~50M
U.S. women affected
Perimenopausal + postmenopausal population in the U.S.
Full investor proposal · Pause-Health.ai
Menopause Clinical Decision Support
Pause-Health.ai is designed to help clinicians diagnose and treat menopause-related symptoms faster and more accurately by combining patient history, wearable signals, and AI guidance inside normal clinical workflows — EHR-native, never a sidecar. The prototype is live and open-source today: a 2,015-provider directory with distance ranking + state license-sanction filters, Salesforce Data Cloud Calculated Insights grounding live in production, and a MuleSoft Experience API contract verifiable from any curl against /api/mulesoft/providers. Provider organizations onboard in 2026 H2.
~67%
Initially misdiagnosed
Of perimenopausal women, per peer-reviewed literature.
~2.5 y
Time to correct diagnosis
From first menopause-pattern symptom to accurate dx.
~$1,685
Avoidable spend / patient
Estimate of waste per patient from delayed dx + over-utilization, derived from menopause-care cost-of-care literature. Illustrative -- this is a published-research-derived figure, not a Pause-Health.ai measurement.
Executive summary
Close the menopause-care gap by meeting clinicians where they already work
Many women in perimenopause and menopause wait too long for correct diagnosis and treatment. Symptoms are often misattributed, care is inconsistent, and providers do not always have enough menopause-specific training or decision support.
Pause-Health.ai is designed to close that gap. The product supports providers at the point of care with clear risk scoring, treatment suggestions, and workflow-ready guidance. The goal is better outcomes for patients and measurable operational value for health systems and payers.
Target
89%
Diagnostic accuracy
Validation target vs. the ~67% national misdiagnosis baseline. Measurement plan documented in /proposal/insights.
Target
< 30 d
Time to first specialist contact
From a 2.5-year baseline to under a month — first MSCP-credentialed clinician contact.
Target
+34%
Patient satisfaction lift
Pilot target with design-partner provider organizations vs. care-as-usual baseline. (Design-partner program kicks off 2026 H2.)
Target
Time-to-MSCP
Closed-loop metric per design partner
Replaces the previous '$1,685 avoidable spend recovered' metric. The cost-recovery number is literature-derived (see hero); the per-design-partner outcome we will actually measure is time-to-MSCP-visit and Care-Router pathway concordance vs. clinician adjudication.
What Pause-Health.ai provides
Six capabilities, each verifiable in the live prototype
AI-assisted triage + risk scoring
Deterministic risk-band scoring (lib/risk-band.ts) plus an Anthropic-backed Care Router policy surface intake-time triage. Wired and inspectable in the public prototype today.
See live intake →Persona-aware care pathway routing
Six canonical pathways (self-care → ED referral) from a single source-of-truth enum (lib/care-router-pathways.ts). Policy-aware, traceable, designed to be overridable by clinicians once design-partner deployments are in force.
See live routing →Federated Data 360 grounding
Phase 1 (Salesforce Health Cloud SOQL) and Phase 2 (Data Cloud Calculated Insights — HRV z-score, vasomotor burden, sleep disruption) are both live in production on the trailsignup org. The grounding endpoint reports 'Phase 2: SOQL (Health Cloud) + Data Cloud Calculated Insights' on every call; each insight falls back to its intake baseline independently if a DC call fails. Phase 2-bis swaps the demo-cohort seeded CIs for real JHE/DBDP wearable math; Phase 3 onboards the customer's EHR-of-record. Interface stays constant across phases.
Architecture brief →Defensible provider graph
2,015 NPPES-derived providers behind a frozen Experience API contract. Census 2020 ZCTA distance ranking, six NPPES board-cert + multi-specialty signals, three state license-sanction filters dropping 1,720 sanctioned candidates at build (CA Medi-Cal + NY OPMC + TX TMB), real-shaped synthetic insurance acceptance. Browseable UI at /provider, per-NPI profiles at /provider/<npi>, MCP tool find_menopause_providers. The Care Router consumes the same query function so triage and the directory stay in lockstep. Closed-loop outcomes scoring (Phase 3) activates with referral volume.
