INVESTOR FIELD GUIDE / REVIEWED OCTOBER 23, 2026

Healthcare AI market map

Compare the job, the buyer, and the evidence.

Browse healthcare artificial intelligence (AI) products by the work they do. Use the company examples to shape your diligence questions, then test each claim in the setting where the product would be used.

16company examples
across four workflow categories

A selective, non-ranked research map. Not an investment recommendation or clinical endorsement.

One primary workflow, visible overlaps

The four groups mix a care workflow, an operating function, an enabling role and a patient interface. They are navigation lenses, not mutually exclusive sectors. Each company appears once for the selected product, with overlaps called out on its card.

Company summaries describe vendor-stated functions. Buyer and commercial lenses and diligence questions are our research interpretation, not verified buying committees or per-customer costs and revenue. Pricing was not checked. No selected product was independently tested for this map.

CATEGORY 01

Clinical-team workflows

Documentation tools support the encounter; imaging products flag suspected findings and notify care teams. Likely buyers include health systems, provider groups and clinical service lines. Adoption depends on review responsibilities, local workflow fit and integration with the health record.

Documentation is not diagnosis. Imaging triage, notification and care coordination are distinct from autonomous diagnosis or treatment. The U.S. Food and Drug Administration (FDA) reviews device functions, not entire companies.

Encounter-note drafts

Abridge

Drafts structured encounter notes from patient-clinician conversations and surfaces chart context for clinicians.

Buyer / model lens

Health systems and clinical leadership; enterprise clinician-workflow software.

Overlap

Administrative documentation; evidence retrieval.

Evidence boundary

Notes still need clinician review. Documentation support does not establish diagnostic or treatment benefit.

Ask in diligence

How much editing remains by specialty, and what consent, signoff and source-traceability controls work in production?

Abridge product

Imaging triage & coordination

Aidoc

Connects imaging AI results with radiology workflows to flag suspected findings and support downstream care-team coordination.

Buyer / model lens

Hospitals, radiology and clinical service lines; a platform with selected applications.

Overlap

Deployment infrastructure; care coordination.

Evidence boundary

Device authorizations are specific to a device and intended use. The linked 2022 FDA record is a historical triage example, not a current platform-wide clearance inventory.

Ask in diligence

Which models and indications are deployed, how are false alerts and misses measured locally, and who monitors performance after deployment?

Aidoc radiology · FDA record (2022)

Ambient documentation

Nabla

Captures clinical conversations and drafts structured notes. Its wider platform also markets dictation and coding suggestions.

Buyer / model lens

Provider groups and health systems; clinician-workflow software.

Overlap

Administrative documentation; coding support.

Evidence boundary

Documentation is not diagnosis. The source says final coding decisions remain with the clinician. Verify local output review and integration.

Ask in diligence

What are the edit burden, clinician retention and error-handling process on the buyer’s own case mix?

Nabla platform

Stroke notification & coordination

Viz.ai

Analyzes imaging for suspected large-vessel occlusion, a blockage in a major brain artery, and notifies specialists through a stroke-team workflow.

Buyer / model lens

Hospital stroke networks and neurovascular teams; clinical-workflow software.

Overlap

Communication infrastructure; care coordination.

Evidence boundary

The cited historical FDA record describes notification-only use, not diagnosis or a replacement for full patient evaluation. Verify the deployed version and labeling.

Ask in diligence

Does earlier notification change the local care pathway, and what evidence separates workflow effects from patient-outcome claims?

Viz LVO · FDA notification record

Back to categories ↑

CATEGORY 02

Administrative operations

These products address payment and operational work around care. Likely buyers include revenue-cycle, intake and hospital operations teams. Compare the work completed with exception handling, audits and integration costs, rather than counting automated actions alone.

Administrative placement is not a low-risk label. Routing, risk flags, patient instructions and escalation can still affect care.

Inpatient coding

AKASA

Reads inpatient records, assigns encounter codes and links code choices to supporting chart language within customer-set boundaries.

