Strategy

Build versus buy, honestly, in the AI era

Ilya MarchettiMay 21, 20268 min read

A framework for deciding when to buy the tool, when to build the system, and when to do nothing at all — including the total costs vendors never quote.

We are a build shop, so treat the following with appropriate suspicion. We also tell perhaps a quarter of the companies who approach us to buy something instead, because that is what makes the other three-quarters worth the money.

Buy when the workflow is common and non-differentiating

Transcription, scheduling, e-signature, expense capture, standard support macros. These are solved, competitively priced, and improving faster than you would. Buying is correct and building is vanity.

Build when the workflow is how you compete

If the process encodes something specific about how your business wins — pricing logic, underwriting judgment, clinical routing, supply allocation — a generic tool will force you toward the industry average. That is the actual cost of buying, and it never appears in the comparison.

The costs nobody quotes

Per-seat pricing that grows with headcount. Integration work that is your problem regardless. The migration cost when the vendor changes direction. The workaround spreadsheet your team maintains because the tool does not quite fit — that spreadsheet is a real line item.

Model those four and the comparison changes shape more often than people expect.

Do nothing when the process should be deleted

The best outcome of a diagnostic is occasionally the discovery that a workflow exists to produce a report nobody reads. Automating it would have been a well-engineered waste. Ask what the output is for before asking how to speed it up.

Next step

Let's find the work your team should stop doing.

A 30-minute discovery call. We come with questions about your process, not a deck about ours. You leave with two or three concrete automation candidates, whether or not you hire us.

Typical reply within one business day