Identifying Leverage Points

Not all improvements are created equal. Some changes cascade through your product and organization, creating multiplicative benefits. Others drain resources with minimal impact. The difference is leverage. Understanding where to apply effort is as important as the effort itself.

High-Frequency vs High-Impact Tasks

Two dimensions define improvement opportunity: frequency and impact. Frequency is how often a task occurs. Impact is how much difference the task makes to outcomes. The most valuable improvements combine both.

                    HIGH IMPACT              LOW IMPACT
                ┌─────────────────────┬─────────────────────┐
HIGH FREQ       │                     │                     │
                │  STRATEGIC PRIORITY │   OPTIMIZE IF       │
                │  These tasks are    │  resources allow     │
                │  worth significant  │                     │
                │  investment         │                     │
                ├─────────────────────┼─────────────────────┤
LOW FREQ        │                     │                     │
                │   BIG BET           │   DEPRIORITIZE      │
                │   Could be          │   Fill in when      │
                │   transformative     │   nothing better   │
                │   but rare          │   to do            │
                │                     │                     │
                └─────────────────────┴─────────────────────┘
            

Applying the Matrix

The Frequency-Impact Matrix guides investment decisions based on where tasks fall across the two dimensions. High frequency combined with high impact indicates tasks where you should invest heavily, as these are the core of your AI strategy and will generate the most value. High frequency combined with low impact suggests tasks to optimize for efficiency, as these may be candidates for automation where volume compensates for individual low impact. Low frequency combined with high impact means you should design carefully for when these important but rare events occur, ensuring they are handled well despite their infrequency. Low frequency combined with low impact suggests you should minimize investment, as these tasks are not worth much attention.

QuickShip: Frequency-Impact Analysis

TASK ANALYSIS FOR QUICKSHIP OPERATIONS:

┌─────────────────────────────────────────────────────────┐
│ TASK                │ FREQ    │ IMPACT  │ PRIORITY      │
├─────────────────────────────────────────────────────────┤
│ Exception handling   │ High    │ High    │ STRATEGIC     │
│ Route optimization   │ High    │ High    │ STRATEGIC     │
│ Customer status      │ Very    │ Low     │ OPTIMIZE       │
│ inquiries            │ High    │         │               │
│ Driver performance   │ Medium  │ Medium  │ SECONDARY     │
│ review               │         │         │               │
│ Weekend volume       │ Low     │ High    │ BIG BET       │
│ prediction           │         │         │               │
│ Weather impact       │ Low     │ High    │ BIG BET       │
│ planning             │         │         │               │
│ New driver training  │ Low     │ Low     │ DEPRIORITIZE  │
└─────────────────────────────────────────────────────────┘

PRIORITIZED AI PORTFOLIO:
1. Exception handling AI (strategic - highest leverage)
2. Route optimization AI (strategic - core to value prop)
3. Predictive volume modeling (big bet - could transform planning)
            

Bottleneck Analysis

Bottlenecks are constraints that limit the throughput of the entire system. Unlike local inefficiencies that affect only one step, bottlenecks constrain overall capacity. Fixing a bottleneck improves the entire system.

Identifying Bottlenecks

Look for these signs of bottlenecks in your workflows. Work piling up, where inventory or backlog builds before a particular step, indicates that step cannot keep pace with upstream flow. Downstream starvation occurs when resources sit idle because upstream cannot deliver work fast enough, showing that the bottleneck is earlier in the process. The longest cycle time identifies a step that takes significantly longer than others, which often constrains overall throughput. Dependency chains reveal steps that cannot run in parallel and must proceed sequentially, creating inherent limitations. Single points of failure occur when one person or system has responsibility that everything depends on, creating vulnerability and constraint.

The Theory of Constraints

Based on Eli Goldratt's Theory of Constraints, improvement efforts should follow a systematic five-step approach. First, identify the system bottleneck by finding the constraint that limits overall throughput. Second, exploit the bottleneck by making maximum use of its capacity, ensuring it is not starved or interrupted. Third, subordinate all other work to support the bottleneck, aligning everything else to keep the constraint running at full efficiency. Fourth, if the bottleneck constraint remains insufficient after exploitation and subordination, elevate by adding capacity to it directly. Fifth, repeat the process when one bottleneck is resolved, as fixing one constraint often reveals the next one in the system.

