Skip to main content

2026.9.7 the assistant stops answering what it cannot answer, and an invitation stops promising a message

A release about not saying things that aren't true. When a patient texts something the approved wording doesn't cover, the assistant now hands the conversation to a coordinator in silence instead of replying with a holding sentence. And an invitation to the portal no longer announces a secure message when nobody has written one.

The assistant stops sending replies it can't stand behind​

If your clinic has approved replies set to Strict, a message the approved wording doesn't cover is now passed to a coordinator without a reply being sent. Previously it was answered with a holding sentence ("Thanks, we've passed this to our team") which, by design, could not relate to what the patient had actually said. A patient answering one of your questions with "YES" received it. So did someone who sent the email address your team had just asked for.

The message still reaches you: it is recorded on the conversation and raised for a coordinator exactly as before. Only the automatic reply is gone.

What still answers, unchanged: emergency wording, consent, identity checks, and the carrier keywords: a patient who texts STOP is still opted out and still told so. Those are approved wording, and Strict has always sent them.

A text from a number your clinic doesn't recognise is now also recorded for staff without the automatic "we can't reply to it here" response.

Clinics on the default setting (Free) are unaffected. The assistant answers in its own words as it always has. See patient messaging.

An invitation no longer promises a message that isn't there​

An invitation to the prospect portal used to say "You have a new secure message waiting for you" (in the subject line, the inbox preview and the body) whether or not a message had been written. Someone invited to see where their referral stands would follow the link and find an empty thread.

An invitation now says their secure access is ready. Only an actual secure message announces a message.

The portal says who a message is from​

In the patient portal, a message from your clinic is shown as being from the clinic. Patients were seeing a raw email address, or an internal account name, in the From line. A thread is also titled "Patient message" rather than "Secure message", so a list of them can be told apart.

The menu is worded for whoever is reading it: a patient sees My care where your team sees your own wording.

AI Scribe: record first, attach the patient after​

A visit can now be recorded before the patient is identified, a walk-in, or a room already under way. Start from the Scribe area or the chart, and attach the patient afterwards; the note cannot be signed until they are attached.

Note templates are versioned rather than overwritten, so a note written months ago still shows the wording it was written against. The templates page shows when each template was last used and who wrote it. See AI Scribe.

Smaller things​

  • The privacy and cookie policy links in the cookie notice were white on white and could not be read.
  • A colour that did not meet contrast is no longer used as text.
  • AI Scribe: the note says when a measurement was stated but could not be reduced to a number, instead of leaving it out.