Provider-graph brief →Outcomes telemetry baked in
Every Care Router decision emits OpenTelemetry-style spans (intake → identity → grounding → routing). The 'clinician's eventual action' attribute is designed (a real clinician needs to be on the other end first); spans are populated today by the prototype. Optional Salesforce Platform Event egress (sink wired 2026-06-24, dormant until customer-org Connected App + Pause_Agent_Trace__e are provisioned) ships every span into the customer's Shield + Event-Monitoring audit pipeline under the Connected App's integration user.
See the trace plane →MCP server + host (both directions)
Pause-as-tool-source: the @pause-health/mcp server (stdio + Streamable HTTP at /api/mcp) exposes get_patient_timeline, get_patient_intake, find_menopause_providers, and experience_api_health to Claude Desktop, Cursor, and the Salesforce Agentforce 3.0 Registry. Pause-as-tool-consumer: the Care Router task endpoint now runs an MCP client per request — registers the loopback /api/mcp plus any external slots from PAUSE_MCP_HOST_REMOTES — and resolves provider recommendations via tool calls instead of direct HTTP. Verified end-to-end: host-on and host-off paths return byte-identical provider lists.
MCP brief →Technology foundation
Built on open + best-of-breed substrates
Pause-Health.ai composes six substrates the customer's data team already trusts. Architecture briefs link out to each.
JupyterHealth Exchange
Consented FHIR R5 data exchange substrate. The interop plane for wearable, EHR, and intake records. Real JHE Django + Postgres stack stood up on the maintainer's box (2026-06-16); pause_ingest round-trips both a raw Oura heart-rate sample (mapped handler) and a derived HRV-time-domain feature (auxiliary handler, with derivedFrom provenance) end-to-end. Followed up 2026-06-23 with an opt-in PAUSE_USE_REAL_JHE=1 pytest marker — same seven contract assertions now run against the live JHE instance, not just the wire-level mock. Federation across customer-org JHE deployments is the Phase-2 target.
JupyterHealth integration brief →DBDP feature engineering
Duke's Digital Biomarker Discovery Pipeline — sleep, HRV, activity, skin temperature features. FLIRT-backed RMSSD math is integrated in pause_ingest today with 20 passing unit tests including a closed-form correctness check; FHIR persistence is Phase 2.
DBDP brief →Salesforce Data 360
Identity Resolution + grounding wired against Salesforce Health Cloud via OAuth Client Credentials (Phase 1, live). Three Data Cloud Calculated Insights (Pause_HRV_RMSSD_30d, Pause_Vasomotor_Burden_30d, Pause_Sleep_Disruption_7d) authored over ssot__Individual__dlm and activated on the trailsignup tenant (Phase 2, live; auth flows through the mandatory a360 token exchange — see PHASE_2_ACTIVATION_CHECKLIST.md for the five gotchas). Segments + federated query against JHE/DBDP wearable math is Phase 2-bis. The Care Router interface doesn't change across phases.
Data 360 brief →MuleSoft + Agent Fabric
MuleSoft CloudHub 2.0 worker live (iterations 1–8 shipped: Flex Gateway runtime enforcement, Auth0 RS256 JWT validation, plain Rate Limiting, OAS 3.0 spec in Exchange, and the Phase-2 contract DataWeave deployed as v1.0.4 on 2026-06-16 with the mock's full shape — distance, signals, sanctions, insurance, dataset provenance — live behind Auth0-JWT). Production /api/mulesoft/providers reports meta._source: 'live-mulesoft' end-to-end. Iteration 9 (persistent VM hosting for the Flex Gateway, moving off the local-ngrok rig) is the remaining piece. The Agent Fabric multi-agent control plane (agent registry, policy catalog, trace plane) is the designed production home for the Anthropic-backed Care Router; the underlying MuleSoft surface is more than 'designed' — it's running.
Agent Fabric brief →Salesforce Headless 360 conformance
Salesforce's TDX 2026 architecture umbrella — 'every Salesforce surface is an API, MCP tool, or CLI command, and agents can use all of it.' Pause covers most of it incidentally: REST (mulesoft + data 360 + agent.json), MCP (the /api/mcp server + the Care Router host), A2A (the care-router/tasks endpoint with an Agent Card). The /proposal/headless-360 audit page names the four explicit gaps; two are now wired dormant — the PKCE External Client App OAuth flow (gap #1) with six routes under /api/salesforce/headless-360/* and 25 unit tests pinning S256 + signed-cookie invariants, and the Salesforce Platform Event egress sink (gap #3) emitting every Agent Fabric span as a Pause_Agent_Trace__e record once the customer-org Connected App is provisioned. Activation runbooks ship alongside the code so the operator-side hand-off is one document.
