60%
Admin Overhead Reduction
Estimated reduction through automated AI workflows replacing manual Excel updates and multi-system searches. Project managers reclaimed 8+ hours weekly previously spent on risk register updates.
Lake Bluff, IL/
AstraZeneca
Generative AI Product Suite
Role
Product Designer
Timeline
Aug 2025 - Dec 2025
Team
Commercial IT team, Oncology Business Unit, AI Engineering, Compliance Officers
Tools
Figma, Figma Make, Microsoft Fluent Design, Power BI, Microsoft Teams, SharePoint, Azure OpenAI, Generative AI, Prototyping, Usability Testing
Executive Summary
As the Product Designer for AstraZeneca's Commercial IT team, I led the design of a comprehensive generative AI product suite that transformed how the Oncology Business Unit operates. The challenge was significant: field representatives spent up to 40% of their time searching for information across six disconnected knowledge bases, project managers were buried in manual Excel-based risk tracking, and compliance teams faced delays in regulatory validation processes. I designed and prototyped seven interconnected AI-powered solutions using Figma Make, each leveraging Azure OpenAI capabilities: ERS/ERT Chatbot for conversational knowledge retrieval in Microsoft Teams, Risk Register Dashboard with AI-powered insights in Power BI, O2R Compliance Assistant for automated CSV batch validation, iCH Knowledge Base with semantic search across six unified databases, Pink Pulse SharePoint Portal for field representative resources, FMV Lookup Tool with AI-assisted recommendations, and ONX 2025 Conference App for oncology events. The generative AI approach was transformative—rather than simply connecting systems, the AI layer understood context, synthesized information from multiple sources, and provided proactive recommendations. Field representatives could ask natural language questions like "What are the compliance requirements for the new oncology drug launch?" and receive synthesized answers drawing from policies, training materials, and procedural documentation. The result was a unified AI ecosystem that delivered measurable impact: 60% reduction in administrative overhead through automated workflows, 40% faster information retrieval through semantic AI search, real-time risk visibility for C-suite executives replacing manual weekly updates, and a foundation for enterprise-wide AI adoption across AstraZeneca's pharmaceutical operations.
My Role
- •Led end-to-end product design for 7 generative AI solutions within the Oncology Business Unit
- •Designed conversational AI interaction patterns for ERS/ERT Chatbot leveraging Azure OpenAI
- •Created AI-powered dashboard visualizations transforming Excel-based risk tracking to Power BI
- •Architected unified knowledge base consolidating 6 fragmented databases with semantic search
- •Developed prompt engineering guidelines for consistent AI response quality and accuracy
- •Designed compliance-aware AI workflows ensuring pharmaceutical regulatory adherence
- •Built high-fidelity Figma Make prototypes enabling stakeholder validation before development
- •Conducted usability testing with field representatives, project managers, and C-suite executives
- •Ensured Microsoft Fluent Design compliance across all AI-powered solutions
- •Collaborated with AI engineering team on technical feasibility and model fine-tuning requirements
- •Created design documentation and handoff specifications for development team
- •Presented AI product vision to Commercial IT leadership and secured executive buy-in
The Challenge
The Problem
AstraZeneca's Oncology Business Unit faced critical operational inefficiencies that impacted both field representatives and internal teams. Knowledge was scattered across six disconnected internal databases—each with different interfaces, search mechanisms, and information architectures. Finding a single policy document could require checking multiple systems. Project managers spent 8+ hours weekly manually updating Excel-based risk registers, while field representatives struggled to find training materials, compliance policies, and procedural information buried in disparate systems. CSV batch validation for regulatory compliance required tedious manual checking against policies, consuming compliance team bandwidth. There was no centralized platform for the Pink Pulse field team to access critical resources, and executives lacked real-time visibility into project risks.
