Last week we talked about bespoke software and whether it has to be out of reach for smaller businesses. This week we pick up the thread we deliberately put a pin in: what even is enterprise — and do you actually need it?
Spoiler: probably not as much of it as you’ve been quoted for.
The conversation kicks off with a real one Matt had recently — someone pushing back on Azure with “Facebook runs on PHP and MySQL, so why can’t we?” It’s a fair question with an uncomfortable answer: Facebook doesn’t run MySQL, it runs Facebook’s MySQL, maintained by tens of thousands of engineers. You don’t have that. The $30–$100/month floor on Azure isn’t a tax — it’s a team of Microsoft engineers doing the boring infrastructure work you’d otherwise have to do yourself.
From there we get into the meat of it:
- What “enterprise” actually means — and why our working definition is internal systems for larger organisations, even though that definition leaks the moment you poke it
- Why the reference architectures Microsoft published for SMBs in the 2000s never matched the businesses we were actually working with
- The all-nines availability trap — and the difference in cost, effort, and engineering maturity between 99.9% and 99.999%
- The CIA triad: why availability is negotiable, but confidentiality and integrity shouldn’t be — even for the smallest shop
- Surface area vs. blast radius — two lenses for thinking about risk that don’t require enterprise budgets to apply
- Why “the server under John’s desk” can genuinely be a more secure option than a half-baked cloud deployment
- How enterprise-level engineering costs get baked into bespoke software quotes — even when the customer doesn’t need them
- The squeaky-wheel CEO problem: how a single bad experience at the wrong moment triggers a million-dollar re-architecture for a system that was working fine
- Why “nobody gives a f#ck about your code” still applies — end users don’t care if it’s enterprise ready, they care that it works when they need it
- And the AI angle: if a developer’s answer to “can you get me this Excel report?” is “sure, that’ll be an n-tier architecture and six months,” Claude is going to eat that work for breakfast
The point we keep landing on: there are a lot of dials you can turn on a software solution — scale, uptime, redundancy, perimeter security, geographic availability — and every one of them costs money. Enterprise customers turn them all to 11 because they need to. Smaller businesses don’t, and the job of a good consultant is to ask which dials actually matter for this customer, this problem.
If you’ve ever looked at a quote for bespoke software and thought “there’s no way that’s worth it” — there’s a decent chance a big chunk of that number was for things you don’t need.
We close the loop on last week and tee up the next question: what changes when the software is consumer-facing? That’s a different conversation. Tune in next time.
🍻 Tonight’s Drinks
Matt – Oakvale Shiraz 🍷 Liam – Glandore TPR Tempranillo (Hilltops) 🍷
Both wineries are in the Hunter Valley. We should probably get back up there.
🔗 Tonight’s Links
- Last episode: The Bespoke Software Myth
- The callback we made: Nobody Gits a Fork About Your Code
- Matt’s upcoming talk at /NEW conf — “Everything is Too Hard” (Thursday, just after lunch)
Any Likes 👍, Shares 📣, Subscriptions 🔔, and Love ❤️ go a long way to helping us keep doing this for fun.
Cheers! 🍻