Discovery
Problem validation, user research, and hypothesis testing. Where AI opportunities are identified and scoped.
Value Stream
The AIVA Value Stream represents the complete journey from identifying an AI opportunity to realizing measurable business value. Understanding where work flows freely and where it stalls is the first step to acceleration.
The Four Phases
Section titled “The Four Phases”Delivery
Implementation, integration, and shipping to production. Where AI solutions are built and deployed.
Validation
A/B testing, customer feedback, and impact measurement. Where AI value is confirmed and quantified.
Foundations
Platform enablers, tooling, and infrastructure. The supporting capabilities that accelerate all phases.
Discovery Phase
Section titled “Discovery Phase”Purpose: Ensure the right problems are being solved before investing in implementation.
Discovery is where AI initiatives begin. Teams identify potential use cases, validate that problems exist, research user needs, and form hypotheses about how AI can create value.
Key Activities
Section titled “Key Activities”| Activity | Description |
|---|---|
| Problem Identification | Surfacing candidate problems where AI could add value |
| User Research | Understanding user needs, pain points, and workflows |
| Hypothesis Formation | Articulating clear, testable hypotheses about AI impact |
| Feasibility Assessment | Evaluating technical and data feasibility |
| Prioritization | Ranking opportunities by value, feasibility, and strategic fit |
Flow Indicators
Section titled “Flow Indicators”| State | What It Looks Like |
|---|---|
| Blocked | No clear process for identifying AI opportunities; ideas stuck in backlogs |
| Constrained | Long queues for discovery work; slow stakeholder alignment |
| Flowing | Regular discovery cadence; validated opportunities flowing to delivery |
| Accelerated | Proactive opportunity identification; rapid experimentation cycles |
Metrics
Section titled “Metrics”- Discovery Duration - Time from opportunity identification to validated hypothesis
- Validation Rate - Percentage of discoveries that proceed to delivery
- Discovery-to-Delivery Handoff - Time between discovery completion and delivery start
Delivery Phase
Section titled “Delivery Phase”Purpose: Build and ship AI solutions to production efficiently and reliably.
Delivery transforms validated opportunities into working AI products. This encompasses development, testing, integration, and deployment.
Key Activities
Section titled “Key Activities”| Activity | Description |
|---|---|
| Development | Building AI models, integrations, and user interfaces |
| Testing | Validating functionality, performance, and safety |
| Integration | Connecting AI solutions to existing systems and data |
| Deployment | Shipping to production with appropriate monitoring |
| Iteration | Responding to early feedback and refining solutions |
Flow Indicators
Section titled “Flow Indicators”| State | What It Looks Like |
|---|---|
| Blocked | Unable to ship; critical dependencies unmet; production access issues |
| Constrained | Slow release cycles; long PR review queues; deployment friction |
| Flowing | Regular releases; predictable cycle times; smooth deployments |
| Accelerated | Continuous deployment; rapid iteration; feature flags enabling safe experiments |
Metrics
Section titled “Metrics”- Cycle Time - Time from work started to production deployment
- Deployment Frequency - How often code reaches production
- Lead Time for Changes - Time from commit to production (DORA metric)
- Change Failure Rate - Percentage of deployments causing issues (DORA metric)
Validation Phase
Section titled “Validation Phase”Purpose: Confirm that shipped AI solutions deliver expected value.
Validation closes the loop. It measures whether the AI solution actually solves the problem it was designed to solve and quantifies the business impact.
Key Activities
Section titled “Key Activities”| Activity | Description |
|---|---|
| A/B Testing | Comparing AI solution against baseline/alternatives |
| Usage Analytics | Tracking adoption, engagement, and retention |
| Customer Feedback | Collecting qualitative input on value delivered |
| Impact Measurement | Quantifying business outcomes (revenue, efficiency, satisfaction) |
| Learning Capture | Documenting insights for future initiatives |
Flow Indicators
Section titled “Flow Indicators”| State | What It Looks Like |
|---|---|
| Blocked | No measurement infrastructure; unable to validate hypotheses |
| Constrained | Slow feedback loops; limited A/B testing capacity; unclear metrics |
| Flowing | Systematic validation process; clear success criteria; regular reviews |
| Accelerated | Real-time impact dashboards; rapid experiment iteration; continuous learning |
Metrics
Section titled “Metrics”- Time to Validation - Duration from production deployment to impact confirmation
- Experiment Velocity - Number of A/B tests or experiments completed
- Hypothesis Confirmation Rate - Percentage of hypotheses validated by data
- Value Realization - Quantified business impact (context-specific)
Foundations Phase
Section titled “Foundations Phase”Purpose: Build and maintain the platform capabilities that accelerate all other phases.
