Heidi II and agentic AI: what might it mean for surgeons?

What has been announced

On 28 September 2026, Heidi announced Heidi II, a new version of its platform. The company describes it as moving from documenting a patient visit to acting on the administrative work around it, with clinician supervision.

According to the company, the release introduces three main capabilities:

• Agents, which can prepare for clinics or ward rounds, chase outstanding results, complete referral forms and draft letters
• Memory, which carries context across a clinician's patients, tasks and systems
• Access to peer-reviewed research within the platform

The company also describes a new home screen that brings scheduled patients, open tasks and preparation into one view. Heidi reports that its scribe has supported more than 175 million visits globally. These figures come from the company and have not been independently verified by ASTI.

From scribe to agent

Ambient voice technology, often called AI scribing, drafts a note from a consultation for the clinician to check. An agent goes further: it can take actions across calendars, inboxes and clinical systems. That is a meaningful change in the kind of risk involved. An error in a draft note can be corrected before signing. An error in an action, such as a referral sent for the wrong patient or an incorrect entry written back to the record, may reach patients or colleagues.

Heidi states that, by default, drafts wait for clinician review and any write-back to the record pauses for approval. How these controls work in practice, and how they are configured locally, will matter as much as the default settings.

Possible relevance to surgical practice

Much surgical administration sits around the operation rather than in it. Areas where teams may be interested in agentic tools include:

• Clinic letters, discharge summaries and GP correspondence
• Pre-assessment preparation and pre-charting before ward rounds
• Chasing pathology, imaging and outstanding investigations
• Referral forms and onward requests
• Handover preparation

Any time released would depend on local systems, integration and how much checking the outputs require. Published independent evidence on agentic tools in surgical settings is currently limited.

Points for surgical teams to discuss

• Which specific tasks, if any, would the team be comfortable delegating, and which should always remain manual?
• Who reviews each action before it leaves the system, and how is that review recorded?
• Does the tool write to the electronic patient record, and can that be switched off or restricted?
• Has the organisation completed clinical safety assessment (DCB0129 and DCB0160), a DPIA and DTAC review for this use?
• What is the product's current regulatory status in the UK for the functions being used?
• How are patients informed that AI is involved in their care administration?
• How would the team detect and report an error made by an agent?
• What happens to workflow and safety if the tool is unavailable?

The existing NHS position

NHS England has published guidance on the use of AI-enabled ambient scribing products in health and care settings. It sets out expectations on clinical safety, data protection, regulatory status and clinician oversight. Organisations considering agentic features may find it a useful starting point, although tools that take actions may raise additional questions the guidance was not written to cover.

Summary

Heidi II is one example of a wider move from AI that writes text to AI that carries out tasks. For surgeons, the potential lies in routine administration. The main questions concern oversight, accountability, integration and evidence. Decisions about adoption rest with each organisation's clinical, governance and information governance leads.

ASTI shares this overview for information only. It does not endorse Heidi, its products or any other company mentioned, and it is not clinical, legal or regulatory guidance.