Building an AI-assisted design delivery pipeline.
Turning research insight into governed, implementation-ready product direction.
McKesson had an opportunity to carry design intent more clearly from research to engineering. As design strategy and delivery lead, I built a governed AI-assisted pipeline inside Provider Solutions Technology that connected research synthesis, live AI prototyping, design-system patterns, Storybook validation, and engineering-aligned implementation into one continuous workflow.
This was not a tool adoption effort. It was an operating-model change: preserve product intent, explore more options earlier, and carry validated direction from research to engineering with human validation at every stage.
AI accelerates execution. Humans remain responsible for direction, validation, and approval.
Secured leadership support for a governed AI-assisted delivery pilot.
The work translated an AI/SDLC vision into a practical workflow that helps teams move from research insight to live prototype to engineering-aligned implementation while keeping human validation at each stage.
Senior Director and VP leadership allocated dedicated engineering capacity outside regular roadmap commitments.
Direction was chosen in working sessions rather than returned for revision — teams evaluated two or three live concepts instead of moving one approach through a multi-week review cycle.
Reduced the need for a 12-week design lead buffer ahead of engineering handoff. Live prototypes validate direction before engineering commits.
Figma Variables, component properties, and variants are mapped to production code through Code Connect and validated through Storybook.
Validated end to end in a sandbox pilot before applying the pipeline to Treatment Readiness and Demand Forecasting.
Six outcomes across delivery, velocity, and organizational readiness
Governed AI delivery path
Created a practical AI-assisted delivery workflow that could operate inside McKesson's enterprise governance model using approved tools and existing engineering practices.
Leadership investment
Secured Senior Director and VP support for a pilot with dedicated engineering capacity outside regular roadmap commitments.
Expanded solution space
AI-assisted concept generation expanded the solution space before teams committed to a direction. Teams could compare multiple approaches in working sessions instead of moving one concept at a time through review cycles.
Earlier validation
Moved workflow direction into live prototype review before engineering commitment, reducing the risk of discovering misalignment after build begins.
Reduced translation loss
Connected research, prototype direction, design-system patterns, and implementation planning so design intent is less dependent on handoff notes and interpretation.
Scalable practice model
Created a repeatable delivery model for applying AI-assisted prototyping, design systems, research infrastructure, and UX operations to broader modernization work.
A governed enterprise environment with an AI delivery mandate
Product leadership had introduced AI-assisted SDLC as a strategic direction for McKesson's Provider Solutions Technology division. The question was not whether to pursue it — it was how to implement it responsibly inside a governed enterprise environment with existing tooling, engineering workflows, and compliance requirements.
The existing delivery model relied on sequential handoffs: research completed, then wireframes, then design iterations, then stakeholder review, then engineering estimation, then build, then QA — each stage creating opportunities for intent to drift from the original research insight.
The Vision
Leadership defined the desired AI-assisted SDLC direction.
The Gap
The presentation showed where to start, but not how to implement it inside enterprise governance, tooling, and engineering constraints.
The Implementation Path
I translated the vision into a pilot-ready workflow using approved tools: Figma Make, Code Connect, Storybook, GitHub Copilot, and existing engineering practices.
Leadership defined the AI-assisted SDLC direction. My role was translating that direction into a governed delivery model the organization could actually pilot.
Leadership defined the AI-assisted SDLC direction. My role was translating that vision into a pilot-ready delivery model the organization could actually operate — using approved tools, inside enterprise governance, alongside existing engineering practices.
Helping research, design, and engineering carry intent through delivery.
We mapped the existing design-to-engineering workflow across research, design, tickets, QA, and implementation. Each stage introduced a translation layer where user needs, interaction behavior, states, edge cases, and decision rationale became harder to preserve. Research findings became screens. Screens became redlines and tickets. Tickets became code — and intent drifted at every step.
The opportunity was not to create more artifacts. It was to help artifacts carry the reasoning behind the decisions — the context engineers needed to build the right thing without reinterpreting intent at every step.
The decision was not whether to explore AI-assisted delivery — leadership had already set that direction. The strategic question was how to make it repeatable, governed, and usable inside the existing enterprise workflow.
Connect handoffs into one workflow
Research, design, and engineering were moving through separate phases with formal handoff points. The opportunity was to create a more continuous workflow with fewer interpretation gaps.
Move validation earlier
Stakeholder feedback often surfaced after engineering had already oriented around a direction. The opportunity was to validate workflow direction before build commitment.
