All sections in this case
My role
Led product discovery, business-rule definition and UAT
Result
Accurate assignment in production
Optimized within the defined operating constraints
What needed to change
Manual assignment could be slow and difficult to explain. The product needed to make valid matches visible while preserving authorized review and override.
The product decision
Check hard eligibility first. Protect priority outcomes before comparing fit among valid options, with a clear reason for the recommendation.
A high-ranked option is useful only while it remains eligible and available. The workflow rechecks those conditions before acceptance and escalates when no option qualifies.
Generalized product model
How the model works
Qualification, availability and capacity remain binding. Business priority comes before fit; unassigned work stays visible for follow-up.
From decision to delivery
Led definition of matching variables, hard and soft constraints, override behavior, and audit needs.
Partnered with coordinators, domain experts, engineering, and operational leaders.
Product behavior and edge cases
- Eligibility checks
- Specialty matching
- Simplified licensure logic
- Availability and capacity
- Recommendation explanation
- Authorized human override
- No-match escalation
- Decision trace
- No eligible provider
- Two equally eligible providers
- Capacity changes after a recommendation, requiring the current eligibility, availability, capacity and timing to be checked again before acceptance
- A restricted case has one eligible provider
- A time-sensitive request enters an existing queue
- An authorized user selects a different eligible provider
When a cheaper match is the wrong assignment
The original product brief estimated that coordinators spent about 30% of their time deciding which cases to assign. The task required more than distributing an equal number of cases: eligibility, urgency, scarce qualifications and capacity all shaped a valid assignment.
The 2026 review records a UAT failure in which blended costs could favor lower-priority unrestricted work while a time-sensitive restricted case remained unassigned. The product question became which outcome must take priority before comparing otherwise valid matches.
| Decision layer | Product rule | Review question |
|---|---|---|
| Eligibility | Remove invalid matches before optimization | Can this reviewer accept this work now? |
| Business priority | Protect the priority outcome before fit and workload preferences | Does a lower-priority assignment displace more urgent restricted work? |
| Fit and scarce capability | Compare eligible matches while preserving scarce qualifications | Is this capability needed for another case with fewer options? |
| Capacity and control | Recheck capacity; explain unassigned work; retain coordinator review | Can the result be accepted, understood and corrected? |
From UAT finding to production
The original specification describes a limited pilot before expansion. The 2026 review captured the priority inversion and the policy choices being worked through at that stage. The correction is now implemented and working in production, as confirmed in the update below.
The correction is working in production
The corrected assignment workflow is operating accurately and optimally within its defined constraints. It protects priority outcomes before optimizing match preferences, while qualification, availability and capacity remain binding.
This closes the priority issue identified in UAT. The reconstruction below makes the decision visible with fictional cases, reviewers and allocations.
Production update · 8 October 2026. The original specification and UAT review document the problem and decision context.
A closer look at the decision
Every slot filled. The wrong case waiting.
Follow the decision behind the production correction with a fictional batch. Explore in 2D or 3D, or read the annotated comparison.
Evidence and attribution
I led the product decisions behind the corrected workflow. It now operates accurately and optimally within its defined constraints. I use fictional cases and simplified rules here to explain those decisions.
What I learned
I make the basis of a recommendation visible so people can review it. The priority correction reinforced the need to settle which outcome matters before optimizing among valid options.
Examine overrides and capacity changes to improve the assignment rules.
Explore the full case
Follow the delivery path, constraints and operating decisions in more detail.
Why the problem mattered
Time-sensitive review requests may need to be matched with an appropriately qualified and available provider under multiple operating constraints.
Manual assignment can create delay, inconsistency, and limited transparency. A rules-based product needed to explain recommendations, support review and override, and escalate safely when no eligible match existed.
The people behind the workflow
- Authorized coordinators
- Provider reviewers
- Operations leaders
- Audit and compliance partners
Partnered with coordinators, domain experts, engineering, and operational leaders.
Constraints and exceptions
- No eligible provider
- Two equally eligible providers
- Capacity changes after a recommendation, requiring the current eligibility, availability, capacity and timing to be checked again before acceptance
- A restricted case has one eligible provider
- A time-sensitive request enters an existing queue
- An authorized user selects a different eligible provider
Product strategy
- Treat qualification and simplified licensure as hard constraints.
- Use capacity and timing as transparent ranking inputs.
- Explain why a recommendation is eligible before asking for acceptance.
- If capacity changes before acceptance, recheck hard eligibility and current availability, capacity and timing; explain the updated recommendation for authorized human review.
- Allow authorized users to review or override a recommendation.
- Escalate to manual review when no eligible match exists.
The product model
- Eligibility checks
- Specialty matching
- Simplified licensure logic
- Availability and capacity
- Recommendation explanation
- Authorized human override
- No-match escalation
- Decision trace
Delivery and adoption
- Facilitated scenario design with operational coordinators
- Aligned domain and operating constraints
- Partnered with engineering on deterministic matching behavior
- Defined review, override, and audit acceptance criteria
- Designed measures for speed, balance, and user trust
Lessons and next decisions
Rules-based automation earns trust when users can understand the recommendation, see the constraints, and intervene when judgment is required.
- Study override patterns
- Refine ranking inputs
- Improve proactive capacity visibility
From the initial problem to delivery
Assignment problem → reviewable matching rules → delivered workflow
- Define
Worked with coordinators on qualification, capacity, timing and no-match conditions.
- Validate
Defined explanation, authorized override, escalation and audit acceptance criteria.
- Deliver
Saw the underlying provider-assignment workflow through initial delivery.
Leadership and delivery
Led product discovery, business-rule definition and UAT
Led definition of matching variables, hard and soft constraints, override behavior, and audit needs.
- Facilitated scenario design with operational coordinators
- Aligned domain and operating constraints
- Partnered with engineering on deterministic matching behavior
- Defined review, override, and audit acceptance criteria
- Designed measures for speed, balance, and user trust
Alternatives and tradeoffs
Alternatives and trade-offs
Make product trade-offs explicit.
Qualification and simplified licensure remain hard eligibility gates; capacity and timing rank only eligible candidates. An authorized coordinator retains review and override, and a no-match case goes to manual review.
01
Use capacity and timing as transparent ranking inputs.
02
Explain why a recommendation is eligible before asking for acceptance.
03
If capacity changes before acceptance, recheck hard eligibility and current availability, capacity and timing; explain the updated recommendation for authorized human review.
04
Allow authorized users to review or override a recommendation.
05
Escalate to manual review when no eligible match exists.
Confidentiality
I’ve generalized the implementation details in this case to protect confidential information.
Every provider identity and scenario is fictional. Licensure and matching logic are simplified and are not used for clinical decisions.