Build vs. Buy in the Age of AI: Why the Old Rules Are Wrong

The standard build-vs-buy framework is taught in every business school and repeated at every strategy offsite. It goes something like this:

1. If the capability is core to your competitive advantage, build.
2. If it's non-core, buy.
3. Adjust for speed to market and internal capability.

This framework was correct when "build" meant $500k, nine months, and a permanent engineering team. It is no longer correct.

What changed

Three assumptions underneath the old framework have quietly broken:

1. "Build is slow." A focused product now ships in 8–12 weeks, often faster than enterprise SaaS implementations.
2. "Build is expensive." A flat $8–15k/mo managed engagement ships four products a year. That's less than one Salesforce seat per day, for four custom apps a year.
3. "Build requires a team." One senior engineer with AI assistance now does what a three-person team did in 2023.

Every one of those bullets compressed the "build" cost by about 10x in under three years.

The updated framework

The question is no longer "is this core?" It's:

1. Is the workflow specific to how we work? If yes, buying forces you to change how you work. Building reflects it.
2. Is the data sensitive or proprietary? If yes, putting it on someone else's multi-tenant cloud is a structural risk that only gets worse over time.
3. Do we expect the workflow to evolve? If yes, SaaS is a frozen snapshot of "best practice three years ago." Custom evolves at your speed.
4. Is there a credible commodity SaaS with 90%+ fit? If yes, buy it (think: Stripe, GitHub, Google Workspace). If not, build.

This framework produces different answers than the old one, especially for lower-mid-market firms. Most of what you were told to buy — CRM, matter management, intake portals, dashboards, client portals — lives in the "specific to how we work + we expect it to evolve" quadrant. Those are build-now items under the new math.

What you should still buy

Buy commodity infrastructure: email delivery, authentication, payments, storage, observability. These are table stakes where you get crushed trying to roll your own.

Buy content platforms and collaboration (Notion, Slack, Linear) where the network effect is the product.

Buy everything else at your own peril. The delta between renting a mediocre fit and owning a perfect one has never been smaller.

The decision test

For any software line item: if you canceled it tomorrow, how long until your firm had a better version of it — owned, custom, and fitted to your actual workflow?

Three years ago: forever. Today: one quarter.