Almost never a model built from the ground up — that question usually turns out to be about a different, narrower layer than most businesses actually need to touch.
Short answer
No, not in the sense of training a foundation model from scratch — that is a distinct, specialized undertaking almost no business actually needs. What most people mean by “needing our own AI” is fine-tuning, prompting or retrieval on top of a model someone else already built, which is a far smaller commitment.
If the real question is an off-the-shelf tool versus a custom-built application, that is a more common decision and a different one — see build vs buy: choosing between an off-the-shelf tool and a custom solution. This question sits one layer beneath that: even a fully custom-built application almost always runs on top of someone else’s underlying model.
Canada’s federal, provincial and territorial privacy commissioners define this role precisely in their joint generative AI guidance, distinguishing “Developers and Providers” — those who “develop (including training) foundation models or generative AI systems, or that put such services onto the market” — from “organizations using generative AI.”
Training a foundation model from the ground up puts a business in the first category. It is a resource-intensive, specialized undertaking that the overwhelming majority of AI-using Canadian businesses never actually attempt; most sit in the second category, working on top of a model someone else already built and validated.
The same joint guidance tells organizations to weigh whether a generative AI use is “necessary and proportionate” to the purpose at hand, and to evaluate “the validity and reliability of the generative AI tool for the intended purpose” before reaching for more than the task requires.
Applied here, the practical test is not whether a business could build something more powerful. It is whether an existing general-purpose model, worked on top of, already does what the task in front of you actually needs — and for almost every Canadian business asking this question, the answer stops there, well short of training anything from scratch.
See how to choose between an off-the-shelf tool and a custom solution, one layer up from this question.
