I led the design process that helped Provation choose between two approaches: engage patients earlier or build a cancellation-risk model. The work turned that strategic uncertainty into a product direction the organization could pilot.
Product direction preview
Patient guidance creates the readiness signals nurses need.Product direction preview · Patient experience and second-phase nurse dashboard
My roleDesign strategy, research, facilitation, and product design lead
PartnersProduct, Engineering, Data Science, Operations, leadership
Pilot6 weeks · 200+ patients · Hershey Endoscopy
DecisionPatient engagement first, cancellation prediction later
01 · Challenge
The prioritized approach depended on data the organization did not yet have.
The roadmap entered the work prioritizing cancellation prediction. Patient engagement was the competing direction. Both promised scale, but research needed to determine whether the data, timing, and workflow required by either approach actually existed.
Approach 01
Patient engagement
Reach patients during preparation, answer questions, and prevent readiness failures before they become cancellations.
Promise: change the outcome earlier.orStarting priority · Approach 02
Cancellation-risk model
Identify high-risk patients so staff could focus outreach where it may matter most.
Research constraint: reliable readiness and cancellation signals were not available.Earlier cancellation-risk model · Prioritized before research
Before choosing what to build, I needed to determine where readiness actually failed, what signals existed, and which intervention could work inside a live clinical pilot.
02 · In-person workshop
WORKSHOP DESIGN + FACILITATION
Competing perspectives became a shared learning agenda.
I brought together 15 cross-functional stakeholders from medical content, legal, technology, product, and data science, with some in the room and others joining virtually.
The goal was to onboard a new contractor team that was eager to start development but did not yet know where to begin. By reviewing the hypothesized journey and overlaying the concepts, I facilitated breakout discussions that exposed the feasibility, desirability, and viability unknowns behind each direction.
Workshop frameworkAmbiguity in, learning agenda out
01
Review the journey
Where did the hypothesized patient journey suggest readiness and cancellation risk changed?
A shared starting point→02
Overlay concepts
Where would patient engagement and cancellation prediction intervene across that journey?
Concepts anchored in context→03
Interrogate unknowns
What did each concept assume about feasibility, desirability, and viability?
Breakout discussions I facilitated→04
Define experiments
Which technical and field-research questions had to be answered before development?
A prioritized experimentation plan
Outcome
A prioritized research and experimentation plan for deciding what the team could and should build.
Research track 01 · Technical experiments
Test what the data could support.
Inspect the data lake, confirm which data we could access, and determine whether cancellation prediction was technically feasible.
Research track 02 · Field + concept research
Test what ASCs actually needed.
Understand ambulatory surgery center (ASC) workflows, how teams assessed risk at each step, and how strongly each concept resonated.
03 · Research
Workshop questions became evidence from the real readiness workflow.
We conducted 12+ virtual and in-person interviews across nurses, schedulers, physicians, and patients. We traced how readiness moved from scheduling through procedure day, learned how teams identified risk, and gathered feedback on where each concept fit.
In parallel, technical experiments tested what data was actually accessible. Together, the two tracks turned workshop assumptions into evidence.
A manual system organized by dateOne ASC stored patient communication materials, including preparation instructions, in physical folders organized by procedure date. The system was extensive and manual.
12+ interviewsVirtual + in person
Workflow investigationRisk across the readiness journey
Concept feedbackResonance with ASC teams
Data experimentsAccess + prediction feasibility
Research evidenceWhat we learnedProduct implication
01
Patients experienced significant anxiety, often lost or forgot their preparation instructions, and sometimes learned the importance of preparation too late to prevent a cancellation.
Synthesized readiness observation
→
Anxiety and missing guidance put readiness at risk
Instructions alone were not enough when patients could lose them, forget key steps, or miss why preparation mattered until it was too late.
→
So what
Reinforce preparation with timely, reassuring guidance before risk becomes a cancellation.
02
Many ASCs designated a full-time nurse to call patients about preparation; the proposed experience could handle much of that routine outreach and focus nursing time on patients who needed judgment.
Synthesized ASC workflow observation
→
Routine outreach consumed clinical capacity
The opportunity was not simply to identify risk, but to reduce repetitive calls while surfacing the patients who needed help.
→
So what
Automate routine guidance and reserve nurse time for exceptions.
03
Readiness risk became visible through staff conversations, but the data lake did not contain the passive signals the proposed prediction model required.
Workflow interviews + data-access experiments
→
The needed prediction data did not yet exist
Useful readiness signals were being created through outreach rather than captured early enough for prediction.
→
So what
Use engagement to create the signals a future model could learn from.
04 · Journey + product direction
PATIENT + NURSE / CLINIC SWIMLANES
The journey became the decision tool.
I synthesized the interview and workflow evidence into the journey map below. It showed when risk became visible, where there was still time to intervene, and where each concept fit. Chat and reminders could prevent breakdowns and create signals for a nurse dashboard; the cancellation model could not proceed because the passive data it required did not exist.
Prioritize now: AI patient chat + automated reminders. Strong journey fit, feasible for the pilot, and creates readiness signals.
Enable next: Nurse readiness dashboard. Uses chat-generated readiness signals, not a predictive score.
Deprioritize for now: Cancellation-risk model. Relied on passive signals and data that did not exist.
