1
Curious builder0 XP earned · 300 to level 2
0 daysFinish a lesson to begin
Badge collection0 of 6 unlocked
51 small wins to finish your pathNext question →
Your company debates building its own LLM stack vs buying a vendor platform (or using a single model provider end to end). How do you guide the decision?
30-second answerSay your answer out loud first, then reveal.
Decision framework
| Factor | Lean "buy" | Lean "build" |
|---|---|---|
| Differentiation | Commodity (chat UI, generic tracing) | Core to the product's value |
| Time to market | Urgent | Can invest months |
| Scale economics | Low or medium volume | Very high volume (unit costs matter) |
| Customisation | Standard workflows | Unusual requirements |
| Compliance | Vendor certified for needs | Strict residency / air-gap |
| Team | Small, product-focused | Strong platform/ML infra team |
| Vendor risk | Acceptable | Single point of failure, pricing power concerns |
Lock-in mitigation
- A gateway abstraction over model providers.
- Store data, prompts, evals and traces in your own systems (portable).
- Use open standards: OpenTelemetry for traces, MCP for tools, OpenAI-compatible APIs.
- Run evals regularly on alternative models so switching is a known quantity.
- Avoid deep coupling to proprietary orchestration features unless they give clear value.
Reassess over time: what's "build" today (e.g. evaluation tooling) may be "buy" next year as the market matures, and vice versa at scale.
Interview signal. A nuanced, layered answer rather than "always build" or "always buy", plus concrete lock-in mitigations.
Related
You understood something today that you didn't yesterday.