October was focused on helping practices manage correspondence more effectively, with a particular emphasis on making tasks easier to find, assign and complete. We've also introduced document merging, continued to improve the way tasks fit into the letter-review workflow, and made significant improvements to the accuracy and reliability of our AI.
Merge Documents
We've introduced the ability to merge multiple documents into a single piece of correspondence, giving users more flexibility when working with letters in BetterLetter.
This makes it easier to bring related documents together without needing to download them and combine them using external software, helping practices keep correspondence organised and process letters more efficiently. Read more about this feature here.
A New Tasks Experience
We've been working on a new tasks experience to give practices a clearer view of outstanding work and more flexibility in how tasks are managed and assigned.
The new experience introduces several improvements, including:
A clearer view of outstanding work: View tasks across different taskboxes, filter by team and access tasks directly alongside the letters they relate to.
2. More flexible task assignment: Assign tasks to individuals or shared team queues, with a dedicated My Tasks view for work assigned directly to you.
3. Improved task management: Add notes to individual tasks, navigate between tasks linked to the same letter and manage taskbox members to better reflect how your practice works.
The new experience is initially being piloted with a small number of EMIS practices, with a wider EMIS rollout planned before becoming available as an option for Docman and SystmOne practices.
Want to find out more? We will share a more detailed overview of these features when it is released.
Vaccinations table
Due to updated guidance from NHSe and a change request from Enlivio, we have changed the way you see vaccinations. They now appear in a table of their own, with the ability to specify whether the vaccination is GMS or not.
AI Improvements
We've continued to improve the accuracy of our AI, with October's work focused on addressing specific errors in the way clinical information is interpreted and coded.
The latest AI release includes targeted fixes across a number of areas, including:
More accurate interpretation of clinical information
We've improved how the AI interprets numbers and quantities in clinical text. For example, a statement such as "smokes 1–3 cigarettes a day" can now be classified more appropriately as Light Smoker.
Fewer incorrect clinical matches
We've addressed cases where ordinary words were incorrectly matched to clinical abbreviations or terminology. This includes situations where "us" was interpreted as ultrasound (US), "toe" as transoesophageal echo (TOE), or words such as "fit" and "fitting" were interpreted as clinical terms when the surrounding context meant something different.
Improved medication and coding accuracy
We've made targeted improvements to reduce medication false positives and overly broad coding. This includes cases where ordinary words were incorrectly identified as medications, or where a specific phrase such as "Heat Stroke" was coded too broadly as Stroke.
Better handling of abbreviations and text formatting
Further improvements address ambiguous abbreviations, including OD, CAP, RTA, MVA and PV, where interpretation depends on context. We've also improved handling of names, initials, qualifications, organisations and variations in hyphenation.
These changes are part of our ongoing work to make AI processing more reliable and reduce the need for manual correction.
Looking Ahead: November and December
Our focus will remain on improving the reliability of AI and making it easier for practices to manage outstanding work.
Following our October AI release, we're working towards a more regular deployment schedule, with the next release currently expected in early December. Our longer-term goal is to move towards monthly releases during 2027, making it easier to deliver improvements consistently and address issues more promptly.
We're also working to strengthen the way we identify, prioritise and track issues, so that known problems have clear ownership and appropriate expectations for resolution.
Looking further ahead, we'll be building on the targeted AI fixes with more systemic improvements to the areas that have the greatest impact on accuracy. This work will help us move towards a system that is more accurate, easier to test and more maintainable over time.
See you next month!