Headless 360 audit →Agentforce Voice
Salesforce announced Agentforce Voice GA on 2025-10-13 and shipped Agentforce Contact Center on 2026-03-10. The partner-web developer surface is sales-gated as of 2026-06-24 (no public LWC, no Agent API voice endpoint), and the audio round-trip requires Contact Center licensing + a CCaaS partner contract (Amazon Connect / Five9 / NiCE / Vonage). Shipped honestly: a 4-env-var seam (PROVIDER + BASE_URL + DEPLOYMENT_REF + AGENT_DEPLOYMENT) plus a public-safe /api/agentforce/voice/config probe and a launch button that renders one of three affordances driven by provisioning state. The button click currently surfaces a 'verification pending' toast — the CCaaS handshake lands on activation day. The page deliberately does NOT claim a Web Speech API browser wrapper is 'Agentforce Voice'; that's a different product.
Agentforce Voice brief →Business model
Target ACV ranges · pre-design-partnerB2B healthcare SaaS — three channels
The ACV / PMPM ranges below are target ranges derived from comparable provider-vendor pricing benchmarks, not booked revenue. Pause-Health.ai is pre-design-partner; final pricing lands with the first executed MSAs.
| Channel | Motion | ACV / PMPM | Detail |
|---|---|---|---|
| Health systems | IDN + AMC enterprise sales | $25K – $75K | Annual contract; bundled with EHR-native install. |
| Payers | Value-based payer partnerships | $0.50 – $2.00 PMPM | Eligible perimenopausal + postmenopausal populations. |
| Medical practices | MSCP-credentialed independent practices | $2K – $8K | Annual subscription; lighter Agentforce-only install. |
Target
$8M
24-month ARR target
Across IDN + payer channels.
Target
$50M+
Year-4 ARR trajectory
Through multi-site IDN expansion + payer scale.
Market + strategic timing
Literature-derived sizingA large, under-served market in a tractable moment
Market size
U.S. menopause-care market estimated at ~$15.4B, growing ~5.7% annually (literature-derived sizing across pharma, devices, services, and digital health; the precise figure varies meaningfully across analyst sources).
Payer + IDN pressure
Both sides under measurable pressure to improve outcomes and reduce avoidable spend.
AI readiness
LLMs are now practical enough for real-time, point-of-care decision support — at acceptable latency and cost.
Interop floor rising
FHIR R5 + SMART-on-FHIR + wearable feature tooling shorten time to clinical validation.
Core problems to solve
Where current care breaks down
1. Misdiagnosis and delays
- Symptoms misattributed to depression, anxiety, thyroid, or 'normal aging'.
- Correct diagnosis frequently takes years.
- Delays erode trust and prolong patient suffering.
2. Cost burden
- Excess referrals + avoidable utilization inflate cost.
- Payers + providers each absorb the inefficiency.
- Workforce productivity losses extend the impact beyond healthcare.
3. Inconsistent treatment
- No universal protocol across systems.
- Recommendations vary widely between providers.
- Outcome feedback loops are frequently missing entirely.
4. Provider workflow burden
- Menopause care crosses specialties — hard to manage in a 12-minute visit.
- Research evolves faster than busy clinicians can track.
- Existing tools are usually not integrated into daily EHR workflows.
24-month objectives
Primary objective
Deploy Pause AI decision support across 25 health systems, improve diagnostic accuracy by 30% over baseline, cut time-to-diagnosis by 50%+, and reach $8M ARR with a clear path to $50M+.
Strategic objectives
Objective 1
Clinical validation
Validate performance against real cohorts and publish peer-reviewed evidence.
Objective 2
Health system adoption
Win anchor IDN customers and prove operational ROI in pilot.
Objective 3
Payer scale
Sign regional payer partnerships and demonstrate PMPM savings.
Objective 4
Revenue validation
Build durable ARR with expansion (NRR) and pipeline health metrics.
Objective 5
Product excellence
Robust EHR integration, performance, uptime, and the planned regulatory progression (HIPAA controls in force at design-partner kickoff, HITRUST CSF then SOC 2 Type II as Year-2 milestones).
