Proposing blind
Rates were set with no visibility into what competing payers were offering for the same providers.
HARMONYIQ
AI-powered contract intelligence for US health plans — turning a weeks-long negotiation process into something a contract manager could actually trust.
Read the case study ↓
30+
INTERVIEWS
INTERVIEWS
WCAG
2.1 AA
2.1 AA
10K+Healthcare providers
30+User interviews
5Contract stages unified
AAWCAG 2.1 throughout
01
THE BRIEF
MY FIRST MONTH WAS ALMOST
ENTIRELY INTERVIEWS.
Healthcare provider contract management is dense. Rates, compliance, network adequacy, renewal cycles — all interconnected, all high-stakes. I joined as PD1 on a team building an AI intelligence layer on top of this.
Before I could design anything useful, I needed to understand what contract managers actually did with their days — not what the brief said they did. Thirty-plus interviews later, the real problem came into focus.
My scope: own user research, ground every design decision in clinical workflow reality, and make sure every screen held up to WCAG 2.1 AA — because a healthcare tool that isn't accessible isn't a healthcare tool.
02
WHAT THE RESEARCH FOUND
EVERY STAGE HAD ITS OWN WAY
OF BREAKING.
The problems weren't abstract. They came up in the same forms across every interview, across every market.
The wrong tools for the job
Word for proposals. Excel for tracking. Email for iteration. No version history. No audit trail.
Scattered data, manual assembly
Claims history, member trends, and quality scores lived in three separate systems. Pulling them together took days.
Uncaught billing errors
When contract terms and actual claims weren't aligned, providers got reimbursed incorrectly. Nobody had an easy way to catch it.
Approval chains with no record
Stakeholders chased sign-offs over email. No flow, no timestamps, no accountability.
03
ARCHITECTURE
THE SYSTEM MAP CAME BEFORE
ANY SCREEN DID.
Once the interviews were synthesised, we mapped the complete contract lifecycle end to end. That diagram became our shared language — with the team and with stakeholders — for the entire project. Three distinct personas emerged from it, each with non-overlapping needs.
I need to walk in informed.
Benchmark data, AI-recommended strategy, and a complete picture of the provider — before a single word of negotiation.
I need to prove every decision.
Full version history, timestamped approvals, traceable edits. If it happened, it has to be findable.
I need status without chasing.
Where is this contract, who has it, what's blocking it — in one view, updated in real time.
04
THE WORK
AI EMBEDDED AT EVERY STAGE —
NOT BOLTED ON AT THE END.
Five screens for five stages. The guiding constraint throughout: the contract manager should always know what the right next action is, and why. WCAG 2.1 AA on every component, every state.
- 01Initiate Proposal
- 02Finalise Proposal
- 03Model Pricing
- 04Initiate Contract
- 05Finalise Contract
01
PROVIDER 360
The complete picture before negotiation starts.
The entry point for every negotiation. Before any proposal starts, the contract manager sees the provider's complete picture — AI-flagged alerts, where the contract currently sits in its lifecycle, the financial relationship, and the full interaction history. No more going in blind.
What needs attention this week — surfaced by the AI before the manager asks.
The current stage in the lifecycle, visible at a glance for every provider.
Escalations, calls, member feedback — in context where they're actually useful.
02
INITIATE PROPOSAL
Draft where the context lives.
Built around how contract managers actually draft — not how a PM imagined they did. A structured proposal editor with embedded rate tables, sitting next to an AI panel that surfaces a recommended strategy, surfaces known friction with this provider, and shows competitive benchmark data in real time.
AI suggests a rate with its reasoning. Accept, reject, or come in with your own number.
Past billing disputes, escalations, known issues — visible before the proposal is sent.
Proposed reimbursement figures sit inside the proposal itself — not in a separate tab.
03
MODEL PRICING
Three scenarios. Visible consequences.
Pricing decisions don't happen in a vacuum. This screen lets contract managers run up to three scenarios in parallel — baseline against two variants — each one showing projected reimbursement and real-time risk positioning relative to the contracting strategy.
Baseline vs variant 1 vs variant 2, with variance from baseline shown for each.
Elevance, UHC, Aetna, Cigna — where does this proposal sit against peer payers.
Pick a scenario, watch the risk position update. Every decision has a visible consequence.
04
FINALISE, DRAFT & SIGN
A reviewable contract in one session.
After a scenario is selected, the AI recalculates the overall risk position and auto-populates the contract with NPI data it has verified against public sources. The result: a fully drafted, reviewable contract ready for provider signature — with a complete edit history from day one.
Every change is re-evaluated before the contract moves forward.
Provider details auto-populated from verified public sources. Twelve fields filled; eight flagged for human review.
Every version, every edit, every signature timestamped. A process that used to take weeks, closed in one session.
05
DESIGN DECISIONS
WHERE THE REAL THINKING
HAPPENED.
INSTEAD OF
A single AI recommendation card
“Recommended rate: $X.” Accept or ignore.
WE USED
A positioning slider that shows consequences
Moving the slider makes the downstream risk visible. Every proposal becomes a strategic decision, not a field to fill.
INSTEAD OF
Keeping provider friction history out of the flow
A clean slate per proposal. Past disputes are the past.
WE USED
Friction history as a feature, not a footnote
Knowing a provider has a billing dispute history changes how you walk into a renewal. That context is too valuable to hide.
INSTEAD OF
An “AI Insights” tab you opt into
AI lives in one place. You go there when you want it.
WE USED
AI embedded at every decision point
In the proposal editor, inside pricing simulation, within the draft. Always present, never a detour.
INSTEAD OF
Full AI automation
Let the AI fill everything. Move faster. Trust it.
WE USED
AI fills, human reviews
12 fields filled from verified sources. 8 fields flagged for human confirmation. Trust is built through transparency, not magic.
06
REFLECTION
WHAT THISBUILT IN ME.
Research is design
The 30+ interviews weren't prep work before the real design started. They were the design. The architecture, the AI touchpoints, the persona logic — all of it came directly from what people said in those sessions.
Accessibility is not a checklist item
Building WCAG 2.1 AA into the component library from day one — rather than auditing at the end — meant we never had to retrofit. The compliance lead persona made this non-negotiable: an inaccessible audit trail defeats the point.
Show the AI's reasoning
Healthcare professionals don't accept black-box outputs. Every recommendation needed to show its logic. Users who could see why the AI suggested something trusted it significantly more than users who just saw the output.
Upstream decisions have long shadows
The system architecture built in the first two weeks set the ceiling for everything designed afterward. Getting that right — or wrong — compounds with every screen.
07
OUTCOME
WHAT WE
SHIPPED.
Contract stages brought into one platform
Providers across the US healthcare market
Interviews driving every design decision
WCAG 2.1 compliance across the full platform
Back to the portfolio
See selected work →