24–30
Build a bounded pilot
Readiness is strong. Define the evaluation set, autonomy boundary, human review, cost ceiling, and rollback plan before building.
AI & Intelligent Systems · 15 factors
Is this AI use case worth building?
A practical 15-point assessment for deciding what to build now, pilot carefully, or avoid. Score one real use case in ten minutes.
Before you score
Define the use case before scoring it
Avoid scoring a vague goal such as 'use AI in customer support.' Score a concrete workflow: 'draft answers to tier-1 support tickets using our approved knowledge base, with an agent reviewing every response.'
Optional. Stays in your browser unless you send the form below.
- 0Unknown / absentThe requirement is unclear, unavailable, or unowned.
- 1Partially readySome evidence exists, but material gaps remain.
- 2ReadyEvidence, ownership and controls are sufficient for the next step.
Score evidence, not optimism. If you cannot point to the data, owner, metric or control, score it 0 or 1.
-
Business value
A successful result changes revenue, cost, risk, speed, or customer experience in a measurable way.
-
Workflow clarity
The current task, inputs, outputs, exceptions, and handoffs are documented.
-
Repetition and volume
The task occurs often enough for improvement to matter.
-
Data access
Required data is available, lawful to use, and accessible in a usable format.
-
Data quality
Examples are representative, current, and sufficiently complete.
-
Ground truth
A reviewer can determine what a good answer looks like.
-
Evaluation metric
Accuracy, quality, latency, cost, or another success measure is defined.
-
Error tolerance
The cost of a wrong output is understood and proportionate to the controls.
-
Human oversight
A named person can review, approve, override, or escalate when needed.
-
Security and privacy
Access, sensitive data, retention, and vendor exposure are addressed.
-
Integration path
The system can receive inputs and return outputs reliably.
-
Operational owner
A business owner is accountable after launch, not only during the build.
-
Adoption fit
Users have a reason to trust and adopt the changed workflow.
-
Cost boundary
Build, model, infrastructure, review, and monitoring costs have limits.
-
Reversibility
The pilot can be contained, rolled back, or switched off without disruption.
Turn the score into a decision
What your total means
16–23
Fix the gaps first
The idea may be valuable, but one or more foundations are weak. Address the lowest-scoring factors before committing to implementation.
0–15
Do not build yet
The current use case is not ready. Clarify the workflow, improve data and evaluation, or select a more suitable problem.
Your weakest factors
Your next three actions
- 1Name the constraintWrite down the lowest-scoring factor that could invalidate the project.
- 2Design the smallest testChoose a bounded sample, named reviewer, success threshold, and time limit.
- 3Decide before scalingCompare evidence against the threshold. Build, revise, or stop — do not drift into production.
Next step
Want an independent assessment before committing budget?
The AI Feasibility & Architecture Assessment tests up to three use cases against your data, cost and risk, and ends in a go / no-go architecture.
AI Feasibility & Architecture AssessmentFrom $5,900 · 2–3 weeks
Discuss my score
Your score and weakest factors are added to the message for you.
Get your personal report
Score every factor, then get a PDF of your AI readiness scorecard built from your answers: your total and result, your weakest factors, all fifteen scores and the next steps. (Before every factor is scored, you get the blank scorecard.)