Buyer / model lens

Hospital coding, health-information management, revenue-cycle and finance teams; automation software.

Overlap

Clinical documentation.

Evidence boundary

Coding a documented encounter is different from making a diagnosis. Vendor comparisons and performance claims were not independently validated here.

Ask in diligence

What are the unsupported-code, correction, denial and audit rates after deployment, and who is accountable for submitted claims?

AKASA coding

Healthcare communication

Infinitus

Automates interactions such as benefits verification, prior-authorization follow-up, enrollment, scheduling and refill coordination.

Buyer / model lens

Provider, payer and life-sciences operations teams; enterprise agent workflows.

Overlap

Patient-facing communication; agent-building infrastructure.

Evidence boundary

Verify permitted scripts, disclosures, human escalation and payer acceptance for the specific program. Vendor compliance or results claims are not independent certifications.

Ask in diligence

What happens when a caller disputes a response, a payer changes its process or a potentially clinical issue needs human escalation?

Infinitus agents

Pre-admission coordination

Qventus

Supports pre-admission testing teams with chart review, risk surfacing, call preparation, records requests and patient outreach.

Buyer / model lens

Hospital perioperative and pre-admission leaders; operational-workflow software.

Overlap

Clinical risk support; patient communication.

Evidence boundary

An operations buyer does not make every function low-risk. Risk surfacing and patient instructions can affect care; classification and clinical appropriateness were not determined.

Ask in diligence

Which actions need clinician review, and how are missed risks, inappropriate instructions and local escalation performance audited?

Qventus coordination

Referral intake & coverage

Tennr

Collects referral documents, extracts information, evaluates coverage requirements and coordinates operational follow-up across providers, payers and patients.

Buyer / model lens

Specialty providers, pharmacies, health systems and intake teams; workflow automation.

Overlap

Patient communication; clinically informed routing.

Evidence boundary

Operational routing is not proof of diagnostic triage or guaranteed payer approval. Check the exact decisions and escalation rules.

Ask in diligence

How are payer-rule changes, missing information, exception queues and inappropriate routing detected and corrected?

Tennr product · Tennr overview

Back to categories ↑

CATEGORY 03

Data, evidence & deployment

These products help structure healthcare data, generate observational evidence, supply training data or deploy and monitor AI. Buyers may sit in quality, research, informatics, technology or model-development teams; software, data licensing and services have different costs and dependencies.

Infrastructure is an enabling role, not a clinical-performance category. A data license or platform listing does not validate downstream outputs.

Observational evidence

Atropos Health

Provides infrastructure for patient-group queries and observational evidence generation: studies of existing healthcare data rather than randomized treatment trials.

Buyer / model lens

Health-system research, evidence and informatics teams, plus life-sciences evidence teams; evidence infrastructure.

Overlap

Clinical evidence support; life-sciences research.

Evidence boundary

Observational analysis is not randomized treatment evidence. Check missing data, alternative explanations, data provenance and permitted uses.

Ask in diligence

Can an independent reviewer reproduce the patient group, assumptions and statistical choices, and assess relevance to the target population?

Atropos GENEVA OS

AI deployment & monitoring

Ferrum Health

Combines deployment tooling, production-model monitoring and a model marketplace for healthcare organizations.

Buyer / model lens

Health-system AI, information-technology, informatics and service-line leaders; platform tooling.

Overlap

Clinical imaging applications.

Evidence boundary

A marketplace listing does not establish authorization for every model or use, or improved local outcomes. Validate each model and deployment separately.

Ask in diligence

Who supplies trusted reference answers, how are changes in model performance and results across patient groups monitored, and which thresholds trigger model rollback or review?

Ferrum platform · Ferrum overview

Clinical registry data

Layer Health

Reads records and suggests registry-field answers with chart evidence. Offers software for internal abstractors and AI-supported abstraction services.

Buyer / model lens

Health-system quality and registry teams; software and optional service delivery.

Overlap

Administrative quality and registry operations.