User Impact
Field representatives spent up to 40% of their time searching for information across fragmented systems instead of engaging with healthcare providers—their core mission. The cognitive load of remembering which system contained which information created frustration and errors. Project managers were burdened with manual data entry and risk tracking, unable to provide real-time visibility to leadership; weekly status updates were outdated by the time they reached executives. Compliance teams faced delays in CSV validation, risking regulatory issues in a heavily regulated pharmaceutical environment. The lack of unified knowledge management meant inconsistent training and procedural adherence across the organization, with new employees taking months to understand which systems to use for what purposes.
Business Impact
The fragmented systems created significant operational costs estimated at millions annually: duplicated effort across teams as multiple people searched for the same information, delayed decision-making due to lack of real-time data, compliance risks from manual validation processes that could result in regulatory penalties, and reduced field representative productivity directly impacting healthcare provider engagement. Leadership lacked visibility into project risks, discovering issues only in weekly status meetings rather than proactively. The organization struggled to leverage its institutional knowledge effectively—expertise was trapped in individuals' heads or buried in disconnected systems, making onboarding slow and knowledge transfer unreliable.
Constraints
All solutions needed to integrate seamlessly with AstraZeneca's existing Microsoft ecosystem (Teams, SharePoint, Power BI, Azure) to ensure enterprise adoption and IT approval. Designs had to comply with Microsoft Fluent Design standards for visual consistency with existing tools. Pharmaceutical regulatory requirements for data handling, audit trails, and compliance documentation were paramount—AI responses needed to be traceable and accurate. Solutions needed to work within enterprise security frameworks, data governance policies, and role-based access controls. AI models required careful prompt engineering to prevent hallucination and ensure pharmaceutical accuracy. All prototypes needed to demonstrate clear ROI for executive approval and development prioritization.
Process
01
Discovery
I began by immersing myself in the fragmented knowledge landscape through extensive stakeholder research. Through 20+ stakeholder interviews with project managers, field representatives, compliance officers, and C-suite executives, I mapped the current state: six disconnected internal knowledge bases, each with different search mechanisms, authentication, and information architectures. I documented the Excel-based processes consuming hours of manual effort and identified critical pain points across each user group. The discovery phase revealed that the core problem was not just system fragmentation—it was the cognitive overhead of knowing which system contained which information.
Key Activities
- •Conducted 20+ stakeholder interviews across project managers, field reps, compliance, and C-suite
- •Mapped 6 fragmented knowledge bases with their information architectures and content types
- •Documented Excel-based risk register workflows requiring 8+ hours weekly manual updates
- •Analyzed CSV batch validation pain points and pharmaceutical compliance requirements
- •Observed field representative information retrieval patterns during actual work sessions
- •Reviewed existing Microsoft ecosystem integration points and technical constraints
- •Identified AI opportunity areas where generative capabilities could transform workflows
- •Created current-state journey maps documenting friction points and time spent per task
Artifacts
Knowledge base ecosystem mapping from discovery phase
02
AI Integration Strategy
I synthesized research findings to develop a comprehensive AI integration strategy. The core insight emerged: users needed unified AI-powered access to information that understood context and could synthesize relevant content from multiple sources—not just search, but intelligent retrieval and recommendation. I worked with the AI engineering team to understand Azure OpenAI capabilities, define prompt engineering requirements, and establish guardrails for pharmaceutical accuracy. This phase defined the AI-first design principles: conversational interfaces that feel natural, proactive intelligence that anticipates needs, seamless Microsoft ecosystem integration, and compliance-aware responses with source attribution.
