An advisory panel inside the EHR, at the referral decision.
Crossbeam helps US health systems keep referrals in network, expand specialist capacity, and verify outcomes on claims, from inside the EHR at the one moment it changes the decision. A draft concept in clinician validation; the clinician always decides.
HIPAA-ready by designSOC 2 · planned, not yet certifiedAdvisory only · the clinician decidesNo new login, inside the EHR
The approach
Calm. Specific. Advisory.
Instead of flooding care teams with alerts, dashboards, and another queue, Crossbeam surfaces one focused panel at the one moment it changes the decision: the referral. Your institution's own specialty protocol, read against the chart that is already open. The clinician takes it from there.
Intelligent prioritizationThe panel reads the chart against your own division's protocol: criteria met, criteria missed, red flags first. Signal, not noise.
Built for action, not reviewThree real doors, each one click: manage with the pathway, send the pre-assembled eConsult, or refer. Minimal clicks, minimal cognitive load.
Native to the workflowOpened by choice inside the EHR, on its own FHIR and In Basket rails. No new login, no pop-ups, no hard stops.
Aligned with financial and quality goalsEvery disposition ties back to in-network retention, specialist capacity, and quality performance, verified on claims.
From signal to action to outcome · the actual product, not a slide deck
A working, AI-native product you can drive today.
One panel carries the whole journey: the signal (the chart, read against your institution's own protocol), the action (three real doors, one click each), and the outcome (verified on claims, feeding the next protocol version). These are real screens from the running Crossbeam demo app, not mockups. The AI drafts; the clinician decides and signs everything.
EHR sidebar · advisory panelTHE ACTUAL PRODUCT
The AI matched this chart against the GI division's protocol v3.2, streamed its read, and drafted the three options. The clinician chooses; nothing signs itself.
01 · SignalAutomated consult triage and routing.
The institution's own protocol, the criteria it met and missed, and three real doors: refer, eConsult, or manage in primary care. The AI streams its read of the chart, every item sourced; the gold ring marks its recommendation. Opened by choice inside the chart that is already open, never a pop-up, and the clinician always decides.
eConsult composerTHE ACTUAL PRODUCT
The AI drafted this eConsult from the chart: the question, labs, and prior workup attached. The clinician edits and sends; the specialist reply and disposition are logged for outcome tracking.
02 · ActionAsynchronous specialist oversight.
The eConsult arrives pre-assembled: drafted, not dictated, and the PCP edits and sends. A complete question the specialist can answer on arrival, routed on the EHR's own In Basket rails; programs at scale answer in a median 0.9 days,12 and a protocol-prepared review takes about 20 minutes instead of a 30 to 60 minute visit.19
Outcomes · claims-verifiedTHE ACTUAL PRODUCT
Every disposition is tracked on claims and chart data, labeled by how it was measured. The specialty division tunes its own protocol on real results, and this session's case flows into the aggregate.
03 · OutcomeOutcome-driven protocols.
Every avoided wait and completed pathway measured on claims and chart data, labeled by how it was measured. Self-report is secondary telemetry, never the headline. Real-world outcomes write the next protocol version: the specialty division tunes its own pathway on its own results.
Every capability the market is naming, and where it lives in the product
Collaborative care platformThe whole product: the four pillars, working as one loop.
Automated consult triage and routingP3: the three doors at the referral decision moment.
Protocol-driven workup and ordersP1: the specialty's preferred workup, pended for signature.
Specialist capacity expansionAsynchronous oversight: a 20-minute review, not a 30 to 60 minute visit.19
Outcome-driven protocolsP4: real-world outcomes write the next protocol version.
Autonomy is earned rung by rung: v1 is the permanent advisory floor (the clinician signs everything), v2 auto-drafts for review, v3 is narrow, evidence-gated auto-resolution only.
Real screens from the running Crossbeam demo app (Next.js + streamed Claude calls). Fictional patients, demo data, and a fictional “Meridian” EHR styled after common US EHR layouts. No affiliation with any EHR vendor. Advisory only; no product or clinical claims are made.
