01
Health system informatics teams
CMIO and CIO teams adding chart summaries and inbox triage to their EHR.
Healthcare / Clinical Copilots
Clinical copilots built into the EHR workflow
Clinicians lose time reading long charts and working full inboxes, and an AI tool that sits outside the EHR adds one more window. Agnotic Technologies builds clinical copilots that open inside the EHR, summarise the record, triage messages and draft work for the clinician to review and sign.
In the EHR
Launched through SMART on FHIR or CDS Hooks
Grounded
Every suggestion points back to the patient record
Signed by a clinician
Drafts only; nothing is filed without sign-off
Trusted by global innovators
Industry Overview
What is a clinical copilot?
A clinical copilot is AI software that works inside the clinician's EHR to summarise the chart, triage the inbox and draft notes, messages and orders that the clinician reviews and signs. Agnotic Technologies builds clinical copilots that launch through SMART on FHIR or CDS Hooks, ground each suggestion in the patient record and log what was shown and accepted.
Industry Challenges
Where do clinical copilot projects go wrong?
Most failures are about workflow and trust, not the language model.
Outside the chart: a separate tab or app means copying text in and out, so clinicians stop using it within weeks.
Ungrounded answers: a summary that cannot show where a statement came from in the record is a summary nobody can sign.
Silent writes: a tool that files text or orders without an explicit clinician action breaks the chain of responsibility.
No record of use: without a log of what was suggested, edited and accepted, quality and safety reviews have nothing to work from.
Decision table
Copilot, scribe or decision support: which one do you need?
Three kinds of clinical AI that buyers often mix up. They start from different inputs and carry different risks, and many health systems end up running all three.
| Question | Clinical copilot | Ambient scribe | Rule-based decision support |
|---|---|---|---|
| What does it start from? | The patient record and the clinician's inbox | The audio of the visit | Coded data and configured rules |
| What does it produce? | Summaries, triage, draft notes, messages and orders | A draft visit note | Alerts and reminders |
| Where does it run? | Inside the EHR, launched in patient context | On a device in the room, note sent to the EHR | In the EHR at defined workflow points |
| Who acts on the output? | The clinician, who edits and signs | The clinician, who reviews the note | The clinician, who accepts or overrides |
| Typical standards | SMART on FHIR, CDS Hooks, FHIR R4 | FHIR or the EHR's note APIs | CDS Hooks or native EHR rules |
| Main design risk | Statements not grounded in the record | Transcription and attribution errors | Alert fatigue |
| Where we cover it | This page | AI medical scribe page | Clinical decision support page |
Our Approach
How does a copilot stay grounded in the patient record?
A copilot is only as good as what it is allowed to read and how clearly it shows its sources. These are the design choices we make first.
Read only what the launch grants
The copilot reads FHIR R4 resources within the scopes granted at SMART on FHIR launch, for the patient in context, and nothing else.
Cite the record
Every statement in a summary links to the note, result or medication it came from, so the clinician can check it in one click.
Drafts, not actions
Orders, notes and replies arrive as drafts or CDS Hooks suggestions. The clinician accepts, edits or rejects each one in the EHR.
Log every suggestion
What was shown, to whom, from which sources, and what the clinician did with it is recorded for review.
Capabilities
What can a clinical copilot do for clinicians?
Copilot functions we build, alone or together, for the workflows your clinicians name first.
Chart summaries
A short pre-visit summary of problems, recent results, medications and open items, each linked to its source.
Visit and handoff summaries
Summaries for shift handoff, transfer or discharge planning, drafted from the record for the clinician to finish.
Inbox and message triage
Patient messages and results sorted by urgency and type, with a suggested reply or routing for each.
Draft orders
Order suggestions offered at order-select or order-sign through CDS Hooks, for the clinician to accept or discard.
Draft notes and letters
Referral letters, patient instructions and note sections drafted from the chart, never filed without a signature.
Questions about the chart
Answers to a clinician's question about this patient, quoted from the record with links, or a clear statement that the record does not say.
How It Works
How does a copilot request flow from the EHR and back?
The request path we build, from the clinician's click to a signed result. Every step is logged.
Launch in context
The EHR launches the copilot through SMART on FHIR, or calls it through a CDS Hooks hook such as patient-view or order-select.
01
1
Fetch within scope
The service reads only the FHIR R4 resources the launch scopes allow, for the patient and user in context.
