Most agencies are paying for software they could now build themselves. Not because anyone made a bad decision – but because “buy it” has always been the default, and almost nobody has updated that default since AI changed what building actually costs.
This post covers why the build-versus-buy line has shifted, the real cost of a subscription that has nothing to do with the monthly fee, how to tell which tools are worth keeping and which you should own, the psychological trap of building things you shouldn’t, and how the Technology pillar of the STANDOUT framework helps you audit your stack with intent rather than inertia.
The Build-Versus-Buy Line Just Moved
A few days ago I cancelled a SaaS subscription I’d been paying for – a scorecard tool I used as the front door to my AI audit. Then I rebuilt the entire thing myself: the questions, the scoring logic, the back-end that emails each person their report and logs the lead into my own database. It took an afternoon working alongside Claude, and it now runs on my own domain, with my own data, for nothing a month.
The point of that story isn’t the saved fee – it’s what the saved fee revealed. For years, the honest answer to “should we build this or buy it?” was almost always buy. Building meant a developer, a budget, a timeline and a maintenance burden most agencies couldn’t justify for anything that wasn’t core. AI broke that assumption without most people noticing. The cost of building a simple, genuinely useful tool has collapsed – things that needed a development team eighteen months ago can now be assembled by someone who understands the problem and can direct the AI well. The market moved the line between build and buy, and the reflex most agency owners use to decide hasn’t caught up.
“Every subscription you keep should be a decision, not a default.”
The Real Cost of a Subscription Isn't the Subscription
The monthly fee is the smallest part of what a tool costs you. Look across an agency’s stack and the individual subscriptions all look cheap – twenty here, fifty there, a couple of hundred for the one nobody remembers signing up for. The problem is never a single subscription; it’s the sprawl, and what the sprawl quietly does to the business underneath it.
Across the audits I’ve run, the hidden costs show up in the same places every time: your data ends up scattered across platforms you don’t control, in formats you can’t easily export; your team’s knowledge lives inside tools, so leaving one means losing the workflow around it; you rent capabilities close to your core offer, so part of your differentiation depends on someone else’s roadmap and pricing; and nobody owns the stack as a whole, so cost and overlap grow faster than the value they add. That last point is the one that should worry you most. When a capability close to your edge is something you rent, you’ve handed a piece of your competitive position to a vendor who can raise prices, change features or shut down. The better question is rarely “is this tool worth the money?” – in isolation it almost always is – but whether the capability should live inside your business or outside it. That one barely gets asked.
When You Should Build, and When Buying Is Still Right
This is where the message gets misread, so let me be blunt: the lesson is not “build everything now.” That’s just swapping one lazy default for another, and it’ll cost you more than the subscriptions ever did. The decision is a judgment call – AI widens the range of things you could sensibly build, but it doesn’t tell you which of them you should.
Build it yourself when the tool sits close to your differentiation or IP and owning it strengthens what makes you distinct; when the data should live with you, not a vendor; when the problem is contained enough that AI genuinely closes the gap; and when you can maintain it without it becoming a second job. Keep buying when it’s critical infrastructure where security, compliance or uptime carry real risk; when the problem is complex or fast-moving; when the maintenance burden would outweigh renting; or when a mature product already does it better and it isn’t central to your edge. Build the things that are core, contained and worth owning. Buy the things that are complex, critical and not your fight. The mistake most agencies make isn’t picking wrong on a single tool – it’s never running the decision at all.
The Trap: Building Things You Shouldn't
Now the counterweight, because this is where my psychology background makes me suspicious of my own advice. The moment building gets cheap, a new failure mode appears: building things you have no business building. Two well-documented patterns make this almost inevitable.
The first is the IKEA effect – we overvalue what we build ourselves, even when the bought version is better and cheaper to run, which quietly pushes you to keep and defend internal builds long past the point they earn their place. The second is what I call procrastinating via automation: building an internal tool feels like progress, but for plenty of agencies that energy would have done far more good aimed at client delivery or new business. So the same shift that lets you cancel a subscription and own your tool can lure you into a graveyard of half-finished builds nobody maintains. Cheaper building doesn’t remove the need for judgment – it raises the stakes on it. AI executes; you still have to decide.
The Bottom Line
The way out of both traps – over-buying through inertia and over-building through enthusiasm – is to stop treating each tool as a one-off and start governing your stack as a whole. In the STANDOUT framework that’s the Technology pillar: what you own versus what you rent, and whether each capability is worth building, buying or cutting. Standstill agencies accumulate subscriptions by inertia and can’t tell you what the stack costs or what half of it does. STANDOUT agencies treat it as a set of deliberate decisions, know exactly what they own and why, and use AI to bring the right capabilities in-house where ownership creates a genuine edge.
Frequently Asked Questions
Should agencies build their own AI tools or buy them?
Build when the tool is close to your differentiation, the data should be yours, the problem is contained enough that AI closes the gap, and you can maintain it. Buy when it’s critical infrastructure, complex, or outside your core. Most agencies err by never asking the question and renting by default.
Has AI really made building software cheaper than buying it?
For simple, well-defined tools, yes. AI has collapsed the cost of building things that used to need a developer, so capabilities you’d have rented can now be owned by someone who understands the problem and can direct the AI. It hasn’t changed the maths for complex or critical infrastructure.
What’s the real cost of too many SaaS subscriptions?
It’s rarely the monthly fees. The real cost is data scattered across platforms you don’t control, workflows locked inside tools, differentiation that depends on a vendor’s roadmap, and a stack nobody owns where overlap grows faster than value.
When should an agency keep paying for a SaaS tool?
Keep buying when the tool is critical infrastructure, the problem is complex or fast-moving, the maintenance burden of owning it would outweigh renting, or a mature product already does it better and it isn’t central to your edge.
