Introduction
As a lead designer, I focused on designing a centralized operational dashboard that showed real-time inconsistencies and established a shared foundation for consistent decision-making across teams.
Background
US health insurance still relies on a patchwork of outdated tools
Insurance operations were built across a disconnected system for benefits, commissions, enrollment, and claims. Most of which were never designed to work together.
Because these tools were already deeply embedded in day-to-day operations, teams couldn't replace them. They had to keep operations running while reconciling mismatched data and working around system gaps.
Vizcare was created as a shared operational foundation, bringing visibility and consistency across existing systems without disrupting them.
Challenges
Disconnection, not task complexity, drove operational risk
The real challenge wasn't slow execution. It was the risk of bad decisions caused by disconnected systems. Teams had to cross-check information across tools that didn't always agree.
Small inconsistencies often go unnoticed until they cause downstream problems.
Split Systems
Benefits, commissions, and claims lived in separate tools, leading teams to interpret the same data differently
Manual Processes
Agents manually verified information between systems, increasing rework and the risk of error
No Unified View
There was no single place to identify what was out of sync, delaying detection and response
Goals
Creating a shared operational baseline that teams could trust
The internal tool didn't fail because it was missing features. They failed because teams had no consistent way to understand what changed, what was reliable, and what needed action.
I defined three system-level goals to reduce operational risk and ensure consistent decisions across teams.
01_ Show What Changed
Make updates and inconsistencies visible as they happen, so issues don't stay hidden
02_ Stay In Flow
Provide the correct information in context without forcing agents to pause and cross-check systems
03_ Work from One Record
Ensure everyone relies on the same source of truth to avoid conflicting decisions
System Snapshot
Understanding failures involved tracking how problems spread across systems
Failures rarely originate in a single spot. They spread because systems were updated at different times and followed different rules. I mapped the operational chain from start to finish, not to show off complexity, but to identify where teams lost visibility and why small mismatches became significant risks.
User Research
Where teams read the same data differently
I spoke with 10 internal operators across membership operations, claim processing, and customer support teams, each with different decision-making responsibilities for the same record (2-3 per role).
By reviewing real support calls and auditing sync logs with product and engineering, we traced how each role relied on different signals to decide whether a case was "moving forward."
User Type 1: Membership Ops
Confirm enrollment status
Membership ops relied on surface status to confirm enrollment, with no signal for what was still pending downstream.
"It shows active on my screen. Why wouldn't it go through?"
User Type 2: Claims Analysts
Track and resolve batch errors
Claims analysts required complete data to proceed, but had no visibility into whether missing fields were still syncing or permanently blocked.
"I never know who to ask when info is missing or changing."
User Type 3: Call Center Agents
Handle calls/chats from clinics and patients
Agents were left to choose which system to trust, often escalating not because rules were unclear, but because ownership was.
"I just tell them I'll escalate. There's nothing else I can do."
So, instead of fixing screens in isolation, I designed features around how decisions actually broke down.
Feature Prioritization
From repeated workarounds to intentional design choices
I grouped recurring decision breakdowns observed across teams and traced each one to a targeted design response.
Information Architecture
Aligning how teams interpret the same system
To address these breakdowns, I needed more than individual feature fixes. Teams were operating on the same systems, but with different assumptions about ownership, scope, and flow.
Before moving into detailed design, I worked closely with business analysts, engineers, and product managers to map how each platform actually supported decisions across teams.
This revealed overlaps, clarified boundaries, and established a shared structure the entire product could be built on.
Solutions
Translating research insights into three focused themes
I structured the final solutions into three themes.
Each theme turns research insights into concrete workflows that reduce guesswork across enrollment, sync, and support operations.
01_ Step-Based Enrollment Flow
Enrollment issues did stem from complex rules in some cases. But more often, breakdowns occurred because decisions were made before the data was fully synced.
I redesigned the flow to control when decisions could happen, separating life events, coverage, and dependents into clear stages so users weren't forced to guess or backtrack when the system lagged behind.
Life events, coverage changes, and dependents are handled in separate stages to prevent incomplete or premature updates.
Inline tooltips show required documents only when they're relevant, reducing back-and-forth and preventing unnecessary rework.
02_ Centralized Sync Dashboard
Failures appeared in various parts of the tools, but there was no single way to trace what happened from start to finish. To address this, I created a centralized sync dashboard to make the entire transaction process visible from start to finish, allowing teams to track where data moved, where it got stuck, and why.
All sync alerts are combined into one view, with the initial failure clearly highlighted.
The team can begin from the point of breakdown and trace how the issue spread across systems, rather than cross-checking logs tool by tool.
I designed the error log to make failures explainable, not just visible. By tying each error to a specific record and handoff, agents could grasp why something failed and determine the next steps without defaulting to escalation.
This view was designed to help teams judge whether data was moving or actually blocked. Showing exchange-level updates in context reduced guesswork and prevented premature fixes based on incomplete signals.
I designed this map to understand where progress signals diverged across the system.
By tracking a single transaction through each handoff, I could identify where progress advanced, where it truly stalled, and how small mismatches escalate into real risks.
03_ Agent-Ready Membership & Commission Data
Membership and commission data technically existed across systems. But agents struggled to make decisions with confidence during live support because of the absence of a clearly owned, authoritative view that agents could rely on in the moment.
I designed consolidated data model that brought membership status, coverage, and commission records into a single, trusted source.
This allowed agents to verify information quickly, resolve disputes without second-guessing, and stay focused on supporting customers.
Part A:
I organized membership data into scannable sections so agents could confirm coverage and enrollment status without cross-checking other tools. This removed updates and gave agents confidence in what they saw.
Commission data was structured by agent, date, and status to identify and resolve discrepancies without manual back-and-forth, reducing disputes and shortening resolution time.
Part B:
During live calls, agents frequently waste time looking for answers while trying to maintain the conversation.
I integrated AI directly into the contact center to provide relevant context and previous answers in real time. The focus was on enhancing the efficiency and accuracy of the live call workflow.
+Detailed Views
AI Contextual Suggestions
I designed contextual suggestions to appear when agents needed them, so common questions could be answered without disrupting conversation flow. This helped agents stay focused on the caller.
Duplicate Question Detection
When repeat questions came up, agents were shown prior answers so they could confidently answer without retracing past steps, reducing backtracking and improving consistency during calls.
Outcome
Faster decisions, fewer escalations with clear trust signals
With clear ownership, traceable sync states, and AI-assisted context, agents spent less time verifying data and more time making confident decisions and supporting customers.
Most workflows required little guidance, even when handling real, anonymized production cases.
What still needs work
Two gaps became clear during testing:
some users hesitated at the very first steps, unsure where to begin
AI handled common flows well, but felt generic in rare edge cases
Next Steps
If I were to take this further, I'd focus on three things for better reassurace:
Explore other projects