02
2
Build a grounded prompt
Relevant record extracts are selected and tagged with their source so the output can cite them.
03
3
Generate in a private environment
The model runs in your cloud account or a covered model service under a BAA, with no training on your data.
04
4
Check before showing
Outputs without a source, or outside the agreed task, are blocked or flagged before the clinician sees them.
05
5
Show as a suggestion
Results appear as a CDS Hooks card or in the SMART app panel, clearly marked as AI drafted.
06
6
Clinician decides and signs
The clinician accepts, edits or rejects. Only accepted content is written back, through the EHR's own order or note workflow.
07
7
Engineering Stack
Standards and technologies behind a copilot
The launch standards, data access and model hosting behind the clinical copilots we build.
EHR launch
SMART on FHIRCDS HooksOAuth 2.0 and OpenID Connect
Data
FHIR R4US CoreHL7 v2 where results arrive as messages
EHR platforms
EpicOracle Health (Cerner)athenahealth
Models and retrieval
Private LLM deploymentsRetrieval over the patient recordEvaluation sets per task
Backend and infra
PythonNode.jsAWSAzure
Measured Impact
Measures for a copilot program
Measures we build reporting around, agreed with your clinical leads before launch.
Acceptance
Suggestions accepted as drafted
Tracked by task, so weak functions are fixed or retired.
Edits
How much clinicians change a draft
A signal of whether drafts save time or create work.
Inbox time
Time spent working messages
Measured before and after the copilot, per clinician group.
Flags
Outputs reported as wrong or ungrounded
Each one reviewed and fed into the test set.
My Lauren Maternal Health Platform | Built by Agnotic
A HIPAA-ready maternal health platform enabling remote patient monitoring, care coordination, and personalized engagement from pre-pregnancy through postpartum.
Case Study
LERA Health: Compliant Women’s Health App Platform
A women’s health product experience focused on guided care journeys, education, and compliance-aware workflows to support long-term engagement.
Case Study
Use Cases
Who builds clinical copilots with us?
Teams that want AI inside the clinician's workflow, not next to it.
02
Digital health startups
Clinical copilots for startups selling into health systems, built as SMART on FHIR apps from the first release.
03
Specialty groups
Copilots tuned to one specialty's chart, such as oncology, cardiology or behavioral health.
04
Virtual care providers
Message triage and visit summaries for clinicians who work mostly in the inbox.
05
Ambulatory networks
Pre-visit summaries and referral letters across many clinics on one EHR.
06
EHR app vendors
Teams adding copilot functions to an existing SMART on FHIR app.
Integrations
Systems a copilot connects to
The systems a copilot reads from and writes through.
Launch
EHR (SMART on FHIR launch)
Workflow
EHR (CDS Hooks service)
Data
FHIR R4 API
Access
Identity provider
AI
Private model hosting
Review
Audit and evaluation store
Self-assessment
Which guardrails should every clinical copilot have?
The checklist we work through with your clinical and informatics leads before a copilot reaches a single clinician.
Compliance-First Development Services Backed by Global Standards
We build secure, scalable products designed for privacy, interoperability, and regulatory readiness from day one across every sector we serve.
Health Insurance Portability and Accountability Act
Protect PHI with privacy-first architecture, encrypted storage and transmission, and strict access controls.
Fast Healthcare Interoperability Resources
Enabling standardized health data exchange through standards-based API mappings.
Our Edge
Why Global Leaders Choose Us
We combine deep technical expertise with industry-specific knowledge to deliver solutions that aren't just functional, but transformational.
Enterprise-Grade Security
We implement rigorous security protocols and compliance standards (HIPAA, GDPR, SOC2) across all industrial solutions to protect sensitive data.
High-Performance Scaling
Our architectures are built to handle massive data loads and user bases, keeping performance steady whether you're serving ten or ten million.
Accelerated Time-to-Market
Using our suite of internal tools and proven frameworks, we reduce development cycles and get your product to market 40% faster.
Embedded AI Integration
Beyond simple wrappers, we build deep-learning integrations and predictive analytics directly into the core of your industry-specific workflows.
Engagement Model
How do we build a clinical copilot?
How to build clinical copilots one task at a time, with clinicians involved from the first week. Timing depends on scope, and we agree a fixed first release on the scoping call.
01
Scoping call
We agree the first task, the EHR and launch method, and who on your side reviews outputs.
02
Task and test set
Clinicians define what a good output looks like and we build a test set from de-identified or synthetic cases.
