We're About to Cancel Our Software Subscriptions. We're Building Our Own Everything Now.

‍

Over $20,000 per year

‍

That's how much our 12-person creative agency spends annually on software. We'd never actually added it up before. Seeing the number felt like opening your credit card statement after a vacation you don't fully remember.

‍

Last Tuesday, we built our PM tool replacement in six hours. The cancellation is coming. And it won't be the last.

‍

The Meeting

‍

It started at our Monday standup. "Standup" is generous—we all sit, because we're creatives and we reject arbitrary corporate rituals. Also, standing is uncomfortable.

‍

One of our designers was clicking through PM tool trying to find a task she'd updated yesterday when the interface reorganized itself in that special way software does when it's had an "improvement."

‍

"Where did they move the dependencies view?"

‍

"Under the three dots now. Or maybe settings. I don't remember."

‍

This is what we pay over $300 a month for. Software that rearranges the furniture while we're trying to work, like a roommate with boundary issues.

‍

"Why don't we build our own?" I said.

‍

The room went quiet. The kind of quiet that happens when someone says something so stupid everyone's trying to find a polite way to ask if you're having a stroke.

‍

"With Claude," I said.

‍

"That Claude thing you won't shut up about?" our designer said.

‍

Fair.

‍

The Inventory

‍

Before doing anything rash, we looked at what we were actually paying for:

‍

Our PM tool — $25/user/mo → $3,600/year. We use maybe 60% of it. The other 40% is features we navigate around like furniture in a dark room.

‍

Slack — $8/person/mo → $1,152/year. Staying. It's where work happens and where we send each other memes at inappropriate times.

‍

Harvest — $12/person/mo → $1,728/year. Time tracking. Replaceable.

‍

QuickBooks — $90/mo → $1,080/year. We use 20% of its features. The team is weirdly excited about building our own.

‍

Figma — $45/editor/mo (6 editors) → $3,240/year. Not going anywhere. You can't rebuild collaborative design software on a Tuesday.

‍

Google Workspace — $12/person/mo → $1,728/year. Not going anywhere. Google Suite is the closest thing to perfect software that exists.

‍

Adobe Creative Suite — $85/person/mo (8 people) → $8,160/year. Still required. Still irreplaceable. Still making us pay monthly for software we used to own.

‍

Total: over $20k/year.

‍

Six Hours

‍

We gave ourselves one afternoon. If we couldn't build something functional by 5 PM, we'd accept that our PM tool $3,600/year was the price of not thinking about project management infrastructure.

‍

We opened our IDE coding terminals. The team gathered around like we were conducting a séance.

‍

What happened next was the closest I've come to a religious experience without involving psychedelics.

‍

It just started building. Not based on what some PM best practices guide says our workflow should be, but based on how we actually move projects from pitch to delivery.

‍

Do you track design revisions separately from dev tasks?

‍

Yes. And our PM tool makes that weirdly hard.

‍

How do you handle client feedback cycles?

‍

With a custom field and a lot of manual updates. It's annoying.

‍

Want me to automate that?

‍

Three hours in, we had something functional. Six hours in, we had something custom-built for us. Views organized by client. Automatic time estimates based on similar past tasks. A client feedback loop that actually matched how we work.

Our team leaned back. "I don't know whether to be impressed or terrified."

‍

"Both is fine. Both is appropriate."

‍

‍

The Domino Effect

‍

If we could replace our PM tool in an afternoon, what about Harvest?

‍

Ninety minutes. Working prototype. Simple interface, tracks time by project and task, integrates with our new PM system, exports to CSV for our accountant who still—bless his heart—prefers spreadsheets.

‍

Then someone suggested we build our own accounting software.

‍

"That's insane," I said.

‍

"Is it though? We do the same bookkeeping tasks every month. Categorize expenses. Track invoices. Generate reports. QuickBooks is bloated. We use maybe 20% of it."

‍

I opened my mouth to argue and realized I couldn't.

‍

The Travel Agent Analogy

‍

I have a rule: if you can't explain something to friends at dinner without them checking their phones, it's either boring or you don't understand it.

‍

Here's the dinner party version: Remember when you needed a travel agent to book flights? They knew the airlines, the cheapest routes, the airport codes. Real expertise. Then the internet happened and you could just... do it yourself. They protested. "We provide value! We know things you don't!" And they were right. But not $75-per-ticket worth of value.

‍

That's what's happening to SaaS companiesright now. All SaaSs are travel agents. Harvest is a travel agent. Half our software stack is travel agents. And Claude is the internet.

