Meta AI Raybans. UX Evaluation on Workplace Productivity.

Industry

Digital Manufacturing

Company

Bayer AG

Timeline

2 Months

Keywords

UX Research

Status

Shipped

Problem Statement

How might we

Methodology

The project kicked off with user interviews of internal Bayer researchers who were already using an early release of the PRINCE chatbot. Users ranged from toxicologists, microbiologists, R&D innovation directors, and machine learning research scientists. The outcome of these user interviews were opportunities that helped to identify persona, user goals, use cases, and challenges users faced when interacting with PRINCE. Questions explored included the following:

What do you consider important when making decisions?
What are your main goals when using the PRINCE chatbot?
Are there any existing features you don’t use?
What do you do with the information you receive?
Which types of questions does PRINCE assist you on?

User interviews were conducted with 6 participants internal to the company.

UX Interviews

Affinity Mapping

After uncovering key insights into how our users interact with AI systems, how they source and gather information, and where current road blocks exist in their workflows, I translated the user findings into a affinity map. I identified actionable use cases for features we could build and iterate into our next product sprint.

Affinity mapping board identify five core clusters

Note: Design assets have been modified per NDA requirements

Writing Assistant

How can we envision a writing assistant with 'human in the loop?'

During user interviews, many PRINCE users noted that they needed help crafting good prompts leading to their desired goals, or just weren't sure where to start depending on their level of using Generative AI tools. As a solution, we explored a writing assistant feature offering users suggestions on prompting, and exploring a way to build a 'human-in-the-loop' interaction experience. The purpose of the writing assistant was to help users at various stages if their research process — whether that involves brainstorming, expanding on a topic, crafting better drafts for papers, or forming follow up questions.

How might we envision a writing assistant to collaborate with users in their research and discovery process?

Conversation Starters

Since we found that new users were unsure how to begin interacting with the AI chatbot, we introduced “conversation starters," a predefined, context-aware example of prompts to suggest useful queries and highlight the system’s capabilities. This improved onboarding and user confidence, reducing hesitation and helping researchers explore the platform more effectively.

What if we add conversational starters and topics modes to help onboard users in?

What if we add conversational starters and topics modes to help onboard users in?

Sources

Researchers needed transparency to trust AI-generated responses and validate the origin of retrieved information.I designed a “sources” component that clearly lists and links back to the original PRINCE reports for every chatbot answer. Strengthened trust, accountability, and compliance by making each AI response fully traceable to its verified data source.

How might we add in a way for users to access the sources of their search requests?

Medial Logs

What about a feature showing the step by step process behind their search results?

New Chat

Recents

Give me a literature review on studie...

Data Table for Study A10z7 with bio...

Find me the biomarkers from Study...

Data Table for Study A88b3 from 20...

Written Report Summary for Study T8

Please draft me a ordered list of all...

Comprehensive literature review on Entrectinib inhibitors

Experiment

Give me a literature review on studies from the past 5 years about Entrectinib inhibitors for tumor-agnostic therapies

Sources

[12]

| More

PRINCE Chatbot steps

Status

Recovering files

50 potential files sourced in internal database

Keyword match

Filtering studies based on keyword match analysis

Sources: 21 found

21 matches filtered based on search and analysis

Tools: Extract data

Extracting data based on findings and metrics

Generate report

Compile all findings into final file for hand off

TRK Inhibitors...

Tissue-agnostic Anti-Cancer Drug...

TRK Inhibitors...

Zebrafish models in Toxicology an...

Pharmaceutic...

Tissue-Agnostic Targeting of NTR...

Pharmaceutic...

NCT02568267 Study of Entrecti...

Journal for Pr...

Integrated Analysis of Efficacy and S...

PRINCE Chatbot

Mechanism and Approval

Entrectinib is an orally available inhibitor targeting TRKA/B/C, ROS1, and ALK, notable for its CNS penetration capability which is critical for treating brain metastases. It has been FDA approved for adult and pediatric patients with solid tumors harboring NTRK gene fusions, based on results from integrated analyses of several clinical trials including ALKA-372-001, STARTRK-1, and STARTRK-2. These trials collectively demonstrated entrectinib's efficacy across different tumor types without regard to tissue origin, marking it as a key example of tumor-agnostic ("basket trial") drug development.

Clinical Efficacy

Entrectinib showed an objective response rate (ORR) of approximately 57% with some complete responses noted across diverse cancers with NTRK fusions. Its intracranial activity was also documented with responses observed in patients with brain metastases. The median progression-free survival (PFS) ranged around 11 months in patients treated with entrectinib. Indirect comparative analyses suggest larotrectinib may have a longer overall survival and duration of response, but entrectinib maintains a comparable overall safety profile.