TORTUS Health and Heidi II approach AI-assisted care from different starting points. TORTUS Health's October 2026 description presents a patient-facing service with clinician-authorised care. Heidi II supplies capabilities for clinicians and practices working around existing visits. The distinction is not simply which agent can do more, but who brings the patient relationship and organises the clinical work.
This comparison is published by iatroX and includes iatroX as a separate clinical-reference and learning option. It is a documentary comparison, not a head-to-head product test or evidence that one platform is clinically safer.
Who brings the patient relationship?
The TORTUS Health service materials, recorded in research on 9 October 2026, describe a California-first, staged virtual-care proposition. Its terms distinguish the technology company from the clinical provider and describe proposed care plans requiring clinician approval.
Heidi's 28 September 2026 launch announcement describes agents operating around clinical work, with rollout beginning on 29 September. That is a tool proposition for clinicians and teams, not evidence that Heidi itself becomes the patient's treating service in every deployment.
The practical difference is visible before any model runs. A patient entering virtual care needs the service to explain who will assess them and what support is included. A practice deploying an agent already has a clinical relationship, but must decide how the software fits its existing duties and communication arrangements.
Who supplies the human work?
AI-assisted care still contains professional decisions, clarification and exceptions. In TORTUS Health's described model, clinician review is part of the care pathway. In a Heidi II deployment, the organisation must establish who reviews and authorises the relevant work under its configured workflow.
A fictional example illustrates the distinction. A patient submits a new concern to a virtual service and requires further questioning before a plan can be authorised. Separately, a practice agent prepares follow-up work after a clinician has already assessed a patient. Both involve human input, but the clinical starting point is different.
The comparison should therefore ask which work the software removes and which work remains with a person. It should not count an AI-generated plan as completed care or assume that a workflow tool supplies additional clinical staff. Those are service-design questions rather than model features.
Heidi II is broader than a scribe
Reducing Heidi to note-taking would make the comparison outdated. Its September 2026 announcement describes agents, scheduled Routines and actions across existing practice systems, with approval controls depending on workflow and settings. These are documented product claims, not independently demonstrated outcomes for every proposed use.
That broader scope creates possible overlap around preparation, communication and follow-through. However, drafting a message, executing an authorised administrative action and independently deciding treatment remain different activities. The same standard should be applied to TORTUS Health and any other platform.
The geographical qualification also matters. Heidi's launch release excludes the new capabilities from the UK and EU. That should not be converted into a claim that every Heidi product is unavailable there, nor should the California-first TORTUS Health proposition be treated as an immediately available alternative for an NHS practice.
A comparison built around responsibility
| Decision | TORTUS Health, October 2026 materials | Heidi II, September 2026 launch description |
|---|---|---|
| Who starts the encounter? | A patient seeking the service's care | A clinician or practice working with existing clinical tasks |
| Who organises professional assessment? | The described clinical-service arrangement | The deploying care team |
| What does AI contribute? | Information gathering and proposed care plans | Preparation, documentation-related work and authorised workflows |
| What remains a human responsibility? | Clinician approval and appropriate care decisions | Clinical judgement and configured review or approval |
| What should a buyer or patient verify? | Eligibility, clinical access, scope and continuity | Supported workflow, regional access, integration and task ownership |
| What outcome matters? | An appropriately managed care episode | Better-controlled completion of the intended clinical workflow |
These are differences in the published propositions, not permanent boundaries between the companies. Products can evolve, and an enterprise deployment may contain more complexity than a public summary reveals. The exact service and contract remain the relevant basis for a decision.
Different economics follow from different operating models
A workflow supplier can aim to make an existing clinical team more productive. A care service must also make the delivery of assessment, support and follow-up workable. The relevant costs therefore cannot be reduced to a software subscription comparison.
For a practice, useful measures might include remaining staff effort, repeated tasks and unresolved actions. For a virtual clinic, they also include patient acquisition, clinical review, continuing contact and the handling of people who fall outside scope. These are proposed evaluation categories, not disclosed company cost structures.
The distinction does not establish that one model is better. A specialist workflow tool could add substantial value without owning the patient relationship. A patient-facing service could improve access without supplying a general-purpose agent to every clinical organisation. Both need evidence for the promise they make.
Where competition could emerge
The strongest potential overlap is not necessarily in the first consultation. It may appear in the information and actions surrounding care: preparing context, explaining next steps, supporting communication and tracking unresolved work. That is a strategic hypothesis rather than a claim that the products already offer identical functions.
A practice could prefer an integrated tool that supports its existing relationships. A patient might prefer a separate service with a simpler route to assessment. A healthcare organisation might consider a partner model that combines selected elements of both. Each choice depends on local needs and the actual service available.
The decisive evidence would be task-specific. Does preparation reduce the need to repeat the history? Does communication preserve the clinical intent? Does a hand-back reach the right person? Broad launch language cannot answer those operational questions.
Evidence and learning remain separate needs
The Medical Board of California's telehealth guidance, checked on 9 October 2026, preserves professional standards across consultation modes. For a UK team considering delegation, the GMC's guidance provides a different jurisdictional reference. Neither should be used as a blanket approval of a vendor.
Under iatroX's September 2026 specification, Ask-iatroX provides source-linked clinical reference, while Tutor and practice tools support professional learning. These functions may help a clinician investigate the reasoning behind a proposed action, but they do not complete the patient's care or validate another agent's output.
A useful learning task is to identify which part of a proposed workflow requires a new clinical decision and which part merely executes a decision already made. That distinction remains valuable whether the clinician is working in a virtual clinic or using automation inside an established practice.
Verdict by reader scenario
For a patient seeking an eligible virtual assessment, examine TORTUS Health's actual access, clinical scope and continuing support. For a clinical team seeking help with work around existing visits, examine the available Heidi II configuration and its approval boundaries. Neither category is an automatic substitute for the other.
For an NHS buyer, regional availability comes before comparison of international promises. For a clinician looking to check knowledge or practise reasoning, a reference-and-learning platform addresses a separate task. The most useful question is who owns the clinical work after the software has produced its output.
Frequently asked questions
Is TORTUS Health a direct replacement for Heidi II?
Not on the basis of the October 2026 descriptions: one is a patient-facing care proposition and the other supports clinicians' workflows. The degree of overlap depends on the exact task and released functionality.
Is Heidi II only a documentation product?
No: the September 2026 announcement describes agents, scheduled Routines and broader workflow actions. Their actual use remains dependent on access, configuration and approval arrangements.
Which is the better option for an NHS practice?
A practice must first establish locally available functionality and deployment requirements. Neither a California-first care description nor an excluded regional launch establishes an immediately usable NHS option.