Buy vs. Build After AI: What's Still Worth Owning - Part 2

markdown*Post 2 of 4 in the **Building With LLMs** series. [See all posts](blog-series-index.md). Previous post: [Should You Build a SaaS in 2026?](blog-series-1-should-you-build.md)*

---

The first post in this series asked the seller's question: should you build a SaaS at all in a world where AI generates code for free? This post asks the same question from the other side. The buyer's calculus is shifting just as fast, and most teams haven't sat down and re-thought what counts as commodity now versus what's still worth paying for.

The buyer's math used to be straightforward. "Pay $200/month for a niche tool" vs. "spend a quarter and three engineers building it ourselves" — the buy answer won almost every time, because the build answer was slow, expensive, and risky. AI didn't kill that calculation, but it changed the inputs enough that the answer flipped for whole categories of tools. A weekend with Claude can produce something that genuinely competes with a third-tier SaaS on functional grounds. For a lot of internal tools — admin panels, simple scrapers, basic dashboards, lightweight CRMs, scheduled job runners — the build option is now real on day one in a way it wasn't eighteen months ago.

The honest version of the new question isn't "build or buy?" It's *"do I want to be in the business of maintaining this for the next five years?"*

## What didn't get cheaper for the buyer

The build is the easy 20%. The other 80% is still real labor: security patches when a CVE drops in a dependency you'd forgotten about, edge cases your customers find that you didn't, on-call when it breaks at 3 AM the night you wanted to sleep, integrations that need updating every time an upstream API changes, the long tail of "weird but legitimate" requirements from real users. AI helps with each of these tasks individually but doesn't make any of them go away. A "$10 in tokens" build often becomes a $10,000-a-year maintenance burden that nobody put in the budget.

There's also the trust and compliance dimension. SOC 2, GDPR, HIPAA, audit logs, single sign-on, SAML for the enterprise customers, role-based access for the team — these are unglamorous but they're real, and they take longer than the headline feature you actually wanted. If the SaaS you're considering already does them and you'd otherwise have to build them, the price you're paying is mostly for that.

## What's commodity now, what isn't

The line of "what counts as buildable" has moved a lot in a year. Things that were obviously buy-not-build are now reasonably buildable: internal admin tools, basic CRMs for small teams, simple analytics dashboards, scrapers for known sites, light workflow automation, internal style guides, internal documentation systems. If your team has any AI fluency at all, you can ship a credible version of these in a week.

Things that look buildable on day one but are still buy-not-build: anything that touches money (payment processing, invoicing, subscription billing — the regulatory tail is enormous), email infrastructure (deliverability is a multi-year reputation game), identity (auth done badly is a security incident waiting to happen), reliability at scale (it's easy to get to 99.9%, hard to get to 99.99%), and anything where vendor lock-in to specific compliance certifications is the actual product.

The middle is where it gets interesting and where most people make the wrong call. Tools that are *almost* commodity — they look simple, you could build a working version in a sprint, but the production version requires handling 50 edge cases your team will discover one by one over two years. Examples: PDF generation that works across every corner case, calendar invites that don't break in Outlook, file uploads that survive every weird mobile browser, search that's actually good. The "we'll just build it" version of these usually ships, then quietly rots, then gets quietly replaced eighteen months later by a SaaS subscription anyway.

## What this means if you're selling

The implication for SaaS sellers is sharper than for buyers. "We built a hard thing" is no longer defensible by itself, because the customer can build a 70%-as-good version in a weekend. What still defends:

Operating the thing well so the customer doesn't have to — reliability, scale, on-call, the boring infrastructure of always being up. The customer doesn't want to be in this business. That's the offer.

Data and network you can't get on day one — competitor benchmarks, aggregated usage signals, integrations that took years of relationship-building to wire up, partnerships your customer can't replicate. This is the strongest moat AI doesn't touch.

Distribution and trust you've earned — being the obvious choice when someone asks. Being SOC 2 audited so the procurement team doesn't have to ask twice. Being the brand a CFO recognises when the renewal lands on her desk.

Support that customers don't want to do themselves — the version of your product where someone picks up the phone matters more, not less, when the alternative is "you maintain this yourself." Self-service has been winning for a decade, but at the high end of the market that trend is reversing as customers realise they don't actually want to be the on-call team for their own infrastructure.

Feature parity is no longer a moat. It was a moat when shipping a feature took six months. When it takes six hours, the only sustainable answer is to compete on the things that take six years.

## The new heuristic

If you're a buyer, here's the question I'd actually ask: *will this become a core part of how my company is different from its competitors, or is it shared infrastructure that everyone needs and nobody differentiates on?*

Build the differentiated thing. Buy the commodity. The line between those has moved — what counts as commodity now extends much further than it used to, and you should re-audit your stack with that in mind. But the principle hasn't changed. The mistake to avoid is the middle: building things that are commodity to you (you'll regret the maintenance), or buying things that are core to you (you'll regret the dependency).

If you're a seller, the question is the inverse: *am I a commodity, or am I differentiated?* If you're a commodity, AI is going to compress your margins toward zero — get to scale, get to operational excellence, or get out. If you're differentiated, AI is the best gift you've ever received, because it lowers the cost of building the next ten things on top of your moat. The hard part is being honest about which one you are.

---

## Coming next in the series

If you've read this far and you're building with LLMs, the next two posts are about the mechanics. **[Post 3: Tokens and Temperature, in Plain English](blog-series-3-tokens-and-temperature.md)** covers the two dials that nearly every LLM cost and quality issue traces back to — short, foundational, the vocabulary you need before the playbook makes sense. Then **[Post 4: Twelve Steps to a Cheaper, Better LLM App](blog-series-4-twelve-step-playbook.md)** is the tactical playbook itself.

← Previous: [Should You Build a SaaS in 2026?](blog-series-1-should-you-build.md) · [Series index](blog-series-index.md) · Next post → [Tokens and Temperature, in Plain English](blog-series-3-tokens-and-temperature.md)