Evidence boundary

Placed here because the selected product structures clinical data. Field-level validation matters; advertised accuracy and time savings are not independent findings.

Ask in diligence

How are trusted reference answers built per registry field, how are guideline changes versioned and what is the error rate after automated submission?

Layer registry product

Training & evaluation data

Protege

Sources and curates healthcare data, including structured records, notes and imaging, for AI training and evaluation.

Buyer / model lens

Healthcare AI builders and research teams; data licensing and evaluation services.

Overlap

Clinical and administrative model development.

Evidence boundary

An enabling data supplier is not a care application. Vendor privacy descriptions are not a legal audit or proof of representative, uncontaminated evaluation data.

Ask in diligence

Do contracts permit the training use, and can the supplier show provenance, population coverage, de-identification review and separation from evaluation sets?

Protege healthcare

Back to categories ↑

CATEGORY 04

Patient-facing interactions

These products communicate directly with patients or consumers. Some distribute to individuals; others are bought by health systems, employers or program operators. The person using a product may not control its budget or approve its clinical workflow.

Patient-facing describes the interface, not autonomy, clinical evidence or regulatory status. Human handoff and emergency boundaries need separate review.

Symptom information

Ada Health

Combines a medical knowledge base with user-reported symptoms to provide AI symptom-assessment information and next steps.

Buyer / model lens

Consumers and enterprise partners; distribution differs by product and market.

Overlap

Provider digital front door.

Evidence boundary

The checked U.S. terms say the selected assessment does not diagnose or provide treatment plans. Do not generalize one region’s device status to every version.

Ask in diligence

What is the intended use of the exact local version, and how are missed urgent needs, emergency messaging and accessibility tested?

Ada app · Ada U.S. terms

Patient outreach & follow-up

Hippocratic AI

AI Care Companion markets preparation calls, recovery check-ins, protocol-based monitoring, caregiver communication and clinical escalation for care episodes.

Buyer / model lens

Healthcare, life-sciences and medical-technology programs; enterprise communication workflows.

Overlap

Care coordination; life-sciences and medical-technology workflows.

Evidence boundary

The homepage says agents do not diagnose or prescribe. Instructions and escalation still need use-case-specific review; patient-facing does not mean autonomous or proven safe.

Ask in diligence

Which instructions and actions are permitted, how quickly do escalations reach a staffed team and what failure and adverse-event data are available?

AI Care Companion · Hippocratic AI limits

Health-system digital front door

K Health

PatientGPT provides a health-system-branded conversational layer for intake, care navigation, scheduling, messaging and handoff to care teams.

Buyer / model lens

Health-system digital, access and clinical teams; an enterprise platform.

Overlap

Intake; scheduling; clinician communication.

Evidence boundary

This example is enterprise PatientGPT, not the historical consumer service. Local policies, protocols and escalation thresholds matter; marketing references do not validate every feature.

Ask in diligence

Which requests can it complete, when must it hand off and does local testing establish appropriate routing across languages and risk levels?

K Health PatientGPT

Mental-wellbeing self-help

Wysa

Offers reflective AI conversation, guided journaling, mindfulness and other self-help exercises, with signposting to support resources.

Buyer / model lens

Individuals and organizations, including employers, insurers and healthcare services; distribution varies.

Overlap

Employer benefits; clinical adjunct workflows.

Evidence boundary

Its FAQ describes support rather than medical advice or diagnosis, and excludes crisis use. It is not a replacement clinician or crisis service.

Ask in diligence

How are distress recognition, escalation availability, drop-off, adverse events and clinical handoffs evaluated for the intended population?

Wysa self-help · Wysa use limits

Back to categories ↑

Compare the deployment, not the demo

Compare companies within the same workflow and intended use. A note draft, a patient call and an imaging alert need different evidence, even when all three are described as AI.