Live on this site · guided and fully interactive
Drive the product yourself. A guide rides along.
Open the full-screen demo and a seven-step guide walks you through the whole journey with callouts that react to what you actually do: nothing here is a video. Ask the hallway, watch the AI assemble the chart, choose a door, send the eConsult, read the outcomes console, then re-run it on a red-flag patient and watch the recommendation flip.
Runs full screen at demo.crossbeamhealth.com. Fictional patients, a fictional “Meridian” EHR, no PHI anywhere. Deterministic streaming by default so the demo always works; the guardrailed live-AI path is hardened against prompt injection with enumerated inputs, sanitized free text, an origin allowlist, and rate limits.
The published evidence
Built on mechanics that are already proven.
Crossbeam is a draft concept, so this page claims no results of its own. The mechanics it is built on are published, measured, and cited in the Sources list below.
0.9 daysMedian specialist response time across 100,000+ eConsult cases at a program operating at scale. Asynchronous rails work when the question arrives complete.12
Champlain BASE, Ontario
About 20 minutesSpecialist review time for a protocol-prepared eConsult, versus a 30 to 60 minute clinic visit. That difference is the capacity Crossbeam recovers.19
PLOS ONE, 2024
50 to 98%of interruptive decision-support alerts are overridden. It is why the panel is opened by choice, and never a pop-up or a hard stop.7
Systematic review, 2023
How it works
Four steps, no new queue.
1OPENThe clinician opens the panel by choice from the referral workflow. Never a pop-up, never a hard stop.
2SEETheir own institution's specialist protocol for this presentation, with the patient's data already mapped against it.
3CHOOSEThree doors: send the referral, send the pre-assembled eConsult, or manage with the pathway and a safety-net recheck.
4LEARNOutcomes flow back from claims and chart data, and the specialty division tunes its own protocol on real results.
Your EHR vs. Crossbeam
Your EHR captures the chart. It was never built to rebuild the hallway.
The EHR records what happened. It does not carry your specialists' judgment to the primary care clinician at the one moment it changes the decision. Crossbeam is the layer that does, without a new login and without a new queue.
At the referral decision
The EHR alone
With Crossbeam
The specialist's protocol
Lives in a PDF, an email, or a specialist's head. Not at the decision.
Your division's versioned pathway, in the panel, at the moment of the referral.
Assembling the chart
The PCP or an RN team gathers labs, imaging, and history by hand.
Pulled and shown item by item, each with its source, ready to reuse.
The eConsult
Free text, if an eConsult rail exists at all.
Drafted from the chart and the pathway. The clinician edits and sends in one click.
The outcome
Rarely closed. No learning loop back to the specialty.
Closed on claims and chart data. The division tunes its own protocol on real results.
How it is delivered
Interruptive pop-ups, 50 to 98% overridden.7
An advisory panel opened by choice. Never a hard stop; the plain referral is always one click away.
Why it matters, why now
The referral system is where health systems bleed capacity, revenue, and clinician time.
58M → 206MIn-basket messages a year at one large academic health system, 2017 to 2024. The referral system's dropped work lands in the primary care inbox.4
50 to 98%of interruptive decision-support alerts are overridden. Pop-ups and hard stops are the wrong delivery for specialist judgment.7
15.2 hrs/wkof total active EHR time for high-volume primary care physicians, and portal message load never returned to pre-pandemic levels. Another queue is not an option.6
40,400upper-bound AAMC-projected US shortage of primary care physicians by 2036, with specialties like thoracic surgery and ophthalmology at 69 to 70 percent of projected need by 2035. Capacity has to come from infrastructure, not headcount.17, 18
US market focus. Numbered citations resolve in the Sources list below; numbering matches the concept deck's reference list.
The clinician always decides.
Crossbeam carries specialty judgment across the doorway of the referral decision. It never blocks, never auto-diverts, and never grades clinicians in their first year. The plain referral is always one click away, and everything is built on your EHR's own FHIR and In Basket rails.