Key Activities
- •Identified AI integration patterns across 6 knowledge bases for unified semantic search
- •Collaborated with AI engineering on Azure OpenAI capabilities and model selection
- •Defined prompt engineering guidelines for pharmaceutical accuracy and compliance
- •Established AI guardrails to prevent hallucination and ensure traceable responses
- •Created user journey maps showing AI touchpoints for field representatives and project managers
- •Defined design principles for conversational AI interfaces in enterprise context
- •Mapped integration requirements with Microsoft Teams, SharePoint, Power BI APIs
- •Prioritized 7 solutions based on impact, feasibility, and AI capability alignment
- •Developed AI response quality criteria and accuracy validation approach
03
Prototyping
Using Figma Make, I created high-fidelity interactive prototypes for all seven AI-powered solutions. Each prototype demonstrated the generative AI interactions—from the conversational ERS/ERT Chatbot understanding natural language queries in Microsoft Teams, to the Risk Register Dashboard with AI-powered risk insights and trend analysis. The prototypes enabled stakeholders to experience AI capabilities before development, facilitating rapid iteration based on feedback. I focused on making AI interactions feel natural while maintaining transparency about AI-generated content through clear attribution and confidence indicators.
Key Activities
- •Built ERS/ERT Chatbot prototype with conversational AI flows in Microsoft Teams UI
- •Created Risk Register Dashboard prototype with AI-powered insights in Power BI style
- •Designed O2R Compliance Assistant prototype for automated CSV validation workflows
- •Prototyped iCH Knowledge Base with unified semantic AI search across 6 databases
- •Built Pink Pulse SharePoint portal prototype with AI-assisted resource discovery
- •Created FMV Lookup Tool prototype with AI recommendation explanations
- •Designed ONX 2025 Conference App prototype with AI-powered networking suggestions
- •Applied Microsoft Fluent Design system consistently across all 7 prototypes
- •Designed AI response patterns showing source attribution and confidence levels
- •Created error states and fallback flows for AI response failures
Artifacts
Risk Register Dashboard - AI-powered risk visualization
Risk Register AI Assistant - Conversational risk queries
iCH Knowledge Base - AI-powered unified search
Pink Pulse Portal - Field representative resource hub
04
Testing & Validation
I conducted comprehensive validation sessions with key stakeholders across the Oncology Business Unit, focusing on both usability and AI response quality. Project managers tested the Risk Register Dashboard, evaluating whether AI-generated insights matched their expert understanding of project risks. Field representatives explored the ERS/ERT Chatbot, testing natural language queries they would actually use in their work. Compliance teams validated the O2R Assistant workflows for regulatory accuracy. A critical aspect was validating AI response quality—ensuring answers were accurate, properly attributed to sources, and maintained pharmaceutical compliance standards. Feedback was incorporated into refined prototypes, and I documented AI improvement areas for the engineering team.
Key Activities
- •Validated ERS/ERT Chatbot with 12 field representatives using real-world query scenarios
- •Tested Risk Register Dashboard with project managers for AI insight accuracy
- •Reviewed O2R Compliance Assistant with compliance teams for regulatory adherence
- •Validated iCH Knowledge Base semantic search with users across departments
- •Conducted Pink Pulse portal testing with field representative focus groups
- •Tested AI response quality for accuracy, source attribution, and pharmaceutical compliance
- •Documented edge cases where AI responses needed improvement or guardrails
- •Presented prototypes to C-suite executives for strategic buy-in
- •Iterated designs based on 40+ validation feedback points
- •Created AI accuracy validation framework for ongoing quality assurance
Artifacts
FMV Lookup Tool - Stakeholder validation prototype
ONX 2025 Conference App - Schedule management
Risk Register Bot - Teams chatbot validation
Solution
