Heidi has launched a new version of its platform that uses AI agents to carry out administrative tasks around patient visits, with clinicians retaining oversight and approval.
Heidi II adds Agents, Memory and Agentic Search to the company’s existing clinical documentation platform, allowing it to act on tasks identified during and around consultations rather than simply surfacing them.
Heidi said it currently supports 2.8 million patient visits each week globally and identifies around 10 million follow-up tasks.
It cited figures showing doctors spend an average of 8.7 hours a week, or 16.6 per cent of their working hours, on administration, with tasks typically spread across four to six systems used during a patient visit.
Heidi co-founder and CEO Dr Thomas Kelly said the new platform was designed to take on some of those tasks while keeping clinical decisions with clinicians.
“We started inside the visit by capturing the note and gathering the care context. Now Heidi can create the tasks and complete them from referrals and chart prep to arranging follow ups,” he said.
“It handles that work as a partner, while clinical judgement and approvals remain with the clinician.”
Agents and routines
Heidi II introduces a new Home interface bringing together a clinician’s patients, sessions and tasks, with relevant clinical context loaded into the system.
Clinicians can ask Heidi to carry out tasks using natural language, with an AI agent then planning and executing the work before returning the result for review.
The company said early use cases include preparing charts for a day’s patients, analysing practice visits or referral communications and identifying patients due for immunisations, blood tests, screening or recalls.
Capital ENT & Sinus Center’s Dr Raymond Brown said the platform was already taking on preparation work between consultations.
“Heidi does the preparation I used to squeeze in between patients,” Dr Brown said.
“It loads my schedule, finds the pathology results in the chart, and writes a pre-visit note I can read in thirty seconds.”
Agents can also be saved and scheduled as Routines for individual clinicians or practices and run across patient lists.
Heidi said the agents can interact with systems including electronic health records, calendars, inboxes and patient portals. If an agent cannot complete a task, the platform is designed to report this and explain why.
Approval controls can be configured according to the workflow and settings established by clinicians or healthcare organisations.
The company said capabilities would also be staged according to region and setting, with its internal clinical safety teams reviewing features before and after release.
Search and integrations
Heidi II also expands the platform’s search and integration capabilities.
Agentic Search will draw on full-text journal articles from publishers and sources including the New England Journal of Medicine, Wiley and Cochrane, as well as BMJ Group clinical decision support information.
Search results will include citations, with subscribers to the relevant journals able to follow them through to full articles.
New connectors are also designed to allow Heidi to work across existing practice systems, including pulling appointment lists, querying EHRs, managing appointments and clinical mail and following documented practice procedures.
Heidi II begins rolling out in English on September 29, with additional languages to follow. The new capabilities will not initially be available in the UK or European Union.
Heidi said its Scribe product has supported more than 175 million visits and 67 million clinical hours globally, while Heidi Evidence has answered more than 10 million queries since launch.
Add Comment