Success criteria · month 24
All targets · not measured yetWhat “done” looks like
- Diagnostic accuracy89%+in validation programs by month 24
- Time-to-diagnosis50%+ reductionfrom the 2.5-year baseline
- ARR$8M+with credible path to $50M+
- Health system customers20+live IDN deployments
- Payer partnerships2+regional value-based deals
- Trust postureHIPAA + roadmap HITRUSTwith strong provider NPS
Strategic priorities
How we execute
Lead with clinical evidence
Publication-grade validation precedes sales pressure. Earned credibility scales.
Build around provider workflow realities
EHR-native, SMART-on-FHIR. Never a sidecar app the clinician has to remember to open.
Align value metrics with payer economics
PMPM, diagnostic-yield, and avoidable-spend metrics — the ones the payer's actuarial team already tracks.
Continuously improve from real outcome data
Closed-loop scoring: every routed pathway, every clinician override, every patient outcome flows back into training + evaluation.
Trust + compliance as long-term differentiation
The trust posture compounds: HIPAA controls designed-for-production today, HITRUST CSF + SOC 2 Type II as planned Year-2 milestones, and ToS-respecting integrations (e.g. The Menopause Society directory deep-link rather than scraping) wired in the prototype. See /security for the per-control proto-vs-prod view.
Read deeper · architecture + implementation
The architecture story (eleven briefs)
Each architecture brief covers one substrate or agent in depth, with a proto-vs-prod table and links to the corresponding live demo or mocked API surface.
Agent Fabric
The multi-agent control plane — registry, policy catalog, trace plane.
Salesforce Data 360
Zero-copy federation + Identity Resolution + Calculated Insights as routing inputs.
MuleSoft integration
API-Led Connectivity tiers (System / Process / Experience), and where each lives in prod.
MCP server + host
Pause's MCP tools (copy-pasteable Claude Desktop + Cursor configs) on stdio AND Streamable HTTP at /api/mcp for the Agentforce 3.0 Registry. The Care Router now also acts as an MCP HOST — calls external MCP servers as tools via PAUSE_MCP_HOST_REMOTES.
Agentforce intake
Why Salesforce Agentforce for intake, and what's live today vs. what's customer-deployment shape.
Agentforce Voice
Salesforce Agentforce Voice (GA 2025-10-13) — partner-web seam is wired and env-driven; audio round-trip gates on Agentforce Contact Center licensing + a CCaaS partner. Honest 'designed' pill until activation.
Headless 360 conformance audit
Maps every Pause surface (Agentforce chat, Data 360, MCP server+host, A2A Care Router, Agentforce Voice, Platform Event sink) onto Salesforce's TDX 2026 three-pattern architecture. Names the four explicit conformance gaps with status pills; gaps #1 (PKCE) and #3 (Platform Event sink) are wired dormant.
JupyterHealth integration
Which JupyterHealth pieces we adopt, in what order, and what we contribute back.
DBDP feature engineering
FLIRT-backed RMSSD shipped in pause_ingest with 20 passing unit tests + closed-form correctness check; FHIR persistence is Phase 2.
Provider graph
Phase 1+2 shipped: 2,015 NPPES-derived providers, distance ranking from Census ZCTA centroids, six NPPES board-cert + multi-specialty signals, three state license-sanction filters dropping 1,720 candidates at build, synthetic-but-real-shaped insurance, /provider browseable UI.
Menopause Society partnership
4 paths to MSCP partnership, with explicit guardrails on what we never claim.
Read deeper · strategy + market
The investment-thesis briefs (six)
Customer selection
ICPs by segment: IDNs, value-based payers, AMCs. Buying-committee personas + sequencing.
Brief →Research-design plan
Provider + patient discovery plan: literature-derived hypotheses, methodology, and the interview program ahead. Pause is pre-design-partner; this is the research we'll run, not research we've run.
Brief →Data inventory + strategy
Sources, integration order, moats, compliance posture.
Brief →Competition
Six competitive categories + a capability matrix + what we don't claim.
Brief →Digital strategy
Architecture pillars, GTM motion by year, the five moats, operating principles.
Brief →Technology choices
Stack, AI approach, evaluation framework, safety stance.
Brief →