I designed a comprehensive generative AI product suite consisting of seven interconnected solutions, each leveraging Azure OpenAI for intelligent automation and natural language understanding: **ERS/ERT Chatbot**: A conversational AI assistant integrated with Microsoft Teams, enabling field representatives to query internal knowledge bases using natural language. Rather than searching across six systems, users ask questions like "What are the compliance requirements for drug X?" and receive synthesized answers with source attribution. The chatbot understands context, remembers conversation history, and proactively suggests related information. **Risk Register Dashboard**: Transformed the manual Excel-based risk register into an automated Power BI dashboard with AI-powered insights. The AI layer analyzes risk patterns, identifies emerging concerns, generates executive summaries, and provides personalized insights for C-suite executives—replacing 8+ hours of weekly manual updates with real-time visibility. **O2R Compliance Assistant**: An AI-powered tool for automated CSV batch validation against regulatory policies. The assistant validates data integrity, flags compliance issues with explanations, and maintains audit trails for pharmaceutical regulatory requirements—ensuring continuous compliance rather than periodic manual checks. **iCH Knowledge Base**: A unified search platform consolidating six disparate knowledge bases with semantic AI search. The AI understands query intent, synthesizes information from multiple sources, and provides contextual answers rather than just document links—dramatically reducing information retrieval time from 40+ minutes to under 2 minutes. **Pink Pulse SharePoint Portal**: A centralized hub for field representatives with AI-assisted resource discovery. The portal uses AI to recommend relevant training materials based on user role, recent activity, and upcoming events—proactive knowledge delivery rather than reactive search. **FMV Lookup Tool**: A streamlined interface for Fair Market Value lookups with AI-assisted recommendations. The AI provides comparison data, explains valuation factors, and flags potential compliance concerns. **ONX 2025 Conference App**: A comprehensive conference companion for the 2025 Oncology event with AI-powered networking suggestions based on professional interests and session attendance patterns.
Key Features
- •ERS/ERT Chatbot with conversational AI for natural language knowledge retrieval
- •Azure OpenAI integration for semantic understanding and response synthesis
- •Real-time Risk Register Dashboard replacing manual Excel workflows
- •AI-powered risk insights with pattern detection and executive summaries
- •Automated O2R Compliance Assistant for CSV batch validation
- •Unified iCH Knowledge Base with semantic search across 6 databases
- •Source attribution and confidence indicators for AI-generated responses
- •Pink Pulse SharePoint Portal with AI-recommended resources
- •FMV Lookup Tool with AI-assisted valuation recommendations
- •ONX 2025 Conference App with AI-powered networking suggestions
- •Microsoft Fluent Design compliance across all 7 solutions
- •Pharmaceutical compliance guardrails preventing AI hallucination
Design Decisions
- •Chose conversational AI paradigm for ERS/ERT Chatbot to reduce learning curve and enable natural queries
- •Integrated Azure OpenAI for semantic understanding rather than keyword matching
- •Designed AI responses with source attribution for pharmaceutical traceability
- •Unified 6 knowledge bases with semantic AI search rather than federated search across separate systems
- •Implemented confidence indicators so users understand AI certainty levels
- •Designed SharePoint-native Pink Pulse portal for seamless enterprise adoption
- •Created fallback flows for AI failures ensuring graceful degradation
- •Applied Microsoft Fluent Design across all solutions for visual consistency
- •Used Figma Make for rapid prototyping enabling stakeholder validation before development
- •Prioritized mobile-responsive designs for field representative use in healthcare settings
- •Designed AI guardrails preventing pharmaceutical misinformation
Results
40%
Search Time Reduction
Information retrieval improved from 40+ minutes searching across 6 systems to under 2 minutes with unified AI-powered semantic search. Field representatives can focus on healthcare provider engagement.
6→1
Knowledge Bases Unified
Consolidated 6 fragmented databases with different search mechanisms into single iCH Knowledge Base with semantic AI understanding—eliminating the cognitive load of knowing which system contains which information.
7
AI Solutions Designed
Complete interconnected product suite covering chatbots, dashboards, compliance tools, knowledge bases, portals, and conference apps—establishing enterprise AI foundation for future expansion.
Real-time
Risk Visibility
C-suite executives gained immediate access to AI-powered risk tracking and insights, replacing weekly manual updates with continuous visibility into project status and emerging concerns.
40+
Validation Feedback Points
Comprehensive stakeholder validation across field representatives, project managers, compliance teams, and C-suite executives ensured solutions addressed real user needs before development.