Journey stage
Scheduling
3 weeks before
→
Preparing
5 days before
→
Questions
2 days before
→
Nurse outreach
1 day before
→
Procedure day
Day of procedure
Readiness window
Readiness can still changeRisk finally becomes visible
Intervention window
Patient engagement acts herePrediction acts here, with less time to intervene
Concept prioritization
Prioritize now
AI patient chat + automated reminders
Where it acts
Scheduling through Questions
Why this decision
Strong journey fit · Feasible for the pilot · Creates readiness signals.
Enable next
Nurse readiness dashboard
Where it acts
Questions through Nurse outreach
Why this decision
Uses chat-generated readiness signals, not a predictive score.
Dependency
Enabled by readiness signals collected through patient chat.
Deprioritize for now
Cancellation-risk model
Where it acts
Nurse outreach through Procedure day
Why this decision
Relied on passive signals and data that did not exist.
Patient
Receives a procedure date and printed preparation instructions
Interprets preparation guidance alone at home
Needs reassurance but hesitates to call the clinic
Finally discusses readiness with a clinician
Arrives ready, or cannot proceed
Nurse / clinic
Books the procedure and moves to the next patient
Has no visibility into progress or confusion
Does not know that support is needed
Calls, assesses, escalates, or reschedules
Proceeds or loses the appointment slot
Workflow breakdown
Readiness begins, but attention and risk are both low.
Misunderstanding stays silent while it is still preventable.
Uncertainty peaks without an easy, trusted way to ask.
Useful risk signals appear late and only when a nurse has time to call.
A late failure is difficult to recover.
Opportunity
Set expectations without overwhelming the patient.
Deliver small, timed actions and reminders.
Answer approved questions and capture readiness signals.
Give nurses a readiness view and a prioritized worklist.
Capture outcomes to improve future readiness decisions.
Evidence + decision
Supporting research
Patients often set instructions aside until preparation felt immediate.
Synthesized research observation
Roadmap decision
Scheduling was ruled out as the primary intervention moment because patients were not ready to act.
Supporting research
Confusion could remain invisible until the patient arrived or contacted the clinic.
Synthesized research observation
Roadmap decision
This became the largest unaddressed window in the readiness journey.
Supporting research
Patients looked for answers elsewhere when a question felt too small to justify calling.
Synthesized research observation
Roadmap decision
This became the target moment for patient engagement because prevention was still possible.
Supporting research
Nurses generated useful readiness signals during outreach, but those signals were not available beforehand.
Synthesized workflow observation
Roadmap decision
Prediction depended on information the engagement experience could generate earlier.
Supporting research
Once a slot failed on procedure day, the organization had little opportunity to recover it.
Synthesized operational observation
Roadmap decision
Readiness outcomes connected the scaling challenge to the longer-term prediction opportunity.
Scroll horizontally to follow the full journey on smaller screens.
The product direction
Prevent, learn, then predict.
Patient engagement could deliver immediate value, fit the pilot, and generate the readiness data a future cancellation-risk model would need.
05 · Designing the intervention
AI CONVERSATION + NURSE WORKFLOW
Designing AI-guided conversations to reduce readiness risk.
Phase 1 used an AI-guided patient conversation to reinforce preparation, answer questions, and surface readiness signals. Phase 2 carried those signals into a nurse workflow so staff could focus outreach where intervention was needed.
Phase 01 · Patient experience
Choose the applicable preparation instructions
01 · Product decision
Choose the applicable preparation instructions
The first phase asked patients to identify the instructions they had received before the experience could provide relevant guidance.
Constraint + tradeoff
Self-selection was not ideal, but the product could not retrieve instructions from the referring site or store them through another available integration.
Guide the patient through preparation
02 · Product decision
Guide the patient through preparation
The conversation used the selected instructions as its boundary, translating static guidance into timely questions, answers, and checklist actions.
Safety decision
Answers had to remain grounded in approved instructions and preserve a clear escalation path when clinical judgment was required.
Phase 02 · Nurse readiness
Turn conversation signals into focused outreach
The second-phase dashboard organized patients around readiness flags created through chat so nurses could see where judgment and outreach were needed.
Prototype approach
AI prototyping tools accelerated exploration and iteration of the dashboard concept. They supported design speed, not the production system or its clinical decisions.
Patient readiness · Prototype
Turn conversation signals into focused outreach
06 · Pilot + impact
6 WEEKS · HERSHEY ENDOSCOPY
Cross-functional partners brought the pilot to life and tested the direction.
More than 200 patients used the experience. The pilot tested readiness engagement and workflow fit; it was not designed to claim a statistically proven reduction in cancellations.
Verified scope200+patients in the pilotA live clinical environment
Verified duration6 weeksfrom launch to pilot readoutHershey Endoscopy
Pilot learningEarlierreadiness signalsCreated through patient engagement
Direction1shared product pathEngagement first, prediction later
Alignment
The team moved from two competing approaches to one shared product direction.
Product direction
Patient engagement came first because it could prevent problems and create better data.
Sequencing
Cancellation prediction remained on the roadmap, supported by signals the pilot could generate.
Evidence
The pilot gave the organization a concrete basis for deciding what to fund and learn next.
I reframed a technology choice as a product decision, built the evidence to choose, designed the experience, and helped the organization move forward with a direction it could test and scale.
