Clarity

Section 01

Venture Overview

Problem

Health system complexity + learned helplessness at first independent enrollment

User

Young adults — first job, first time choosing their own insurance

Entry Point

Open enrollment window — 2–3 weeks, maximum confusion, no support

Product

AI health navigation app + on-demand human consultant

Model

B2B2C — employer pays, employee trusts

Moat

Structural privacy — health data never touches employer systems

"Employer funded. Employee trusted. Structurally independent."
Clarity — Core Architecture

Section 02 — Session 4

Knowledge Graph — 100 Triples

5 domain clusters + 15 cross-cluster reasoning chains. User-centered — every entity exists in relation to one specific person's situation.

👤

User

Triples 1–5 ✓

📋

Insurance

Triples 6–25 ✓

🏥

Care System

Triples 26–45 ✓

💊

Health Events

Triples 46–65 ✓

💰

Financial

Triples 66–85 ✓

🔗

Cross-Cluster

Triples 86–100 ✓

👤 User — Triples 1–5

#	Entity A	Relationship	Entity B
1	User	HAS	Insurance Plan
2	User	EARNS	Income Bracket
3	User	LIVES_IN	Location
4	User	HAS_CONDITION	Health Status
5	User	CAN_AFFORD	Monthly Premium

📋 Insurance Plans — Triples 6–25

#	Entity A	Relationship	Entity B
6	Insurance Plan	HAS_TYPE	HMO
7	Insurance Plan	HAS_TYPE	PPO
8	Insurance Plan	HAS_TYPE	HDHP
9	Insurance Plan	HAS_COST	Monthly Premium
10	Insurance Plan	HAS_COST	Deductible
11	Insurance Plan	HAS_COST	Copay
12	Insurance Plan	HAS_COST	Coinsurance
13	Insurance Plan	HAS_LIMIT	Out-of-Pocket Maximum
14	Insurance Plan	INCLUDES	In-Network Providers
15	Insurance Plan	EXCLUDES	Out-of-Network Providers
16	Insurance Plan	REQUIRES	Primary Care Physician
17	Insurance Plan	COVERS	Preventative Care
18	Insurance Plan	COVERS	Mental Health
19	Insurance Plan	COVERS	Prescriptions
20	Insurance Plan	COVERS	Emergency Care
21	Employer	CONTRIBUTES_TO	Monthly Premium
22	Employer	SETS	Enrollment Window
23	Employer	DEFINES	Waiting Period
24	HDHP	PAIRS_WITH	HSA Account
25	User	SELECTS	Insurance Plan

🏥 Care System — Triples 26–45

#	Entity A	Relationship	Entity B
26	Care System	HAS_PROVIDER	Primary Care Physician
27	Care System	HAS_PROVIDER	Specialist
28	Care System	HAS_PROVIDER	Urgent Care
29	Care System	HAS_PROVIDER	Emergency Room
30	Care System	HAS_PROVIDER	Telehealth
31	Primary Care Physician	SERVES_AS	First Point of Contact
32	Specialist	REQUIRES	Referral (HMO)
33	Urgent Care	COSTS_LESS_THAN	Emergency Room
34	Telehealth	IS_COVERED_BY	Most Modern Plans
35	Visit Type	HAS_CATEGORY	Preventative
36	Visit Type	HAS_CATEGORY	Acute
37	Preventative Visit	IS_FREE_UNDER	ACA Guidelines
38	Acute Visit	TRIGGERS	Copay
39	Emergency Room Visit	TRIGGERS	Deductible
40	Provider	HAS_STATUS	In-Network
41	Provider	HAS_STATUS	Out-of-Network
42	In-Network Provider	COSTS_LESS_THAN	Out-of-Network Provider
43	HMO Plan	REQUIRES	Referral for Specialist
44	PPO Plan	ALLOWS	Direct Specialist Access
45	User	SHOULD_VERIFY	Network Status Before Visit

