Period covered: Improvements delivered over the last four weeks.
May Updates:
Overview of these updates
This release focuses on two themes:
Faster analytical workflows through parallel conversations
More robust and predictable analysis experience across visualisation, interaction, and platform behaviour.
The objective is to remove friction in exploratory work while increasing confidence in outputs and system behaviour. Analysts can move between questions without interruption and rely on clearer, more consistent interactions and results.
This is most impactful for:
Exploratory analysis: Parallel conversations
Interpretation quality: Chart robustness and clarity
Workflow continuity: Query readability and handover
Work Faster Across Questions
What has changed?
Protégé now supports multiple active conversations within a single browser tab, allowing analysts to explore different analytical directions in parallel. The conversation overview has been updated to clearly indicate which conversations are still generating responses and which are complete.
Why it matters
Patent analysis is inherently exploratory and multi-threaded. Analysts often investigate adjacent technologies, compare alternative segments, or pursue separate lines of reasoning at the same time.
A single-threaded interaction model forces sequential work, introducing artificial waiting time and breaking analytical momentum.
Parallel conversations remove this constraint, enabling continuous exploration across multiple lines of inquiry and increasing overall analytical throughput.
-
Analytical behaviour
Multiple conversations can run concurrently within one session
Each conversation streams responses independently
Conversations in progress are indicated with a spinner
-
Constraints & implications
In-progress responses cannot be cancelled individually
Conversations are not synchronised across devices or sessions
-
Strict correctness guarantees apply:
All streamed content is scoped to its originating conversation
No cross-conversation leakage is permitted
Responses render in the correct order within each conversation
-
Example
A patent analyst investigating how Pfizer’s patent portfolio has evolved post-Covid explores this question from multiple analytical angles in parallel, rather than committing to a single approach upfront:
One conversation analyses portfolio size and filing trends over time, examining changes in publication volume and growth rates before, during, and after Covid
Another conversation assess technology composition, identifying shifts in focus areas such as vaccines, mRNA technologies, or adjacent therapeutic domains
A third evaluates portfolio strength and impact, using indicators such as citations or Patent Asset Index to understand whether Pfizer’s post-Covid activity reflects expansion, consolidation, or quality shifts
While these analyses run simultaneously, the analyst can see at a glance signaled by a spinner which conversations are still generating insights and which are complete, allowing them to move between lines of inquiry without waiting or losing context.
More Reliable and Predictable Analysis Experience
What has changed?
A set of improvements has been introduced across visualisation, query handling and interaction design to make Protégé outputs and behaviour more consistent and easier to interpret.
Why it matters
Analysts rely on Protégé outputs for decision-making and stakeholder communication. Unclear visuals, limited query visibility, or inexplicit interaction behaviour introduce friction and reduce confidence.
These updates improve clarity, reduce ambiguity, and make workflows more predictable from exploration through to reporting.
-
Analytical behaviour
Charts handle negative values more explicitly where applicable
Data grouping excludes undefined ultimate owners labeled as
<unknown>Prompt input is auto-focused after sign-in and when starting a new conversation
Long search queries are fully readable within the interface
More prominent way to open queries directly in Custom Analysis
-
Example
A patent analyst reviewing a complex query can now read the full syntax directly in Protégé, and open it in Custom Analysis for further refinement:
Protégé will also now filter out unknown ultimate owners: