- Richer Patent Data with New Data Fields
- Better Conversation Management with Rename, Delete, and Retention Transparency
- Improved Patent Number Matching
- Better Intent Recognition for Simple Tasks
- Lower Task Failure Rates and Clearer Recovery Behaviour
- Parallel Conversations for Exploratory Analysis
(Period covered: Improvements delivered over the last six weeks.)
Overview of updates
This release focuses on three themes:
Stronger governance and transparency
Faster and more reliable execution
Clearer visual and search behaviour
The goal is practical: manage work in progress with less friction, trust how web and query driven outputs are produced, and complete day to day analysis with fewer dead ends.
Most impactful for:
Exploratory work - parallel conversations, conversation management, and better continuity across sessions
Daily analytical reliability - lower task failure rates, faster simple tasks, stronger search infrastructure and improved activity signalling
Traceability and defensibility - query transparency for document level insights
Richer Patent Data with New Data Fields
What has changed?
Protégé now incorporates additional patent-level data fields related to citations, annuity fees, and legal events, expanding the underlying data available for analysis.
Why it matters
For patent analysts, technical relevance alone is rarely sufficient. Understanding how a patent is used, maintained, and evolves legally is often what determines its real strategic value.
These new data dimensions enable more nuanced assessments across three critical lenses:
Citations: indicators of technological influence and knowledge flow
Annuity fees: signals of portfolio maintenance decisions and perceived asset value
Legal events: visibility into lifecycle changes such as pending, in force, in active
Together, this moves Protégé closer to supporting portfolio quality and lifecycle analysis.
Analytical behaviour
Protégé can now incorporate these fields when retrieving, filtering, and explaining results. This enables:
surfacing patents with high citation relevance when assessing influential technologies or key prior art
referencing legal events to contextualise changes in status, ownership, or enforceability
grounding answers in both technical and legal lifecycle signals, rather than relying solely on publication data
Constraints and implications
These fields improve analytical depth but should be interpreted with care:
Citation signals are directional, not definitive indicators of value or validity
Legal event interpretation remains context-dependent and does not replace formal legal assessment
The feature therefore strengthens contextual understanding, rather than acting as a standalone legal evaluation tool.
-
Example
An analyst assessing a competitor portfolio can now go beyond identifying relevant patents to:check which assets appear in force versus inactive or pending, indicating strategic commitment
highlight who cites the portfolio, checking technological relevance
Better Conversation Management with Rename, Delete, and Retention Transparency
What has changed?
Users can now rename and delete auto-saved conversations. Protégé also introduces a 90 day retention policy for conversations, with prior notification before deletion.
Why it matters
This improves continuity for ongoing work while helping users keep their workspace organised. It also makes data retention behaviour clearer and more predictable.
Analytical behaviour
Users can now: a) reopen recent conversations with context restored, b) rename conversations to reflect the analytical topic, c) delete conversations that are no longer useful
Conversations older than 83 days are marked as Pending Deletion, and users are notified 7 days before automatic deletion.
Constraints and implications
Conversation history is retained for 90 days from creation date and then deleted automatically. Deleted conversations are not recoverable.
Example
An analyst renames a conversation from a generic timestamp to "Waymo’s patent portfolio", making it easier to return to the work later. Older exploratory threads can be deleted to keep the workspace focused.
Improved Patent Number Matching
What has changed?
Protégé now matches patent numbers more robustly, including incomplete or differently formatted variants, and aligns matching logic more closely with PatentSight+ where feasible.
Why it matters
Patent analysts often work with inconsistent patent number formats. Better matching reduces manual rechecking and improves continuity between Protégé and PatentSight+.
Analytical behaviour
Protégé now:
accepts more valid patent number variants
normalises input into a canonical matching form
attempts to resolve inputs to a unique PatentSight patent identifier
presents disambiguation when multiple plausible matches exist
provides clearer feedback when no plausible match is found
Constraints and implications
This release improves number matching, but does not add bulk import flows.
Example
An analyst enters a list of patent numbers without kind codes or with unusual separators. Protégé can still resolve the intended patent document, or guide the user to refine the input if ambiguity remains.
Better Intent Recognition for Simple Tasks
What has changed?
Protégé is now better at identifying simple requests and avoiding unnecessary complex analysis when a straightforward answer is sufficient.
Why it matters
Not every task needs a multi-step analysis plan. By responding more proportionately to simple requests, Protégé can answer faster and fail less often.
Analytical behaviour
For tasks where a direct number or simple factual response is sufficient, Protégé now aims to avoid over analysing, over explaining, or invoking more complex reasoning than necessary.
Constraints and implications
This optimisation is intended for clearly simple tasks. More interpretive or multi-dimensional questions still require broader reasoning and explanation.
Example
If an analyst asks for the number of patents in a narrowly defined set - for instance owner Nintendo - Protégé no longer overcomplicates the task with unnecessary thinking and contextualising, but returns the number.
Lower Task Failure Rates and Clearer Recovery Behaviour
What has changed?
Protégé now includes better failure detection, classification, user messaging, and observability to reduce task failures and make recovery clearer when completion is not possible.
Why it matters
Reliability is not only about completing successful tasks. It is also about being honest and useful when something goes wrong. This release aims to reduce dead ends while helping engineering teams monitor and improve failure patterns systematically.
Analytical behaviour
Protégé now:
detects and logs incomplete task outcomes more clearly
classifies failure types such as ambiguity, system error, unsupported capability, and safety restriction
presents clearer user facing explanations when tasks fail
suggests practical next steps such as retrying, narrowing scope, rephrasing, or switching workflow
Constraints and implications
This release improves handling and observability of failures. The focus is on reducing hard failures and making failure states more transparent and actionable.
Example
If a request is too ambiguous to complete safely, Protégé now explains what is unclear and suggests how to refine the input rather than guessing silently or stalling.
Parallel Conversations for Exploratory Analysis
What has changed?
A single user can now run multiple conversations in parallel in Protégé by opening multiple tabs in the same web browser. Responses stream independently into the corresponding browser tab.
Why it matters
Patent analysis is rarely linear. Analysts often want to pursue a follow up question or adjacent topic while another answer is still being generated. Parallel conversations reduce waiting time and better match real exploratory workflows.
Analytical behaviour
Submitting a prompt in one conversation no longer blocks work. Streaming responses are routed to their originating browser tab, and failures in one conversation do not interrupt others.
Constraints and implications
This release enables parallel work, but does not yet include parallel conversation in one browser tab, per conversation cancellation, live synchronisation across devices, or a broader redesign of conversation navigation.
Example
While Protégé is generating a response on Battery Materials, an analyst can open a new browser tab and start a second conversation by asking about Main competitors in solid state electrolytes without waiting for the first answer to finish.