PATIENT FLOW BOTTLENECK ANALYSIS:

System throughput: Patients processed per day

Bottleneck Identification:
├─ ER arrivals: 120 patients/day
├─ Triage: 125 patients/day capacity
├─ Bed assignment: 80 patients/day capacity ← BOTTLECK
├─ Treatment: 110 patients/day capacity
└─ Discharge: 90 patients/day capacity

Bottleneck: Bed assignment (80 patients/day)
Constraint: Limited by nurse availability for intake

EXPLOIT STRATEGIES:
1. Pre-populate bed assignment forms during triage
2. Use AI to predict bed availability 2 hours ahead
3. Flag potential bottlenecks proactively to administration

SUBORDINATE STRATEGIES:
1. Release beds earlier when discharge is predicted
2. Stagger ER arrivals when predictable surge expected
3. Move stable patients to "pending discharge" status

ELEVATE STRATEGIES:
1. Cross-train nurses from other departments
2. Implement floating nurse pool for bottleneck coverage
            

Quick Wins vs Strategic Investments

A balanced AI portfolio includes both quick wins and strategic investments. Quick wins build momentum and prove value. Strategic investments create transformative but longer-term capabilities.

Characteristics of Quick Wins

Quick wins share several distinct characteristics that make them valuable portfolio additions. They have narrow scope, being limited to a specific, well-defined task rather than sprawling across multiple workflows. They have clear success metrics that make it easy to measure whether the initiative worked. They carry low technical risk because they use proven AI approaches with established track records. They provide fast feedback with results visible within days or weeks rather than months. They have limited dependencies, meaning they can be built and deployed independently without waiting for other initiatives or coordinated cross-team efforts.

Characteristics of Strategic Investments

Strategic investments differ from quick wins in several fundamental ways. They have broad scope that affects multiple workflows or user types rather than targeting a single task. Their success is multi-dimensional, with value that is complex and long-term rather than easily measurable in the short term. They carry higher technical risk because they require new capabilities or approaches that have not been proven in your context. They have long feedback cycles, potentially taking months to see full value rather than weeks. They have cross-cutting dependencies that require coordination across teams, making them more complex to manage and deploy.

HealthMetrics: Building a Balanced Portfolio

EVALUATING TWO AI INITIATIVES:

INITIATIVE A: AI-Powered Discharge Prediction
├─ Scope: Predict which patients will be ready for 
│         discharge in next 24 hours
├─ Frequency: Daily for every patient
├─ Impact: High (frees beds, reduces wait times)
├─ Technical Risk: Medium (requires new model)
├─ Time to Value: 3-4 months
└─ Classification: STRATEGIC INVESTMENT

INITIATIVE B: AI-Assisted Exception Note Writing
├─ Scope: Help nurses draft exception notes
├─ Frequency: 20-30 times per day
├─ Impact: Low (saves 2-3 minutes per note)
├─ Technical Risk: Low (standard text generation)
├─ Time to Value: 4-6 weeks
└─ Classification: QUICK WIN

DECISION:
- Pursue BOTH in parallel
- Quick win (B) builds AI culture and proves value
- Strategic investment (A) creates major competitive advantage
- Different timelines, different teams, different metrics

PORTFOLIO BALANCE:
- 60% of AI resources to strategic (A)
- 30% to quick wins (B and similar)
- 10% to exploration
            

Calculating ROI for AI Initiatives

Before committing to AI initiatives, estimate return on investment. AI investments have costs that are often underestimated and value that is often overestimated.

Cost Categories

AI initiatives incur costs across four distinct categories that must all be factored into ROI calculations. Development costs include engineering time, data preparation, and infrastructure setup required to build the system. Operational costs cover model serving, API calls, and ongoing maintenance that keep the system running in production. Governance costs encompass compliance requirements, monitoring systems, and error handling processes that ensure the AI operates responsibly. Failure costs account for errors the AI may make, system downtime that disrupts operations, and reputation damage that results from AI failures.

Value Categories

AI initiatives generate value across four distinct categories that together determine true return on investment. Labor savings represent the time saved by automation or augmentation, reducing the human effort required to accomplish tasks. Error reduction captures the value of avoiding mistakes that would otherwise require correction or cause downstream problems. Revenue impact includes increased sales, improved customer retention, or better conversion rates that directly affect business revenue. Strategic value encompasses capabilities that create future advantage, positioning the organization for long-term success even if short-term ROI is harder to quantify.

NET VALUE = (Labor Savings + Error Reduction + Revenue Impact) 
            - (Development + Operations + Governance + Failure Costs)

PAYBACK PERIOD = Development Costs / Monthly Net Value

Example:
├─ Development: $200,000
├─ Annual Operations: $50,000
├─ Annual Labor Savings: $180,000
├─ Annual Error Reduction: $30,000
├─ Annual Net Value: $180,000 + $30,000 - $50,000 = $160,000
└─ Payback Period: $200,000 / $160,000 = 15 months
            

Cross-References to Strategy and UX

Leverage point analysis connects directly to product strategy and user experience design: