Should the company buy an enterprise AI copilot or build its own assistant?
Short answer
Buy for broad productivity over standard workplace content where the vendor integrates natively and governance meets requirements; build (on managed AI platforms) for domain-specific workflows, proprietary data sources, custom actions, special security or cost needs. Many enterprises do both under a common governance and evaluation framework.
Detailed answer
| Factor | Buy (e.g. Microsoft 365 Copilot, Gemini for Workspace, Glean) | Build (Bedrock/Vertex/Foundry + RAG + tools) |
|---|---|---|
| Use case | General productivity (email, docs, meetings) | Domain workflows (claims, underwriting, support) |
| Data | Standard SaaS sources | Proprietary systems, structured data |
| Speed | Fast | Months |
| Control | Limited | Full (retrieval, models, UX) |
| Cost model | Per seat | Per usage + engineering |
| Differentiation | None | Possible |
Decision: pilot bought copilot with measured productivity; build targeted assistants where value is domain-specific; share governance, identity and evaluation practices.
What the interviewer expects
Use-case-driven decision, cost models, governance consistency.
Common mistakes to avoid
All-or-nothing decisions; ignoring data permissions readiness (oversharing in content repositories).