01
Workflow & budget
Name the user, budget owner and person approving deployment. Identify the task being replaced or added, the existing alternative and the cost of leaving exceptions for staff.
02
Human responsibility
Trace who reviews outputs, signs off and handles failures. For patient contact, verify disclosures, staffed escalation and responsibility when a handoff fails.
03
Evidence in context
Ask for study methods and local results for the exact version, workflow, population and site. Separate vendor benchmarks, workflow measures and patient outcomes; examine misses and subgroup performance.
04
Data & integration
Check data provenance, consent, contracts, training rights, health-record connections and ongoing support. Ask how updates, security incidents, data access and model changes are governed.
05
Intended use & obligations
Verify the exact function, market, labeling and responsible organization. A device authorization or privacy statement does not apply automatically to a whole company, every feature or every deployment.
06
Economics & adoption
Separate pilot activity from repeat paid use. Test price, implementation effort, human-review costs and retention. Do not assume payer reimbursement, budget availability or savings from an automated-task count.

A PRACTICAL COMPARISON

When comparing two documentation tools, use the same specialty and encounter types. Measure note-editing time, signoff effort and errors alongside usage and total cost. An imaging-notification product belongs in a separate comparison with its own intended use and local alert-review pathway.

Methodology & evidence limits

Reviewed October 23, 2026. This is a dated source review, not a publication date or a live company database.

Selection & ordering

We selected 16 publicly marketed, healthcare-specific AI or enabling data/deployment products with enough first-party detail to describe their workflow. This is an editorial sample, not a census or ranking. Payment, funding totals and valuations were not selection or ordering criteria. Companies are alphabetical within each group.

Included does not mean investable

The examples include smaller startups and more established businesses. Funding, solvency, independent ownership and current investment access were not audited. Inclusion is not a recommendation, a portfolio-status statement or evidence that an angel can invest.

Sources & verification

Official product sources linked on the cards were reviewed October 23, 2026. Most are live product pages rather than dated research publications. This review checks the marketed workflow, not clinical efficacy, adoption claims, safety, accuracy, returns or promised savings. Sources and products can change.

Clinical & regulatory limits

FDA device review concerns specific functions and intended uses; do not call an entire company FDA approved. The linked Aidoc and Viz records are narrow historical examples, not current clearance inventories. Privacy obligations depend on the product, data flow and organizations involved; a Health Insurance Portability and Accountability Act (HIPAA) claim is not proof of clinical safety.

Outside this map

We excluded general-purpose model/cloud vendors, public-stock selection, drug-discovery pipelines, research concepts without public product detail and an exhaustive digital-health directory. An omitted company is not a negative product or investment judgment.

Read reference maps carefully

Menlo’s October 2025 report uses a narrower healthcare-specific generative-AI spending definition. Bessemer’s September 2024 roadmap supplies investor lenses; Elion supplies a live product taxonomy. None is a current market census or independent validation of the selected products. This map’s entries were checked against company sources.

Practice the questions before writing a check

Our Angel Squad angel-investing community combines education and peer learning. Use those conversations to test your assumptions, then do company-specific diligence.

Explore Angel Squad

This educational map is research, not investment, medical or legal advice, a procurement endorsement or an offer to invest. Company inclusion does not establish clinical effectiveness, safety or regulatory status. Startup investments are speculative, illiquid and long-term; most may fail and you can lose all invested capital.

One primary workflow, visible overlaps

Clinical-team workflows

Abridge

Aidoc

Nabla

Viz.ai

Administrative operations

AKASA

Infinitus

Qventus

Tennr

Data, evidence & deployment

Atropos Health

Ferrum Health

Layer Health

Protege

Patient-facing interactions

Ada Health

Hippocratic AI

K Health

Wysa

Compare the deployment, not the demo

Workflow & budget

Human responsibility

Evidence in context

Data & integration

Intended use & obligations

Economics & adoption

Methodology & evidence limits

Selection & ordering

Included does not mean investable

Sources & verification

Clinical & regulatory limits

Outside this map

Read reference maps carefully

Practice the questions before writing a check

Building something hilariously early?