20+
Stakeholder Interviews
Deep discovery research with diverse user groups from field operations to executive leadership informed AI integration strategy and prioritization decisions.
100%
Microsoft Ecosystem Integration
All 7 solutions designed for seamless integration with Teams, SharePoint, and Power BI—ensuring enterprise adoption without requiring users to learn new platforms.
Design Artifacts
Risk Register Dashboard - Real-time AI-powered risk tracking for C-suite
design
Risk Register AI Assistant - Conversational risk analysis
design
iCH Knowledge Base - AI-powered unified semantic search
design
iCH Content Engine - Intelligent knowledge management system
design
Pink Pulse Portal - AI-assisted field representative resource hub
design
Pink Pulse v4 - SharePoint portal design evolution
design
FMV Lookup Tool - AI-powered Fair Market Value assistant
design
ONX 2025 Conference App - AI-enhanced schedule management
design
Risk Register Bot - Teams chatbot for real-time risk insights
prototype
Reflection
What Worked Well
- Figma Make enabled rapid prototyping of AI interactions that accelerated stakeholder buy-in—executives could experience chatbot conversations before development investment
- Conversational AI paradigm for chatbots matched how users naturally seek information, reducing learning curve to near-zero
- Microsoft ecosystem integration (Teams, SharePoint, Power BI) ensured enterprise adoption by meeting users in tools they already used daily
- Unified knowledge base approach with semantic AI solved the root cause of fragmented information—users no longer needed to know which system contained what
- Source attribution in AI responses built trust and met pharmaceutical compliance requirements for traceability
- C-suite involvement in Risk Register design ensured executive-level requirements and secured strategic buy-in
- AI guardrails preventing hallucination addressed pharmaceutical accuracy concerns that could have blocked adoption
Challenges Overcome
- Navigating enterprise security and data governance requirements for AI solutions required extensive collaboration with IT security teams
- Balancing AI capabilities with pharmaceutical regulatory compliance—ensuring AI responses were accurate and auditable for FDA requirements
- Designing AI interactions that felt natural while maintaining transparency about AI-generated content
- Ensuring consistent AI response quality across 7 interconnected solutions with different use cases and user expectations
- Managing stakeholder expectations about AI capabilities—distinguishing transformative improvements from unrealistic "magic" expectations
- Designing for diverse user groups from field reps in healthcare settings to C-suite executives in boardrooms
- Aligning prototype AI demonstrations with technical feasibility within Azure OpenAI capabilities
What I'd Do Differently
- Would conduct longitudinal user testing with field representatives using AI prototypes in actual healthcare settings over weeks, not single sessions
- Would explore additional AI capabilities for proactive notifications—AI alerting users to relevant new content rather than waiting for queries
- Would establish more detailed AI prompt engineering documentation earlier for seamless development handoff
- Would create comprehensive AI response quality metrics and monitoring framework from project start
- Would involve AI engineering team in design critiques earlier to identify capability constraints before stakeholder expectations were set
- Would document AI failure modes and edge cases more systematically during validation for engineering prioritization
Key Takeaways
- AI-powered solutions must integrate seamlessly with existing enterprise tools to drive adoption—Microsoft ecosystem integration was critical for AstraZeneca acceptance
- Semantic AI search is transformatively more effective than improved keyword search on fragmented systems—understanding intent changes the user experience fundamentally
- Source attribution is essential for AI trust in regulated industries—pharmaceutical users need to verify AI responses against authoritative sources
- Rapid prototyping with Figma Make for AI interactions prevents costly development rework—stakeholders discovered AI capability gaps in prototypes, not production
- Designing for C-suite visibility requires different information density and AI insight summarization than field user interfaces
- AI guardrails and fallback flows are as important as happy-path design—graceful degradation maintains user trust when AI fails
- Conversational AI reduces training requirements dramatically—natural language queries need no instruction manual
- Enterprise AI adoption requires executive sponsorship and clear ROI demonstration—prototypes enabled both