A structured evaluation tool for vetting AI use cases across 4 pillars and 12 blocks. Score each block as Red, Yellow, or Green to identify critical gaps before you build.
By Ajay Pundhiraskajay.ai/toolsVersion 1.1 · Updated 2026-05-03
Score each of the 12 blocks as Red (critical gap), Yellow (manageable gap), or Green (validated). Circle the appropriate color for each block.
Foundation or Value Engine reds = STOP or PAUSE. These are structural problems. Do not proceed until they are addressed.
Execution reds = Proceed with caution. You can move forward, but build mitigation plans for each red block.
Sustainability reds = Plan for it before scale. These won't kill the pilot, but they will kill the product at scale.
Revisit the canvas at every development stage — what was green at ideation may turn yellow at production. Reassess continuously.
Scoring Principle: Pillar-Specific Severity
Not all reds are equal. A red in Foundation (Blocks 1-3) or Value Engine (Blocks 4-6) means the use case has a structural flaw — stop and fix it. A red in Execution (Blocks 7-9) means you can proceed carefully with a mitigation plan. A red in Sustainability (Blocks 10-12) means you can pilot, but must resolve before scaling.
Brand commitments behind this canvas
Stop promising demos from becoming costly failures. A 90-minute pilot with a slick UI is not validation. The Canvas exists to surface the unanswered questions — data foundations, defensibility, unit economics, governance — before you commit budget and reputation to a use case that does not survive contact with reality.
Score the use case that is, not the use case you wish for. Each block is scored on what is operationally true today — current data quality, current talent, current governance, current contract structure. Aspirational scoring breaks the framework. The Canvas is a diagnostic instrument, not a marketing deck.
1
Foundation — Strategic Intent
Does this use case solve a real, high-value problem with a viable data and technology approach?
1
Problem Definition
What specific, high-value problem does this AI use case solve?
Is this a painkiller (urgent, must-solve) or a vitamin (nice-to-have improvement)?
Can you quantify the pain? (dollars lost, hours wasted, errors per month, customer churn rate)
Score
R
Y
G
2
Data Strategy
What data is required to power this use case? Is it available, accessible, and of sufficient quality?
Where does the data come from? (internal systems, third-party, user-generated, public)
What is the cold start plan? How do you get enough data to deliver value on day one?
Is there a data flywheel — does usage of the product generate data that makes the product better?
Score
R
Y
G
3
Technology Core
Build from scratch, fine-tune an existing model, or call an API? What is the right approach and why?
What does the production architecture look like? (latency requirements, throughput, reliability)
What are the key technical risks and unknowns? What needs to be proven before committing?
Score
R
Y
G
Foundation Gate Check
If any block in this pillar is Red, STOP. A use case without a clear problem, viable data strategy, or feasible technology approach is not ready to proceed. Go back to problem discovery.
2
Value Engine — Business Value
Does this use case deliver provable business value with real market demand and defensibility?
4
Value Proposition
What specific outcome does this AI use case deliver to the end user or business?
Can you frame the value in concrete business terms? (revenue generated, cost reduced, time saved, risk mitigated)
Is the value measurable and attributable — or is it vague ("improves efficiency")?
Score
R
Y
G
5
Market Validation
Who specifically wants this? Can you name them, describe them, and talk to them?
What evidence do you have beyond "everyone needs AI"? (interviews, pilots, letters of intent, usage data)
Is there demonstrated willingness to pay — or change workflows — for this solution?
Score
R
Y
G
6
Competitive Landscape
What alternatives exist today? (competitors, manual workarounds, status quo)
What is your defensibility? Why can't someone else replicate this quickly?
What moats are you building? (proprietary data, deep workflow integration, domain expertise, network effects)
Score
R
Y
G
Value Engine Gate Check
If any block in this pillar is Red, PAUSE. A use case without clear value, validated demand, or defensibility will fail commercially — even if the technology works perfectly.
3
Execution — Implementation
Can you actually build, ship, and deliver this use case to real users?
7
User Experience
What is the interface? How do users interact with this AI? (embedded, standalone, API, ambient)
How does the system handle errors, uncertainty, and low-confidence outputs?
Can the AI explain its outputs in terms users understand? Is explainability required?
Who are the ACTUAL users? (not the buyers — the daily users) Have you observed their workflow?
Score
R
Y
G
8
Technical Architecture
What does the production architecture look like? (not the notebook — the real system)
How does this integrate with existing systems, data sources, and workflows?
How does the system scale? What happens at 10x, 100x current load?
What is the rollback plan if the AI system fails or produces harmful outputs in production?
Score
R
Y
G
9
Go-to-Market
How do users discover and adopt this? What is the distribution strategy?
How do you build trust? (pilots, case studies, guarantees, transparency about AI involvement)
What is the pricing model? Does it align with value delivered? (per-use, subscription, outcome-based)
What change management is needed? How much behavior change are you asking of users?
Score
R
Y
G
Execution Gate Check
Reds in Execution mean proceed with caution, not stop. Build specific mitigation plans for each red block. Assign owners and timelines. Review weekly.
