line
Administrative support is deployable now. Clinical decision support is a regulated device question. Draw the line in code.
EHR integration reality
FHIR makes Epic and Oracle Health tractable, not fast. Plan 8-16 weeks and design so nothing blocks on it.
04
Part 3 · Getting it done
Where the business and the technology meet
Every use case here stays on the administrative side of the clinical line — which is precisely why they deploy. They join a stage of the chain to the layer that does the work.
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 provider or payer deployment actually runs.
01
Week 1 — start on the revenue side
Pick an administrative workflow with a CFO-visible metric. Clinical ambition is what stalls healthcare deployments in legal review.
02
Week 2 — settle PHI scope first
Agree the minimum-necessary field list and the BAA position before writing code. This is the gate everything else waits on.
03
Weeks 3-6 — build around the EHR, not through it
Assume FHIR integration takes 8-16 weeks. Build against a conformant sandbox so nothing blocks.
04
Weeks 7-9 — label with the people who do the work
Coders, authorisation specialists and adjusters build the eval set. Their edge cases are the test.
05
Weeks 10-12 — attest and hand over
Produce the access attestation, audit trail and runbook. A privacy officer who is satisfied is the actual go-live gate.
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 healthcareWhere to land, which line not to cross, and how to scope past legal review.10 hrs · 5 labs
Delivery mandateAdministrative vs clinical scopeValue sizing on cost to collectStakeholder mapLegal and privacy pathThin-slice scoping
01
Workflow selectionScore four administrative workflows on value, data access and risk.
02
Clinical lineWrite the explicit scope boundary for an agent and encode a refusal for it.
03
Value sizingQuantify cost to collect and denial cost for a service line.
04
Privacy pathDraft the PHI scope and BAA position for a named use case.
05
Thin sliceScope a two-week slice that does not block on EHR integration.
Assessment 01 — Operating model & scope
02Healthcare data foundationsFHIR, X12, terminology and the unstructured chart.10 hrs · 5 labs
FHIR R4 resourcesHL7 v2X12 837/835SNOMED/ICD/CPT/LOINCChart structureConsent and entitlements
01
FHIR traversalQuery encounters, conditions and observations from a FHIR sandbox.
02
Claims parsingParse an 837 and match it to an 835 remittance with denial codes.
03
Terminology mappingMap free-text findings to SNOMED and expand for retrieval.
04
Chart scopingBuild encounter-scoped retrieval that cannot reach the whole record.
05
EntitlementsImplement field-level entitlements with purpose-of-use enforcement.
Assessment 02 — Healthcare data landscape
03Context engineering for clinical textClinical language is abbreviated, negated and templated. All three break naive prompting.10 hrs · 5 labs
Negation and uncertaintyAbbreviation ambiguityStructured outputsPHI-safe promptingOmission riskRefusal design
01
Negation handlingBuild extraction that does not turn “no chest pain” into chest pain.
02
Abbreviation disambiguationResolve ambiguous clinical abbreviations using context.
03
Structured noteEmit a schema-valid encounter note with evidence spans.
04
PHI-safe promptsDesign prompts and logs that never leak identifiers.
05
Omission testBuild an eval that specifically detects omitted findings.
Assessment 03 — Context engineering
04Grounding on charts and payer policyTwo retrieval problems: a patient chart and a shifting library of payer rules.10 hrs · 5 labs
Encounter-scoped retrievalPayer policy indexingTerminology-aware searchCitation contractsFreshnessRetrieval evals
01
Scoped retrievalRetrieve within an encounter boundary and prove the boundary holds.
02
Payer policy indexIndex payer criteria and retrieve the right plan-specific rule.
03
Terminology expansionImprove recall using SNOMED hierarchy expansion.
04
Evidence citationCite the exact chart span supporting each extracted fact.
05
Retrieval evalsBuild a labelled necessity-evidence set and measure recall.
Assessment 04 — Retrieval & grounding
05Agent design for administrative workflowsAuthorisation, coding support and denial management — multi-step, high-volume, reviewable.10 hrs · 5 labs
Workflow vs agentReview gatesQueue-working agentsMulti-agent RCMStateContainment
01
Authorisation agentAssemble evidence and draft a submission for one payer criterion set.
02
Denial agentClassify denials and draft appeals from chart evidence.
03
Review gateRequire clinician sign-off before any note is filed.
04
RCM pipelineCoordinate classification, evidence and drafting agents.
05
ContainmentPrevent an agent from acting outside its assigned queue.
Assessment 05 — Agent design & autonomy
06Integrating with the EHRFHIR write-back, SMART on FHIR launch, and living inside a clinician’s existing screen.10 hrs · 5 labs
SMART on FHIRWrite-back safetyIdentity and contextRate limitsWorkflow embeddingError handling
01
MCP over FHIRExpose FHIR read tools with typed schemas and scope limits.
02
SMART launchLaunch an app in patient context and inherit the user’s permissions.
03
Safe write-backWrite a draft note without filing it, and prove it cannot auto-file.
04
Rate limitsHandle EHR API throttling without losing queue position.
05
EmbeddingDesign the interaction so it removes a click rather than adding one.
Assessment 06 — EHR integration
07HIPAA, safety and securityMinimum necessary in code, BAAs across the stack, and the attack surface of a chart pipeline.10 hrs · 5 labs
Minimum necessaryBAA coverageDe-identificationOWASP LLM Top 10Injection via clinical documentsAccess attestation
01
Minimum necessaryRestrict an agent to a field list and prove it cannot exceed it.
