The 4 C's of enterprise AI.
Context, Control, Cost, and Choice. Four questions every agentic AI system must answer before it earns production. Models change every quarter — these four do not.
The four questions
C1 · Chapters 5–12
Context
Does the agent know what our words mean?
Certified metrics, a shared glossary, semantic layer, lineage.
Read the Context guide →
C2 · Chapters 13–17
Control
Can we prove what the agent did and was allowed to do?
Action permissions, human review ladders, audit logs, agent identity.
Read the Control guide →
C3 · Chapters 18–21
Cost
Does each task earn more than it spends?
Model routing, budgets per task, caching, cost-per-correct-answer.
Read the Cost guide →
C4 · Chapters 22–25
Choice
Could we leave our vendor in a quarter?
Open formats, portable context, standard tool interfaces, multi-model evals.
Read the Choice guide →
Miss one C, and this is how it fails.
Most failed AI projects are not model failures. They are one weak C showing up in production.
| If you miss… | You get… | What it looks like | The fix | Owner |
|---|---|---|---|---|
| Context | Wrong answers | Confident numbers that disagree with the finance dashboard. | Certified metrics, a shared glossary, semantic layer, lineage. | Data / analytics lead |
| Control | Risky actions | An agent emails a customer, edits a record, or leaks a row nobody approved. | Action permissions, human review ladders, audit logs, agent identity. | Governance / security |
| Cost | High bills | The pilot was cheap; the first production invoice ends the project. | Model routing, budgets per task, caching, cost-per-correct-answer. | Platform / FinOps |
| Choice | Lock-in | A price hike or model deprecation forces a rewrite. | Open formats, portable context, standard tool interfaces, multi-model evals. | Chief architect |
Where does your team stand?
Twelve questions, three per C. Nothing is saved or sent anywhere.
0 of 12 answered — finish all twelve to see your score.
Questions about the 4 C's
What are the 4 C's of enterprise AI?
Context, Control, Cost, and Choice. Context is what the AI knows about your business. Control is what it is allowed to do and how you prove it. Cost is whether each task pays for itself. Choice is your freedom to change models, tools, and vendors.
Why use a framework instead of comparing models?
Models change every few months; the four C's do not. A project that is weak on any one C fails in production regardless of which model it uses, so the framework survives model churn.
Which C should a team invest in first?
The one whose time horizon is biting you. A regulatory deadline means Control. A usage spike means Cost. Scaling a flagship use case means Context. A vendor renewal means Choice.
How do I score my team on the 4 C's?
Answer three questions per C as No, Partial, or Yes. The self-score on this page does it in about two minutes and shows your weakest C. Re-score every quarter.
Who is the 4 C's framework for?
Data engineers, AI engineers, analytics engineers, architects, platform and governance teams, and data leaders who need a shared language for designing and funding trusted AI systems.
Where can I learn the 4 C's in depth?
The Context Advantage is a 36-chapter living book by Team BricksNotes that covers each C with stories, architecture patterns, checklists, and anti-patterns. Chapters 1 to 3 are free to read.
Go deeper than a scorecard.
The Context Advantage covers each C across 36 chapters — stories, architecture patterns, anti-patterns, and checklists. One payment, lifetime updates.