4
Sustainability — Long-Term Viability
Will this use case survive contact with reality — at scale, over time, under scrutiny?
10
Unit Economics
What is the cost per inference / per transaction / per user? Is it sustainable?
How do economics change at scale? Do costs decrease (economies of scale) or increase (compute scaling)?
Are margins improving over time? Is there a clear path to healthy unit economics?
Score
R
Y
G
11
Governance & Compliance
What regulations apply? (GDPR, HIPAA, EU AI Act, industry-specific rules) Are you compliant?
How are errors, harms, and unintended outputs handled? Is there a documented incident response process?
How will you adapt as compliance requirements evolve? Who owns regulatory monitoring?
Score
R
Y
G
12
Continuous Improvement
What feedback loops exist? How does user behavior, outcome data, and error reporting flow back into improvement?
What are the success metrics? How do you know this use case is working — and when do you know it's not?
What is the iteration plan? How frequently will you retrain, update, and improve the system?
Score
R
Y
G
Sustainability Gate Check
Reds in Sustainability won't kill your pilot — but they will kill your product at scale. Address these before moving from pilot to production rollout.
Canvas Scorecard Summary
Circle one score per block. Transfer your scores from each section above.
Pillar 1: Foundation
1. Problem Definition
R
Y
G
2. Data Strategy
R
Y
G
3. Technology Core
R
Y
G
Pillar 2: Value Engine
4. Value Proposition
R
Y
G
5. Market Validation
R
Y
G
6. Competitive Landscape
R
Y
G
Pillar 3: Execution
7. User Experience
R
Y
G
8. Technical Architecture
R
Y
G
9. Go-to-Market
R
Y
G
Pillar 4: Sustainability
10. Unit Economics
R
Y
G
11. Governance & Compliance
R
Y
G
12. Continuous Improvement
R
Y
G
Scoring Interpretation Rules
What to do based on where your reds appear.
Pillar with Red
Action Required
Foundation Blocks 1–3
STOP. The use case has a structural flaw. Return to problem discovery. Do not invest further until all Foundation blocks are Yellow or Green.
Value Engine Blocks 4–6
PAUSE. The technology may work, but the business case doesn't hold. Validate demand, sharpen the value proposition, or find defensibility before continuing.
Execution Blocks 7–9
PROCEED WITH CAUTION. Build mitigation plans for each red block. Assign owners, set timelines, and review weekly. These are solvable with effort.
Sustainability Blocks 10–12
PLAN BEFORE SCALE. You can pilot, but do not scale until these are addressed. Sustainability reds become fatal at production volume.
All Green Blocks 1–12
FULL SPEED AHEAD. Rare but powerful. Execute with confidence. Continue reassessing at each development stage — green today doesn't guarantee green tomorrow.
Notes & Observations
Key risks and blockers identified:
Immediate next actions:
Open questions to resolve:
Glossary
Definitions used throughout this worksheet. These align with the canonical methodology article.
AI Use Case Canvas. A 4-pillar / 12-block decision framework for evaluating any AI initiative before resource commitment. Pillars: Foundation (Strategic Intent), Value Engine (Business Value), Execution (Implementation), Sustainability (Long-Term Viability).
STOP / PAUSE / PROCEED WITH CAUTION / PLAN BEFORE SCALE. The four pillar-level verdicts. STOP = critical block failures preclude this use case. PAUSE = significant block failures need remediation before proceeding. PROCEED WITH CAUTION = use case is viable but requires explicit risk acknowledgement. PLAN BEFORE SCALE = use case is sound but operating model needs design before scaling beyond pilot.
Cold Start Problem. The challenge of launching an AI use case with insufficient data to train or evaluate the model. Affects Block 2 (Data Strategy) scoring.
Data Flywheel. The compounding advantage when AI use creates more/better data, which improves the model, which drives more use. Affects Block 4 (Value Proposition) and Block 12 (Continuous Improvement) scoring.
Defensibility Moat. The mechanism by which an AI use case sustains competitive advantage over time. Most often a function of data, distribution, or proprietary workflow integration — rarely the model itself. Affects Block 6 (Competitive Landscape) scoring.
Foundation Model. A large, general-purpose AI model (GPT-4, Claude, Gemini) that serves as the base for downstream applications. Buying access to a foundation model is partly vendor evaluation, partly platform-risk acceptance.
Unit Economics. The cost-per-output and revenue-per-output of the AI use case at production scale. The single most under-tested element in pilot phase. Affects Block 10 (Unit Economics) scoring.
Governance & Compliance Block (Block 11). The Canvas block where regulatory exposure (EU AI Act, GDPR, HIPAA, sector rules) is scored alongside internal governance maturity. EU AI Act penalty tiers under Article 99: 7%/€35M for prohibited practices (Art. 99(3)), 3%/€15M for high-risk system non-compliance (Art. 99(4)), 1%/€7.5M for misleading authorities (Art. 99(5)).
Evidence Base
This worksheet is the operational layer of a published, sourced framework. The methodology, viability data, and companion frameworks live in the references below.