Expand concept exploration
Manual wireframing and sequential review cycles meant teams typically evaluated one direction at a time. The opportunity was to compare multiple directions earlier, while the team still had room to choose.
Align design system patterns to code
Design system components existed in Figma, but they were not consistently mapped to repository code. The opportunity was to make design-system intent easier for engineering to inherit.
Carry research intent forward
User needs, workflow constraints, and decision rationale could become disconnected once research moved into static design artifacts. The opportunity was to keep that reasoning visible through delivery.
Give engineering clearer direction
Engineers needed more than a mockup. The opportunity was to provide validated workflows, component mappings, states, constraints, and decision logic before build began.
The pipeline reframed delivery from a sequence of handoffs into a connected workflow — reducing the steps between research insight and implementation-ready direction, and keeping design intent visible at every stage.
The before model shows where intent weakened through sequential handoff. The AI-assisted model shows how research synthesis, live prototyping, design-system mapping, Storybook validation, and GitHub-aligned delivery carry decisions forward with less reinterpretation.
Eight stages for carrying intent into implementation
The pipeline maps Design Thinking to the enterprise SDLC so research insight, prototype direction, design-system patterns, and implementation decisions move together instead of being reinterpreted at each handoff.
- Phase I — Human-Led Discovery
-
01
Research Synthesis Design Thinking: Empathize + Define
- Customer interviews, field observation, and behavioral analytics run continuously
- AI-assisted synthesis surfaces patterns across research sessions
- Research repository captures findings, themes, and decision rationale — available to the full team
- Human responsibility: problem definition and direction-setting remain with the designer
-
02
AI-Assisted Concept Generation Design Thinking: Ideate
- Figma Make and Replit generate multiple directional concepts faster than manual wireframing
- Teams evaluate 2–3 live concepts in a single working session instead of one concept per review cycle
- Expands the solution space before the team commits to a direction
- Human responsibility: designers evaluate, edit, and decide — AI accelerates execution
-
03
Live Prototype Validation Design Thinking: Prototype + Test
- Interactive prototypes validated in live sessions with customers and account managers
- Feedback moves from follow-up review cycles into real-time working sessions
- In early use, concept iteration moved from days to hours between feedback and revised direction
- Human responsibility: direction validated before engineering commitment
- Phase II — Design-System Integration
-
04
Design-System Mapping SDLC: Design + Plan
- Validated concepts mapped to Lynx design-system components via Figma Code Connect
- Figma Variables and Dev Mode make token intent visible in the engineering environment
- Components validated through Storybook become candidates for production-ready system components
- Outcome: design system moves closer to a shared source of truth across design and engineering
- Phase III — Engineering Delivery
-
05
Implementation-Ready Code Direction SDLC: Build Planning
- AI-assisted output explores component structure, interaction logic, and implementation direction
- Based on mapped design-system patterns — not generated from scratch
- Goal: give engineering clearer, component-aligned direction before build begins
- Not: autonomous production deployment — human review required before any code is used
-
06
GitHub Copilot Refactor SDLC: Build + Enforce
- GitHub Copilot supports refactoring and implementation alignment inside the repository
- Design-system rules, component usage, and code quality checks applied closer to production
- Keeps implementation aligned with design-system decisions as code evolves
-
07
Storybook Validation SDLC: Test
- Component-level review before product integration — issues caught in isolation cost less than issues found after assembly
- Every state and variant documented before release
- Components validated here feed back into the Figma design system — closing the loop
-
08
GitHub Release SDLC: Deploy
- Research insights that enter at Stage 1 reach production with less interpretation loss
- Engineers receive design-system-aligned implementation artifacts alongside Figma files
- Less reliance on handoff notes and interpretation — intent travels with the work
Keeping discovery connected to delivery.
The pipeline keeps discovery connected to delivery by carrying research intent, prototype direction, design-system decisions, and implementation planning through the same workflow — with Figma Code Connect mapping design components to production repository components at every step.
Proving the pipeline before applying it to complex modernization work
The pipeline was first tested in a sandbox pilot for adding patients to Lynx. The goal was to make sure the full workflow functioned end to end — and that it could operate inside McKesson's governed environment — before applying it to more complex modernization efforts.
