"Should we build our own AI or buy a SaaS tool?" is the wrong first question. Asked that way, leadership ends up making a single all-or-nothing bet on a category of decisions that should never be made at the category level.
The framework we use
In every AI consulting engagement we run, we split the AI surface area into four tiers and decide separately on each:
1. Commodity capability. Speech-to-text, translation, generic OCR, basic summarization. These are commodities. Buy. Use the cheapest API that hits your accuracy threshold and move on. There is no defensible moat in building this yourself.
2. Differentiated workflow on commodity capability. A clinical note summarizer for nurse practitioners. An invoice classifier that has to match your chart of accounts. A customer support agent trained on your knowledge base. Build the workflow, buy the model. The value is in the system around the model, not the model itself.
3. Proprietary data advantage. Predictions or generations that meaningfully improve when you fine-tune or embed against your private data. Build, with a serious data plan. This is where bespoke AI earns its keep — and where pilots usually fail because the data plan was an afterthought.
4. Frontier research. Pushing the state of the art in a research-heavy way. Hire a specialized team or partner with a lab. Almost no operating business actually needs this, and the ones that do usually know it.
Why this works
The framework forces leadership to confront where the actual moat is. You quickly discover that 70 percent of your "AI initiatives" are tier-1 commodity work you should just buy and move on from, 20 percent are tier-2 workflow plays that justify a real engagement, and only 5 to 10 percent are tier-3 data plays worth deep investment. Tier-4 almost never shows up.
If the AI conversation in your organization keeps stalling because the answer keeps being "well, it depends," it usually does — on which tier the specific use case lives in. Sort that out and the rest of the strategy stops being a debate.