Healthcare & Life Sciences
2026
AI-Powered Clinical Authoring Platform
Designing a dual-platform AI system that helps clinical teams author regulatory documents faster - reducing a 10–15 year drug development bottleneck.
The Problem
A 10–15 year journey to bring medicine to patients
Bringing a new drug to market takes an average of 10–15 years (PhRMA) and costs approximately $2.6 billion (Tufts CSDD) — yet only about 12% of drugs entering clinical trials ever reach approval (FDA). A significant share of that timeline is consumed by clinical documentation: protocols, clinical study reports, informed consent forms, and regulatory submissions that must be authored, reviewed, cross-referenced against regulatory standards, and coordinated across multiple specialized roles.
Clinical authors work primarily in Microsoft Word, drafting documents manually over months or years, cross-referencing regulatory guidelines from agencies like the FDA and EMA, and coordinating reviews with statisticians, medical directors, and regulatory affairs teams. The process is fragmented across disconnected tools, prone to human error in data transcription, and bottlenecked by sequential handoffs between roles.
The design challenge: How might we integrate AI-powered assistance into the clinical authoring workflow — without disrupting the tools and habits authors already rely on — to help Life Sciences organizations author faster, catch errors earlier, and ultimately shorten the time it takes for medicine to reach patients?
Key Findings
Understanding the ecosystem before designing for it
01
A 7-role ecosystem, not a single-user tool
Mapping the pipeline revealed seven interconnected roles — Authors, Template Admins, Content Admins, Reviewers, Workflow Managers, System Admins, and Auditors — plus an AI layer. Each has distinct goals, but their workflows are deeply interdependent.
01
Competitors miss the mark on author workflows
Competitive analysis of Veeva Vault, Intellinotion, and Workiva revealed a gap: no platform deeply integrates AI authoring within the tool authors already use. This reinforced the strategy for a Word-native solution.
01
Three primary personas, three distinct workflows
Clinical Authors need AI in their writing environment. Template Admins need to codify knowledge into reusable templates. Content Admins need orchestration tools for document packages and team coordination.
Approach
From concept to dual-platform product
1
Platform Strategy & the Pivot to Word
Early concepts explored building a full web-based authoring platform with AI chat, document editing, collaboration, and review capabilities (including real-time co-editing, commenting, version history, and TLF data table insertion). Through iteration and stakeholder feedback, a critical insight emerged: clinical authors have spent years — sometimes decades — working in Microsoft Word. Rather than asking them to adopt an entirely new platform, we pivoted to a dual-platform architecture:
Web app → The orchestration layer: package management, template configuration, document assignment, source file review, lifecycle tracking, and team collaboration.
Word plugin → The authoring layer: AI-powered content generation, document evaluation, compliance checking, and source file management — embedded directly in the author's existing workspace.
Old Version
New Version
3
Designing for Three Journeys
The dual-platform required three distinct but interconnected user journeys: Template Admins configuring AI-powered smart templates, Authoring Managers orchestrating document packages, and Clinical Authors generating and refining content — all connected through shared source files, parameters, and evaluation workflows.
4
Usability Testing with Healthcare SMEs
Conducted structured usability testing with 4 task-based scenarios covering the core authoring workflow — from first impressions of the ribbon interface through content modification, source management, and prompt interaction.
Insights
Placeholder affordance
Users initially tried editing directly in Word — led to stronger visual cues and onboarding nudges on placeholders
@ commands: powerful but hidden
Once discovered, users loved the pattern (compared to Slack/Notion) — added more prominent hint text.
Source management was intuitive
Expected vs. additional distinction worked well. Auto-fetch from regulatory systems reduced friction.
Evaluation scoring resonated
Completeness, Relevance, Accuracy, Fluency dimensions mapped to clinical users' existing mental models.
5
Design System & Scalable Delivery
Built a comprehensive design system used across both platforms. Later encoded into an AI-powered design skill that reduced delivery timelines by ~75% — from ~12 weeks to ~3 weeks per new use case.
Designs
Three journeys, one connected platform
Template Admin
Authoring Manager
Author
Impact
Measurable outcomes
~75%
Faster delivery — design system +AI skill cut timelines from ~12weeks to ~3 weeks
8+
Clinical R&D use cases shipped to
Life Sciences & Healthcare clients
Reusable
Designs became foundations for
client proposals and RFP
responses
What began as an exploration of AI-assisted clinical authoring evolved into a comprehensive dual-platform product for a complex 7-role ecosystem. The pivot from a standalone web app to a Word-native plugin — driven by user research and validated through structured usability testing — embodied a core design principle: meet users where they already work, and augment their expertise with AI rather than replacing their tools.
This project has been white-labeled to protect client confidentiality. Brand names, data values, and visual identity have been modified. The design methodology and UX decisions shown are my original work.