Why we started here
Adding patients to Lynx was focused enough to test the complete pipeline without the full complexity of Treatment Readiness. It still involved real product patterns, design-system decisions, component mapping, and engineering collaboration — making it a useful proof point rather than a synthetic exercise.
Live AI prototyping
Figma Make prototypes were validated before engineering commitment — moving iteration earlier in the delivery cycle rather than correcting direction after build began.
Design-system mapping
Component patterns, states, and interaction logic were mapped through Figma Code Connect before implementation, establishing the design-to-code relationship ahead of engineering work.
Storybook validation
Components were reviewed and documented in isolation before product integration — validating design-system decisions at the component level rather than after assembly.
Engineering-aligned delivery
GitHub Copilot-supported implementation inside the repository confirmed the pipeline could operate within McKesson's governed engineering environment from end to end.
A more connected path from concept to implementation
Governed workflow
Created an AI-assisted delivery workflow using approved enterprise tools rather than unapproved external tooling — making the pipeline adoptable inside McKesson's compliance model.
Leadership and engineering support
Secured Senior Director and VP support for a pilot with dedicated engineering capacity outside the existing roadmap — the organizational signal that the approach was worth investing in.
Earlier validation with more options
Moved iteration earlier by using live AI prototypes during discovery and stakeholder review — shifting feedback from follow-up cycles into working sessions where direction could be evaluated and adjusted in real time.
Reduced interpretation gaps
Connecting prototype direction to design-system patterns and engineering-aligned implementation artifacts reduced the distance between design intent and what engineering received.
Repeatable workflow
Established a reusable process for testing design-to-code practices on active Lynx modernization work — not a one-time exercise but a scalable delivery model.
Research infrastructure
Built the research repository, participant database, and customer issues tracker that support ongoing discovery — infrastructure that now connects research findings to pipeline inputs systematically.
The sandbox pilot is complete. The pipeline is governed, validated, and ready to scale.
- Pilot surface: Adding Patients to Lynx — end-to-end validation in a governed environment
- Next: Treatment Readiness and Demand Forecasting — higher complexity, multiple roles, exception recovery
- Foundation established for applying the pipeline across broader Lynx modernization work
Building the infrastructure around the work
The pipeline is one part of a broader operational model built for the PST UX practice — the systems and processes that allow the team to operate consistently and scale the work across the division.
Team lead
- Lead UX team at McKesson PST — design direction, process, and output
- Onboarded a second UX designer and established shared working model
- Training sessions, working agreements, and UX Kanban board in Jira for cross-team visibility
Research infrastructure
- Built research repository from scratch — structured participant database by role and practice type
- Enables systematic recruitment across 360+ oncology and specialty practices
- Customer issues tracker surfaces patterns that inform roadmap decisions and design priorities
Delivery integration
- UX story writing connects design decisions to engineering-ready work items in Jira
- Design work enters delivery as structured, trackable artifacts — not handoff notes
- Closes the gap between discovery and implementation
Design system governance
- Workflow for moving validated patterns into reusable components and Storybook documentation
- Design system evolves through real product work, not separate documentation efforts
- Implementation-aligned guidance available to engineering at each release
How the work shaped delivery beyond the pilot
Delivery model
Established a governed AI-assisted delivery model that can be applied across Lynx modernization work — not a one-project tool but a reusable operating approach.
Design system foundation
Component patterns, Storybook validation, and Code Connect mappings are building a shared implementation layer that reduces interpretation work at every future handoff.
Research-to-delivery connection
The pipeline establishes a direct path from research findings to prototype direction to implementation planning — reducing the distance between what research reveals and what engineering builds.
Broader division readiness
Leadership investment and dedicated engineering capacity signal organizational readiness to scale the model. The sandbox pilot is the foundation for that expansion.
Related work — Treatment Readiness
The pipeline was built alongside the Treatment Readiness modernization effort — the two case studies show end-to-end systems thinking: one modernizing the product experience, the other modernizing the way product experiences get delivered. View Treatment Readiness →
What This Changed
The previous model required design to run a full release cycle ahead of engineering — a lead buffer of up to 12 weeks before handoff. The pipeline brings research, prototyping, design-system validation, and implementation planning into one connected workflow, designed to reduce that buffer by 6–8 weeks per release cycle.
The impact went beyond efficiency. The pipeline shifts the designer's role from producing documentation after decisions are made to driving direction and intent through every stage of delivery. AI accelerates execution. Humans remain responsible for direction, validation, and approval — at every stage.