Choosing the Wood

Opening Scene

A furniture maker chooses their wood deliberately, based on the specific piece’s needs — strength, weight, cost, and how it’ll actually be used — not by default habit. The wrong material choice, made carelessly at the start, undermines everything built on top of it. Choosing the underlying model foundation for an internal AI tool deserves this same deliberate, upfront consideration.

In Plain English

Selecting a foundation model for an internal tool means weighing the same considerations covered throughout this content library’s generative AI series: whether a general-purpose frontier model, a fine-tuned model covered in this content library’s fine-tuning-versus-prompting series, or a small language model covered in this content library’s dedicated series best fits the specific task’s complexity, cost, and deployment requirements. This decision shapes everything the internal tool will build on top of.

The Old Way

Before this deliberate model selection process was widely applied to internal tool projects specifically, teams sometimes defaulted to whatever model was most familiar or available:

• Internal tool teams sometimes defaulted to whichever model was most familiar or readily available, without a genuine, deliberate comparison against the specific task’s requirements.
• There wasn’t yet a well-established practice of applying the fine-tuning-versus-prompting and small-language-model decision frameworks specifically to internal tool foundation selection.
• Model selection was sometimes made once, early, without revisiting it as the tool’s actual usage patterns and requirements became clearer over time.

A genuine, deliberate model selection process, applying the frameworks covered throughout this content library, reflects a maturing understanding of how much this choice shapes an internal tool’s eventual success.

What’s Changing (and Why AI Is the Reason)

1. Internal tool teams increasingly apply the decision frameworks covered in this content library’s fine-tuning-versus-prompting and small-language-models series deliberately, rather than defaulting to a familiar model.
2. This connects directly to the cost considerations covered throughout this content library’s generative AI series, since foundation choice directly determines an internal tool’s ongoing operating cost.
3. As internal tool projects mature, model selection is increasingly revisited periodically as actual usage patterns become clearer, rather than fixed permanently at the start.

The Metaphor, Fully Extended

The Custom Furniture Maker	Model Foundation Selection Concept
Choosing wood deliberately based on the piece’s specific needs	Choosing a model deliberately based on the tool’s specific requirements
Strength, weight, and cost all factoring into the decision	Capability, cost, and deployment requirements all factoring into the decision
A careless choice undermining everything built on top	A careless choice undermining everything the tool builds on top of
Not defaulting to whatever material happens to be on hand	Not defaulting to whatever model happens to be most familiar

Quick Recap

• Selecting a foundation model for an internal tool means weighing capability, cost, and deployment requirements deliberately.
• This connects directly to the decision frameworks covered in this content library’s fine-tuning-versus-prompting and small-language-models series.
• This choice shapes everything the internal tool builds on top of.
• Model selection should be revisited periodically, not fixed permanently at the start.

Where This Fits in the Series

Article 4 covered choosing the underlying model foundation. Article 5 turns to a joint that holds without nails: how the different pieces of an internal tool’s architecture fit together.