Welcome to the second article in our series, “Conversations with the Team That Built LexisNexis® PatentSight+™ with Protégé™”. In the previous article, you saw how we are building a trustworthy agentic AI system, and in this blog post, you’ll learn how our team is bringing greater clarity to AI-driven patent analytics by incorporating a transparent thought process and a structured rationale into AI responses.
Trust in AI-driven patent analytics depends on more than useful outputs. Through our research and interactions with IP and business leaders around the world, we have learned that users need to understand what is happening in the background, what data is being used, and how the analysis is unfolding.
IP professionals deliver critical insights that support million-dollar decisions. That’s why validating the sources, assumptions, and methodology behind a conclusion is already part of responsible decision-making. The role of transparency in agentic AI systems like LexisNexis® PatentSight+™ with Protégé is to facilitate this validation. AI-supported analysis should clearly show the data behind its outputs. It should also explain its rationale and limitations. This helps users assess the results. It also helps them explain the results to stakeholders and apply them with confidence.
The graphic below shows how we built features that address general user concerns through each step of their analytics journey:
The features that make an AI system dependable are not always fully visible to the user. Patent analysts can verify outputs by checking whether the search syntax is valid, executable in LexisNexis® PatentSight+™, and consistent with the data shown in the system’s visualizations. But these checks happen after the analysis is complete, and they only build part of the trust. You also build trust by understanding the path the AI system followed to reach the results presented to you.
Because Protégé in PatentSight+ delivers answers to complex, critical business questions, users need visibility into the smaller decisions that shape the final output. For this reason, it breaks down these questions into many smaller actions, like:
One early tester said that “the LexisNexis PatentSight+ data source and also… the explainability” are what distinguish their experience with Protégé. Another user put it even more simply: “I also know the limitations and advantages of this database.” These reactions speak to the same point. In patent analytics, trust grows when users understand both the source of the analysis and the boundaries of the underlying data.
Early Protégé user on data and transparency A highlight from our experience is the PatentSight+ data source and also the explainability of the results. I also know the limitations and advantages of this database.
Early Protégé user on data and transparency
A highlight from our experience is the PatentSight+ data source and also the explainability of the results. I also know the limitations and advantages of this database.
When you ask Protégé a question, you will first see a reasoning summary that updates as the analysis progresses. This gives you visibility into the progress of the work. Contextual widgets show the plan, actions, and methodology in separate sections, so you can stay focused on the insights. Together, these elements give you more information, flexibility, and clarity about how the analysis develops.
Early access users described their experience in direct terms: “It very clearly explains to you every step of the way.” That is an important part of how transparency works in practice. The value comes not only from receiving an answer, but also from understanding how the system arrived at it.
Watch the video below to see how Protégé presents its rationale and analysis plan as work progresses.
Transparency also depends on making sources easier to inspect. The system has access to approved external web sources for additional, up-to-date information about public events. It cites these sources in its analysis and provides clickable hyperlinks. This helps users distinguish between patent data used in the analysis and public information brought in for context.
Figure: Shows a Protégé response with clickable hyperlinks to the original sources of information that the system referred to respond to your question.
That distinction is straightforward in principle. As Jakub Hudak, Machine Learning Manager working on Protégé in PatentSight+, put it, “Any patent data, be it charts of aggregated patent metadata or summaries of claims from individual patent families, comes from us, from PatentSight+. And anything that is not specifically from our patent data is rendered as clickable hyperlinks that the user can access.” This matters because users need to know what is derived from curated patent data and what is derived from other sources.
Users who had early access to Protégé said, “I like that I actually have links… instead of having to trust blindly that the AI is not making things up.”
Early Protégé user on transparency I like that I actually have links that I can check myself, instead of having to trust blindly that the AI is not making things up.
Early Protégé user on transparency
I like that I actually have links that I can check myself, instead of having to trust blindly that the AI is not making things up.
Compared with more familiar digital technologies such as email, search engines, and digital commerce, AI systems in expert domains such as IP analytics and innovation management solutions still have work to do before users feel fully comfortable relying on them. As Jakub puts it, “Trust is built over time through reliability and the confidence that users attain when they work alongside AI and see that they can inspect its work and scrutinize every detail of it.”
That is also how early testers describe the experience. One said the system’s trustworthiness stems from the combination of the data source and the workflow’s explainability. Another said simply, “I can really rely on the data that I see there.” That kind of confidence comes from making the system easier to inspect, easier to validate, and clearer about how it works. Jakub and his team say that the journey does not end with launching Protégé to the global audience; it has only begun now. Their aim is to continually improve functionalities and features to retain and build on the trust users place in our analysis.
See how Protégé makes AI-driven patent analytics easier to inspect, validate, and explain. Talk to an expert for a full demo.
Your AI assistant for fast, decision-ready patent insights that drive strategic business decisions.