‍

Except this time it's faster. Because building software is now as easy as describing what you want in complete sentences.

‍

What's Staying, What's Going, For now..

‍

Staying: Figma ($3,240), Google Workspace ($1,728), Adobe ($8,160). Infrastructure plays. You can't prompt-engineer your way to professional design software. I asked Claude. It politely suggested I keep paying Adobe.

‍

Going (in the next few months): Our PM Tool ($3,600), Harvest ($1,728), QuickBooks ($1,080).

‍

That's $6,408/year we're about to save by spending a few weekends building exactly what we need. A 31% reduction in software costs. With an AI assistant we already had.

‍

This has made us insufferable at industry events.

‍

"Oh, you use Monday.com? That's nice. We built our own PM system. Took a weekend."

‍

Someone's going to punch me eventually. It'll be deserved.

‍

The Taste Argument

‍

Here's what keeps me from spiraling into existential panic about AI: taste is the final frontier.

‍

People don't pay Azai Studios for project management or time tracking or bookkeeping. They pay us for taste and experience building comapnies. To know that this shade of blue works better than that one. To konw what will acutally make money. To understand that this layout feels premium while that one feels cheap. To recognize when something is technically correct but aesthetically wrong, or won't move the bottom linen.

‍

Taste is judgment refined through experience. It's cultural context. It's knowing the difference between what works and what works for this specific audience, in this specific moment, for this specific purpose.

‍

Can AI help with execution? Absolutely. Can it know, instinctively, that a logo feels like it's trying too hard?

‍

Not yet.

‍

That "not yet" is doing a lot of work. But for now, taste is still human. And taste is what we sell.

‍

You can't vibe-code good taste.

‍

Who Survives This

‍

The people and companies that thrive through this shift share two things: taste and high agency.

‍

Taste tells you what to build. High agency means you actually build it …  that when you see a $3,600/year problem, you don't complain about it on LinkedIn, you solve it on a Tuesday afternoon. The people who treat obstacles as puzzles instead of walls. The ones who hear "that's not how it's done" and take it as information, not instruction.

‍

Most of the agency world is going to sit through another PM tool demo next quarter. They're going to renew their annual contracts. They're going to keep paying for software that doesn't fit how they work because switching feels hard and building feels impossible.

‍

It's not impossible anymore.

‍

We've also hedged our bets at Azai Studios by investing in companies and sectors that aren't caught in AI's blast radius… real estate, finance, physical goods. Businesses where the moat isn't software, it's atoms. If we're wrong about the pace of change, the downside is survivable. If we're right, we're positioned on both sides.

‍

The Uncomfortable Part

‍

Here's what keeps me up at night: if we can replace PM tool in six hours, what else becomes trivial?

‍

Not just software. What about the consultants who specialize in configuring these tools? The entire ecosystem of jobs that exist to manage, optimize, and train people on SaaS products that their customers can now build themselves?

‍

Every $50–300/month "nice interface for a database" company is now competing with the reality that their customers can build the same thing in an afternoon.

‍

Not all SaaS dies. Infrastructure plays survive. Network effects survive. Things requiring real scale, compliance, and deep expertise survive.

‍

But the middle layer? That's getting hollowed out. And most people don't see it yet.

‍

We Might Be Wrong

‍

There's a non-zero chance we're delusional. That in six months we'll sheepishly re-subscribe to our PM tool while our custom system collects dust like that exercise bike we definitely use.

‍

"These things need maintenance," our engineer reminded us. "What happens when something breaks?"

‍

"We fix it."

‍

"With what expertise?"

‍

"Claude's expertise. Plus our ability to describe what's wrong in increasingly frustrated language."

‍

He gave me a look that suggested I might be an idiot. He might be right.

‍

But here's the thing: PM tools breaks too. And when it does, we can't fix it. We just wait. And hope. And refresh the page.

At least with our own system, when something breaks, we can actually do something about it.

‍

The Bottom Line

‍

We don't know how this ends. Maybe with every agency building their own tools. Maybe with something weirder none of us can imagine.

‍

What we do know: we're canceling PM Tool in the next few months. Harvest is next. QuickBooks after that.

We're living through one of those moments that only makes sense in hindsight. Like the early internet, or the iPhone, or that weird year everyone realized working from home was actually possible.

‍

Except this time, we're paying attention.

‍

And we're building.

‍

Azai Studios is a Brooklyn-based creative agency that talks to robots more than is probably healthy and is about to cancel over $20k/year in software subscriptions. We will definitely bring this up at your next meeting.

‍

‍