Foundations is the enablement layer. Rather than directly delivering user value, it creates the conditions for Discovery, Delivery, and Validation to flow faster and more reliably.
Key Activities
Section titled “Key Activities”| Activity | Description |
|---|---|
| Platform Development | Building shared AI/ML infrastructure |
| Tooling | Creating developer tools, templates, and accelerators |
| Data Infrastructure | Ensuring data availability, quality, and governance |
| Security & Compliance | Maintaining guardrails that enable safe experimentation |
| Training & Enablement | Building organizational capability |
Flow Indicators
Section titled “Flow Indicators”| State | What It Looks Like |
|---|---|
| Blocked | Foundational gaps preventing any AI work; no platform, no data access |
| Constrained | Platform exists but bottlenecks teams; manual processes; limited self-service |
| Flowing | Self-service platform capabilities; teams can move independently |
| Accelerated | Platform as competitive advantage; enabling capabilities teams didn’t know they needed |
Metrics
Section titled “Metrics”- Platform Adoption - Percentage of AI initiatives using shared infrastructure
- Self-Service Rate - Percentage of requests handled without platform team involvement
- Time to First Deployment - How quickly a new team can ship their first AI feature
- Platform Reliability - Availability and performance of shared services
Flow States Explained
Section titled “Flow States Explained”Every phase can be in one of four flow states. Understanding these states helps prioritize where to focus improvement efforts.
Blocked
Section titled “Blocked”Work cannot proceed. There are fundamental blockers that must be resolved before any progress is possible.
Common causes:
- Missing critical capabilities (no ML platform, no data access)
- Organizational blockers (no budget, no mandate)
- Dependency failures (waiting on external parties indefinitely)
Priority: Highest. Blocked phases require immediate intervention.
Constrained
Section titled “Constrained”Work proceeds, but slowly. Bottlenecks limit throughput and create queues.
Common causes:
- Limited capacity (too few ML engineers, overloaded reviewers)
- Process friction (manual approvals, slow environments)
- Technical debt (brittle systems, missing automation)
Priority: High. Constraints compound over time and should be addressed systematically.
Flowing
Section titled “Flowing”Work moves steadily and predictably. The phase functions well without exceptional effort.
Common causes:
- Right-sized capacity for current demand
- Effective processes and tooling
- Strong team capabilities
Priority: Maintain and monitor. Flowing phases should be protected from regression.
Accelerated
Section titled “Accelerated”Work moves faster than baseline expectations. Improvements compound.
Common causes:
- Investment in enablement paying dividends
- Cultural excellence (psychological safety, learning orientation)
- Platform leverage (shared capabilities multiplying team output)
Priority: Learn and spread. Understand what enables acceleration and apply elsewhere.
Measuring Value Stream Health
Section titled “Measuring Value Stream Health”GuideMode provides analytics across the value stream:
| Flow | GuideMode Analytics |
|---|---|
| Discovery | Discovery Flow metrics, validation rates, handoff times |
| Delivery | Delivery Flow metrics, DORA metrics, cycle times |
| Validation | Survey feedback, assessment results, team health |
| Foundations | Platform adoption, infrastructure metrics |
Common Anti-Patterns
Section titled “Common Anti-Patterns”Delivery Without Discovery
Section titled “Delivery Without Discovery”Shipping AI features without validating the problem. Results in solutions looking for problems.
Symptoms: High delivery velocity but low impact; features unused after launch.
Delivery Without Validation
Section titled “Delivery Without Validation”Shipping without measuring outcomes. Results in uncertainty about actual value.
Symptoms: Many launches but unclear ROI; inability to make data-driven decisions.
Foundations Without Flow
Section titled “Foundations Without Flow”Over-investing in platform before validating demand. Results in infrastructure not used.
Symptoms: Sophisticated ML platform but few teams using it; high cost, low utilization.
Next Steps
Section titled “Next Steps”- Capability Lens - Understand what enables or constrains each phase
- Why Acceleration - The philosophy of continuous improvement
- Getting Started - Take the assessment