Amplifire
Amplifire is an adaptive learning company in Boulder, Colorado, founded in 2000, whose healthcare business trains clinicians and staff for health systems. Its platform asks learners both what they know and how confident they are, then targets confidently held misinformation, the answers a clinician is sure of but gets wrong. Cognix, the company's own intelligence layer trained on more than five billion learner interactions, powers Cognix Learn, an AI instructor that answers voice or text questions inside a course while staying within approved training content, and Cognix Author, which turns existing documents into adaptive modules and draft questions for authors to review.
The Healthcare Alliance brings together more than 40 health systems with over 600 acute hospitals, including Intermountain Healthcare, MedStar Health, UCHealth and Providence, to build and share more than 100 courses on EHR training, safety and quality, obstetrics risk, revenue cycle and compliance. In a KLAS Arch Collaborative case, UC San Diego Health cut initial Epic training time by 75% to 79% for 508 providers compared with trainer led sessions, and Stanford Health Care reports onboarding 50% faster with ServiceNow automation.
Epic and Oracle Cerner are subjects of the training rather than systems the platform connects to. The cloud service has held FedRAMP certification since May 2021, and Cognix material claims SOC 2 and HIPAA compliance without a report or business associate agreement behind it, while also stating that Cognix does not hallucinate instructional content. Pricing for the platform, courseware and alliance membership comes through a demo.
Capability Axes
An AI Health Index grade measures what a buyer can verify from public sources on the date shown. It is not a rating of how good the product is. A vendor can build an excellent system and grade low on an axis because it publishes nothing an outsider can check. How grades read
Cognix, the intelligence layer trained on more than five billion learner interactions, drives both halves of the product. Cognix Learn adapts each lesson to what a clinician knows and how confident they are and answers questions by voice or text inside the course, and Cognix Author turns existing material into adaptive modules and draft questions in hours rather than weeks.
Underneath sits the older confidence based learning platform, which still delivers standard courses, reporting and Healthcare Alliance courseware without the generative features. The adaptive engine is what health systems buy, while the conversational instructor and authoring assistant are newer layers on a platform that predates them.
The Cognix page states that Cognix does not hallucinate instructional content and that its constrained architecture is designed to eliminate error risk, and the homepage repeats the first claim. Authors can still inspect and revise generated questions, version history tracks changes, and organizations decide whether the authoring assistant may go beyond source material, so course creation has a review step.
The learner facing instructor answers in real time without instructor review, and the pitch stresses reinforcement without added faculty time. A health system that takes the no hallucination claim at face value has no stated threshold, escalation route or recovery step for a wrong answer already given.
Cognix is described as a purpose built instructional architecture rather than a general purpose language model or one fine tuned on public content, trained on more than five billion learner interactions, decades of domain specific data and the company's patented adaptive learning algorithm.
The platform separates learner facing adaptation and conversation from authoring help, and the help center explains how confidence signals drive targeted learning and how organizations choose what material the authoring assistant may use. Model versions, training set composition, release notes and validation results are not provided, and question version history tracks content changes rather than the models that produced them.
Cognix is presented as the company's own architecture, not bolted onto an existing platform and not a general purpose model, which names a first party intelligence layer for both Cognix Learn and Cognix Author. The privacy policy allows service providers and agents to handle personal information under obligations, without naming them.
Speech handling for voice questions, any outside text generation inside the authoring flow and the cloud host are not identified, and no subprocessor list sets retention and training terms for each party. The FedRAMP listing covers the cloud offering without naming the parties behind the current AI features.
KLAS Research's Arch Collaborative case, issued in November 2025, compares initial EHR training time at UC San Diego Health for 508 providers trained in Amplifire against providers taught by human trainers: 75% to 79% less time across ambulatory, hybrid, acute and emergency settings, with 330 hybrid users dropping from 5.5 hours to 1.25.
In a survey of 19 Amplifire users and 154 trainer taught learners, Amplifire learners were twice as likely to call training relevant, and 577 instances of confidently held misinformation surfaced in the first emergency department wave. How groups were assigned is not stated. Stanford Health Care reports 50% faster onboarding and 8% to 17% higher retention, without learner counts.
The privacy policy covers account details, learner answers and confidence selections, limits their use to delivering the contracted service, keeps them while an account is active or as needed for service and legal obligations, and stores personal information in the United States. The AI instructor stays within approved course content.
Whether customer courses and learner interactions feed Cognix, which is described as trained on billions of learner interactions, is not said, and there is no retention schedule for instructor conversations or author uploads and no rule against placing patient information in source documents. How a harmful generated instruction would be contained and reported is not described.
The Cognix page calls the platform SOC 2 and HIPAA compliant, and the privacy policy covers learner accounts, course answers and confidence data reported to instructors and administrators. No business associate agreement is offered or described, and the training service is not positioned as handling protected health information.
Courses on EHR workflows and clinical safety can draw on patient scenarios, yet no rule keeps real patient information out of documents an author uploads to Cognix Author or questions a learner types to the AI instructor. The compliance claim stands without the agreement that would make it binding for a health system.
FedRAMP lists the Amplifire cloud service as certified since May 14, 2021 through an agency authorization under Rev5, as a low impact software as a service offering in ongoing certification, package FR2012366611. The listing's description cites 3.5 billion learner interactions, against more than five billion in current Cognix material, so whether Cognix Learn and Cognix Author sit inside that boundary is open.
The Cognix page also claims SOC 2 compliance without a report or trust portal behind it. Cloud Security Alliance shows a Level 1 self assessment, now marked deprecated, which is not a third party attestation.
The product teaches clinicians and staff: EHR workflows, safety and quality essentials, obstetrics risk reduction, revenue cycle and regulatory compliance, measured by what learners know and how confident they are. It makes no decision about real patients and gives no diagnostic or treatment advice, and no device claim appears for the platform or for Cognix. FedRAMP certification concerns the security of the federal cloud offering, a different regime from device regulation. Scoped as workforce education, the product sits outside clinical decisions, though the AI instructor teaches clinical content to staff whose later practice it can shape.
The AI instructor's answers stay bound to the organization's approved training content, and course owners decide whether Cognix Author may draw on knowledge beyond the uploaded source material, both real controls over what learners are taught. Learner analytics flag confidently held misinformation and struggling learners for instructors.
What is missing is any look at whether adaptation or instructor answers work differently across professional roles, prior EHR experience, languages or accessibility needs, and no governance framework covers how Cognix changes are reviewed before they reach a health system's learners. Outcome results by care setting measure training time rather than fairness.
Course authors review and edit every question Cognix Author drafts, and version history lets them restore an earlier item, while administrators use learner analytics to revisit curriculum and follow up where clinicians hold confident but wrong beliefs. The site terms govern use of the website, and no service agreement on the site offers a warranty on the accuracy of generated instruction or a remedy when the AI instructor teaches something wrong.
The homepage claim of a 40% reduction in compliance violations carries no source or sample. For courses that teach safety and quality practice, correcting a bad answer rests with the health system's own educators.
Epic and Oracle Cerner are subjects of the training, not systems the platform exchanges data with. UC San Diego Health and Stanford Health Care both used Amplifire for Epic onboarding, with learning paths adjusted for prior Epic experience, and Stanford paired it with ServiceNow automation so analytics on struggling learners generated support requests. Nothing shows the learning platform reading from or writing to a clinical record, and no EHR interface is offered. The ServiceNow link connects training to workforce support workflows, which suits an education product and stays outside the clinical record.
The platform is hosted, and the privacy policy stores and processes personal information in the United States. The cloud service also carries FedRAMP certification as a low impact software as a service offering, listed since May 2021, which defines one assessed environment. No self hosted option exists for health systems, and how institutions are separated on the shared platform is not described.
Whether Cognix Learn and Cognix Author run inside that same assessed environment, or rely on separate infrastructure for model processing, is not stated, so the residency statement covers personal information rather than every AI component.
Healthcare buyers reach Amplifire through a demo or consultation, and no price appears for learners, courses, sites or Cognix features. Healthcare Alliance membership brings an annual virtual summit, AI authoring tools, early access to new courseware, more than 100 peer built courses, shared efficacy insights and affinity groups, but its fee is not stated either.
The pieces of a purchase are clear, namely the adaptive platform, authoring, courseware and alliance membership, and KLAS rates the UC San Diego Health implementation as a budgeted cost delivered within six months. How charges scale with staff numbers or facilities only emerges in a sales conversation.
Health systems are the healthcare buyer, and the validated settings are named: Epic onboarding at UC San Diego Health across ambulatory, hybrid, acute care and emergency providers, including physicians, advanced practice providers and house staff, and at Stanford Health Care from inpatient nurses through specialty providers.
The Healthcare Alliance spans more than 40 health systems with over 600 acute hospitals and 2 million employees, including Intermountain Healthcare, MedStar Health, UCHealth, Providence, Texas Health Resources and UW Medicine. Courses cover EHR training, safety and quality, obstetrics risk and revenue cycle, and the same platform is sold to life sciences, higher education and other industries.
Alternatives to Amplifire
Other products that do the same kind of work as Amplifire, and what each one does.
Xuron provides AI virtual humans for practicing healthcare conversations. Learners speak with simulated patients, caregivers or clinicians, receive feedback and repeat an activity.
Perceptors is an AI native learning platform for medical societies, universities, health systems and healthcare employers that want to turn their own approved clinical expertise into certified courses.
MedVR Education provides virtual patient encounters and XR clinical skills training for universities, nursing programs and hospital educators.
MedLern sells a workforce development platform for hospitals and medical colleges. Training departments can create courses, enroll staff, schedule classroom or simulation sessions and track completion and compliance.
AI clinical simulation for nursing education and health system workforce training, delivered in immersive VR and screen-based formats.
SimX provides multiplayer VR simulation for healthcare education, combining virtual patients, clinical equipment and shared team encounters.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
| Entry Price | Pricing Basis | BAA Tier | Implementation | Source |
|---|---|---|---|---|
|
Contact sales
|
The adaptive learning platform, Cognix authoring and healthcare courseware are sold through a demo and contract, and Healthcare Alliance membership carries no listed fee. | Cognix material claims HIPAA compliance, but no business associate agreement or healthcare tier is offered. | Cases describe curriculum design and deployment support, and KLAS rates the UC San Diego Health implementation as a budgeted cost delivered within six months, with no fee stated. | Vendor Published |
No price appears per learner, site or course, and no minimum or bundle price is given for Cognix Learn or Cognix Author.