💊 Health Events — Triples 46–65

#	Entity A	Relationship	Entity B
46	Health Event	HAS_TYPE	Preventative
47	Health Event	HAS_TYPE	Acute
48	Health Event	HAS_TYPE	Chronic
49	Preventative Event	PREVENTS	Acute Crisis
50	Annual Physical	IS_COVERED_AT	Zero Cost (ACA)
51	Blood Panel	DETECTS	Early Risk Indicators
52	Early Risk Indicator	TRIGGERS	Preventative Action
53	Preventative Action	COSTS_LESS_THAN	Crisis Treatment
54	Acute Event	TRIGGERS	Unplanned Spending
55	Mental Health Episode	IS_COVERED_BY	Modern Insurance Plans
56	Chronic Condition	REQUIRES	Ongoing Care Plan
57	Chronic Condition	INCREASES	Annual Healthcare Cost
58	Life Stage Trigger	REQUIRES	Insurance Re-evaluation
59	New Job	IS_A	Life Stage Trigger
60	Aging Out of Parent Plan	IS_A	Life Stage Trigger
61	Moving to New City	AFFECTS	Provider Network Access
62	EOB Received	REQUIRES	Plain Language Translation
63	Lab Result	REQUIRES	Contextual Explanation
64	Enrollment Window	HAS_DURATION	2–3 Weeks
65	User	EXPERIENCES	Life Stage Trigger

💰 Financial — Triples 66–85

#	Entity A	Relationship	Entity B
66	Financial Document	HAS_TYPE	EOB
67	Financial Document	HAS_TYPE	Medical Bill
68	EOB	IS_NOT	Medical Bill
69	EOB	SHOWS	What Insurance Paid
70	Medical Bill	SHOWS	Patient Responsibility
71	Itemized Bill	CONTAINS	CPT Codes
72	CPT Code	REQUIRES	Plain Language Translation
73	Surprise Bill	VIOLATES	No Surprises Act (2022)
74	Patient Responsibility	EQUALS	Bill Minus Insurance Payment
75	User	HAS_RIGHT	Request Itemized Bill
76	Hospital	OFFERS	Financial Assistance Program
77	Financial Assistance	IS_UNKNOWN_TO	Most Patients
78	HSA	REQUIRES	HDHP Enrollment
79	HSA	PROVIDES	Triple Tax Advantage
80	FSA	HAS_RULE	Use It or Lose It
81	Medical Debt	CAN_AFFECT	Credit Score
82	Payment Plan	IS_AVAILABLE_AT	Most Hospitals
83	Payment Plan	IS_UNKNOWN_TO	Most Patients
84	User	SHOULD_NEGOTIATE	Medical Bill
85	Clarity	TRANSLATES	Financial Document → Plain Language

🔗 Cross-Cluster — Triples 86–100

#	Entity A	Relationship	Entity B
86	User (healthy, low income)	SHOULD_SELECT	HDHP + HSA
87	User (chronic condition)	SHOULD_SELECT	PPO
88	User (new city)	MUST_REVERIFY	In-Network Providers
89	Life Stage Trigger	INITIATES	Full Plan Re-evaluation
90	Preventative Visit	REDUCES	Long-Term Financial Risk
91	Unmet Deductible	AFFECTS	Cost of Acute Visit
92	HSA Balance	OFFSETS	Out-of-Pocket Cost
93	Out-of-Network Visit	GENERATES	Surprise Bill Risk
94	Surprise Bill	TRIGGERS	No Surprises Act Protection
95	Mental Health Visit	IS_COVERED_EQUALLY	Physical Health Visit
96	Annual Physical	PREVENTS	Undetected Chronic Condition
97	Undetected Chronic Condition	INCREASES	Medical Debt Risk
98	Employer	FUNDS	Clarity Access
99	Clarity	BUILDS	User Health Literacy Over Time
100	Health Literate User	COSTS_LESS_THAN	Uninformed User (to employer)

Health Literate User → COSTS_LESS_THAN → Uninformed User
Triple 100 — The Business Case in One Line

Section 03 — Session 5 · Ontology Linking

Entity Linking — Grounded in Real-World Standards

Each key entity in Clarity's KG maps to an authoritative real-world standard — proving the graph is verifiable, interoperable, and trustworthy.