03
Fixed-scope first release
One copilot task in an EHR sandbox, launched in context, with grounding, logging and the guardrails checklist in place.
04
Pilot with clinicians
A small clinician group uses it in production, and acceptance, edits and flags are reviewed each week.
05
Widen and support
More tasks and clinician groups, with regression tests on every change, under agreed support terms.
Facts and sources
Which standards and rules shape clinical copilots?
Short facts behind copilots that run inside the EHR, each linked to its primary source.
The SMART App Launch implementation guide uses OAuth 2.0 and OpenID Connect so that apps can launch from an EHR with the user's and patient's context and only the access scopes they were granted.
Source: HL7 International (SMART App Launch)CDS Hooks is an HL7 specification that lets an EHR call an external decision support service at points in the clinician's workflow, such as opening a chart or signing an order.
Source: HL7 International (CDS Hooks)FDA's Clinical Decision Support Software guidance explains which decision support functions meet the non-device criteria in section 520(o)(1)(E) of the FD&C Act and which remain device software functions.
Source: FDA guidance, Clinical Decision Support SoftwareThe NIST AI Risk Management Framework is a voluntary framework that organizes AI risk work into four functions: Govern, Map, Measure and Manage.
Source: NIST AI 100-1
Voice of success
We don't just build products; we forge lasting partnerships. See how we've helped industry leaders transform their vision into technical reality.
“I can clearly see how Agnotic has a unique way of handling end-to-end development. They are always active on quick chat and provide support quickly.”
Aaron Phelan
Founder, Benchmark
“Agnotic combines deep technical expertise with strong domain knowledge. They understand the business context, anticipate challenges, and make collaboration smooth and effective.”
John Pasmore
Founder, Latimer
“Agnotic is the best technical team we evaluated. Their engineering excellence made our work dramatically easier and allowed us to stay focused on what matters most for maternal care outcomes. They took full ownership of the technical execution, and we are always happy to continue working together.”
Kim Smith
Founder, My Lauren
Related services
Related clinical AI services
Pages for the clinical AI that a copilot complements rather than replaces.
Generative AI healthcare development
Generative AI products beyond the EHR workflow, from patient tools to internal assistants.
Clinical decision support AI
Rule-based and model-based alerts at defined points in the workflow.
AI medical scribe and ambient documentation
Visit notes drafted from the conversation, the other half of documentation work.
Chat with medical records
Retrieval over clinical documents, the pattern behind chart questions.
SMART on FHIR app development
The launch framework that puts a copilot inside the EHR.
HIPAA requirements for software teams
The safeguards we build into every system that holds patient data.
More in Healthcare
More Healthcare Solutions
Explore other production-grade engineering platforms we deliver across healthcare.
Cross-Industry Reach
The Same Engineering Patterns, Across Other Industries
The technical patterns behind this platform translate naturally into adjacent verticals.
Frequently Asked Questions
An ambient scribe starts from the audio of a visit and drafts the visit note. A clinical copilot starts from the patient record and the inbox, and drafts summaries, replies, letters and orders. Many health systems run both; our AI medical scribe page covers the scribe side.
It depends on what each function does. FDA's Clinical Decision Support Software guidance sets out which functions fall outside the device definition. We map each copilot function against those criteria with your regulatory team; the determination stays with you and your advisers.
No. We build copilots that suggest orders as drafts or CDS Hooks suggestions. A clinician accepts, edits or rejects each one in the EHR's own ordering workflow, and only accepted orders are filed.
EHRs that support SMART on FHIR app launch or CDS Hooks, which includes the major platforms such as Epic and Oracle Health (Cerner). Which launch points and write-back options your EHR version supports is confirmed during scoping.
With HIPAA-ready engineering: Agnotic signs a business associate agreement, the model runs in your cloud account or a covered model service, data access is limited to the launch scopes, everything is encrypted, and every suggestion is logged. HIPAA compliance itself is assessed for your organization, not certified for a product.
The number of copilot tasks, the EHRs and launch methods, whether write-back is needed, the size of the clinician test sets, model hosting, and the length of the pilot. We describe these drivers on the scoping call rather than quote a price before it.
It depends on scope. We agree a fixed first release on the scoping call, usually one copilot task in an EHR sandbox with grounding and logging in place, and give you a timeline for that release.
AI tools help our engineers draft code and tests. An engineer reviews every change, and every copilot task is checked against its clinician-reviewed test set before each release.
Industries