02
De-identificationBuild and test a de-identification pipeline with re-identification checks.
03
BAA mappingMap every PHI-touching component to its BAA status.
04
Injection defenceDefend against instructions embedded in a scanned referral letter.
05
AttestationProduce an access attestation report a privacy officer would accept.
Assessment 07 — HIPAA, privacy & security
08Production, outcomes and costTying the agent to a metric the CFO already reports, then keeping it honest.10 hrs · 5 labs
Eval suitesOutcome measurementDriftJudge calibrationCost per encounterHandover
01
Eval gateGate deployment on extraction and evidence-recall regression.
02
Outcome studyMeasure denial overturn or cost to collect before and after.
03
Drift watchDetect drift after an annual coding update.
04
Cost per encounterMeasure and reduce cost per encounter.
05
HandoverProduce runbook, rollback and attestation 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 — Prior authorisation agent
Build an agent that retrieves payer-specific criteria, assembles cited chart evidence and drafts a submission for review — with a minimum-necessary attestation for every field it touched.
Capstone B — Denial management agent
Build an agent that classifies denials, judges appeal viability and drafts appeals with chart citations. Measure overturn rate against a labelled historical set.
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 07 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 07 · HIPAA, safety and security
Minimum necessary enforcement
Prove an agent cannot reach PHI fields its task does not require.
The brief
A prior-authorisation agent needs six fields. The FHIR sandbox returns forty. Build the access layer so the extra thirty-four never reach the model, the logs or the traces — then produce the attestation a privacy officer would accept.
Provisioned for you
- FHIR R4 sandbox, 500 synthetic patients
- OTel collector capturing all spans
- Field-level entitlement service
- Attestation report template
You start from this
ALLOWED = {"Condition.code", "Observation.value", "Encounter.period",
"Coverage.payor", "Procedure.code", "DocumentReference.date"}
def fetch_scoped(client, patient_id, task):
"""Return only the fields `task` declares. Everything else is dropped
before it can reach the model, the log or the trace."""
raise NotImplementedErrorWhat the grader asserts
- No field outside ALLOWED appears in the model prompt
- No identifier appears in any captured OTel span, including error spans
- Dropped fields are not merely redacted in output — they are never fetched into memory
- Attestation lists every field touched, with the task that justified it
- Attempting an out-of-scope field raises rather than silently returning None
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 & scopeDelivery mandateAdministrative vs clinical scopeValue sizingPrivacy path | 25 | 35 min | 4 |
| 02 | Healthcare data landscapeFHIRX12 claimsTerminologyScoping and entitlements | 25 | 35 min | 4 |
| 03 | Context engineeringNegation and ambiguityStructured outputsPHI-safe designOmission | 25 | 35 min | 4 |
| 04 | Retrieval & groundingScoped retrievalPayer policyTerminologyCitation and evals | 25 | 40 min | 4 |
| 05 | Agent design & autonomyWorkflow designReview gatesMulti-agent RCMContainment | 25 | 40 min | 4 |
| 06 | EHR integrationSMART on FHIRWrite-back safetyIdentity and contextResilience | 25 | 35 min | 4 |
| 07 | HIPAA, privacy & securityMinimum necessaryBAA and de-identificationLLM securityAttestation | 30 | 45 min | 4 |
| 08 | Production & operationsEval gatingOutcome 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.
Q1Where should the clinical line be enforced?
A boundary that depends on the model choosing to respect it is not a boundary. It must be architectural and testable.
Q2You need to answer a question about one encounter. What retrieval scope is correct?
Scope must be enforced, not merely intended. Encounter-scoped retrieval is both the compliant and the more accurate choice.
Q3Your extraction turns "no chest pain" into a positive finding. What is the failure?
Negation, uncertainty and family history are the three classic clinical extraction traps. All three need explicit evals.
Q4Payer prior-authorisation criteria vary per payer and plan and change without notice. What does this demand?
Stale criteria produce confident, wrong submissions. Freshness is a first-class requirement, not a maintenance task.
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 |
|---|---|---|---|
| Healthcare domain fluency | Knows the settings. | Maps care and revenue chains to decisions. | Sizes admin value and designs the operating change. |
| PHI engineering | Aware of HIPAA. | Implements minimum necessary in code. | Designs an attestable PHI architecture. |
| Interoperability | Reads a FHIR resource. | Builds against FHIR and X12. | Designs EHR-embedded workflows that clinicians keep using. |
| 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
- PHI
- Protected Health Information — the data class HIPAA governs.
- Minimum necessary
- The HIPAA principle that access is limited to the fields a task requires.
- BAA
- Business Associate Agreement — required with any vendor processing PHI.
- FHIR
- Fast Healthcare Interoperability Resources — the modern healthcare API standard.
- Prior authorisation
- Payer approval required before a service is delivered.
- RCM
- Revenue Cycle Management — everything from registration to collection.
- CDI
- Clinical Documentation Improvement — closing gaps between care given and care recorded.
- CMS-0057-F
- The US rule mandating FHIR-based prior authorisation APIs by January 2027.
- 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
- No clinical background required; the program stays on the administrative side of the line
- Willingness to take privacy seriously is non-negotiable here
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 providers, payers and health tech
- 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 — Healthcare.
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
Healthcare
- Specialisation
- Healthcare
- 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-HEALTHCARE-<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.