Clarity Entity	Standard / Ontology	Authority	Notes
Insurance Plan (HMO/PPO/HDHP)	HL7 FHIR InsurancePlan Resource	HL7 International	FHIR R4 defines plan types used in all US payer APIs
Deductible / Copay / OOP Max	ACA §1302 Cost-Sharing Rules	HHS / CMS	45 CFR §156.130 — federally mandated disclosure fields
Primary Care Physician	FHIR Practitioner Resource + NPI Registry	HL7 / CMS	NPI = unique provider identifier; searchable via CMS API
Specialist	NUCC Health Care Provider Taxonomy	NUCC / CMS	237 specialty codes; maps to NPI registry
Emergency Room	SNOMED CT: 225728007	SNOMED International	"Accident and emergency department" clinical concept
Telehealth	CPT Modifier 95 / GT	AMA CPT	Telehealth billing modifiers; CMS coverage rules apply
Annual Physical / Preventative Visit	CPT 99381–99397	AMA CPT Code Set	ACA mandates zero cost-sharing for these specific codes
Blood Panel / Lab Result	LOINC Codes	Regenstrief Institute	Universal standard for lab tests and clinical observations
Chronic Condition	ICD-10-CM Diagnosis Codes	WHO / CDC	70,000+ diagnosis codes; mandatory for all US billing
Mental Health Visit	CPT 90832–90838 + Mental Health Parity Act	AMA / DOL	Parity Act (2008) mandates equal coverage with physical health
CPT Code	AMA CPT Code Set	American Medical Association	~10,000 procedure codes; annual updates
EOB	FHIR ExplanationOfBenefit Resource	HL7 / CMS	CMS Blue Button 2.0 uses FHIR EOB for all Medicare data
HSA	IRS Publication 969	Internal Revenue Service	Federal tax code defines eligibility, contribution limits, rules
No Surprises Act	42 CFR Parts 749, 800, 826	US Federal Law (CMS)	Effective Jan 1 2022; prohibits balance billing out-of-network
Enrollment Window	45 CFR Part 155	HHS / CMS	Federal regulation defines Special and Open Enrollment Periods

Why Ontology Linking Matters

FHIR compliance → Clarity can ingest real insurance + EOB data via API
CPT / ICD-10 mapping → Clarity can read and translate actual bills
IRS Publication 969 → HSA guidance is legally grounded
SNOMED CT → Health event taxonomy is clinically validated
The KG is not invented terminology — it's grounded reality

Section 04 — Session 5 · AI Reliability Analysis

AI Reliability Analysis

Where will Clarity's AI fail — and what is the architecture to catch it? Eight failure modes mapped with severity and mitigation design.

#	Failure Mode	Severity	Mitigation
1	Plan recommendation error — AI suggests HDHP to user with undiagnosed chronic condition	Critical	Complete health intake required before any plan recommendation; incomplete profiles flagged
2	Network status hallucination — AI states provider is in-network using stale data; user gets surprise bill	Critical	Live FHIR network API only; network status never cached >24hrs
3	EOB misread — AI tells user they owe money when EOB is informational only	Critical	Triple 68 (EOB IS_NOT Medical Bill) hardcoded as non-negotiable rule; never generated
4	State regulation blind spot — AI applies federal rules in state with different requirements	High	User.location (Triple 3) gates all regulatory guidance; state rule layer required at launch
5	CPT code hallucination — AI invents or misdescribes a procedure code	High	CPT descriptions from AMA database only; no generative description of medical codes
6	Mental health stigma — AI frames mental health coverage in ways that discourage use	High	Parity framing locked in prompt layer; human review on all mental health responses
7	Deductible math error — AI calculates remaining deductible incorrectly from stale claims data	High	Deductible tracker requires live EOB feed; uncertainty shown when data >7 days old
8	Financial assistance invisibility — AI fails to surface charity care programs the user qualifies for	Medium	Triple 77 hardcoded as proactive trigger; always surface assistance before payment options

Three-Layer AI Safety Architecture

Layer 1 — Grounded KG: All factual assertions from ontology-linked triples, not LLM memory
Layer 2 — Deterministic Logic: Financial math, legal rights, network status = code not generation
Layer 3 — Human Escalation: Chronic conditions, disputed bills, denied claims → human consultant

The rule: AI explains. Humans decide. Clarity never tells you what to do.

Section 05

What We Don't Know — Risk Analysis

#	Open Question	Risk	Mitigation Path
1	Will employees trust an employer-funded health app with their personal data?	High	Structural privacy firewall; independent data store; explicit user consent flow
2	Does plan recommendation trigger insurance advisor licensing requirements?	High	Legal review pre-launch; "guidance not advice" framing; human consultant as licensed layer
3	Can Clarity access real-time FHIR data from employer insurance carriers?	Med	API partnerships with carriers; CMS interoperability rules help but vary by carrier
4	How do we handle 50-state insurance regulation variation at scale?	Med	Location entity gates all regulatory guidance; state-specific layer is a significant build
5	What happens when Clarity's recommendation conflicts with employer's preferred plan?	Med	Fiduciary-like duty to user written into employer contract language explicitly