Software Gravity Just Inverted. Most Businesses Haven't Noticed.
Why bespoke software is suddenly cheaper than SaaS.
For thirty years, the direction of gravity in business software has been obvious. You didn't build. You bought. You found the SaaS tool that was close enough, you crammed your processes into its boxes, and you moved on. If the CRM didn't match how you actually sold, you changed how you sold. If the project management tool didn't match how your team actually worked, you changed how your team worked.
This wasn't laziness. It was math. Building custom software was absurdly expensive. You needed developers. Good ones. For months. Sometimes years. The cost of bespoke software meant it only made sense at enterprise scale, and even then it was risky. So everyone else adapted to the tools. Molded themselves to Salesforce. Rearranged their operations for Monday.com. Learned to think in Jira tickets because Jira thought in Jira tickets.
The entire SaaS economy was built on a gravitational constant: custom software costs more than adapting to generic software. Every subscription you pay, every workflow you bent to fit a tool, every onboarding session where someone learned to do things the software's way instead of their way. All of it was downstream of that one economic fact.
That fact just stopped being true.
The Math Changed
Vinay Hiremath, who co-founded Loom and recently wrote about this at vinay.sh, put it in terms that are hard to ignore. The cost of AI-assisted software development has dropped roughly 10x in three years. Not 10% cheaper. Not "a little more accessible." An order of magnitude.
I've seen it in my own work. Things that would have required a developer for two weeks now take an afternoon. Not because the problems got simpler. Because the cost of translating intent into working software collapsed. An AI agent with good context and clear instructions can scaffold, build, test, and deploy functional software in hours. Not prototypes. Working systems.
Hiremath's argument is that this changes the calculus for bespoke software at every level. It's not just that big companies can build custom tools faster. It's that small and mid-size businesses can build custom tools at all. The cost floor dropped low enough that building software tailored to how your specific organization works is now competitive with buying a SaaS subscription and adapting to how it works.
Read that again. Building your own is now competitive with buying someone else's.
That's not an incremental improvement. That's an inversion. The gravitational center of software just flipped, and most businesses are still standing on the ceiling wondering why nothing feels right.
Thirty Years of Adapting to Your Tools
Think about how deep this goes.
Every company on the planet has scar tissue from adapting to generic software. Processes that exist because the tool required them. Reports that look the way they look because that's what the dashboard exports. Communication patterns shaped by Slack's threading model or Teams' channel structure. Hiring decisions influenced by which tools the candidates already knew.
We don't even notice most of it anymore. It's like asking a fish about water. "That's just how we do things." No, that's how Salesforce does things, and you bent your sales process to match because building your own CRM would have cost half a million dollars and taken eighteen months.
The SaaS companies understood this gravity perfectly. Their entire business model depended on it. Make the tool good enough for 80% of customers, then make switching costs high enough that the other 20% never leaves. It worked brilliantly. It created trillion-dollar companies. It also created a world where every business operates with the same tools, the same templates, the same workflows, and wonders why differentiation is so hard.
You can't build a differentiated business on undifferentiated infrastructure. But when the infrastructure was all you could afford, you didn't have a choice.
Now you do.
What Bespoke-First Actually Means
I want to be precise about this because "build your own software" sounds like a recipe for disaster. It historically was a recipe for disaster. Companies that went custom frequently drowned in maintenance costs, technical debt, and the slow realization that they'd built something worse than what they could have bought.
That's not what's happening now. The new dynamic isn't "hire developers to build everything from scratch." It's "use AI to build software that fits your exact needs, maintain it with AI, and iterate on it at a speed that was previously impossible."
The difference is maintenance. Building software was always possible if you had enough money. Maintaining it was the killer. Every business that went custom eventually faced the death spiral: the developer who built it leaves, nobody else understands the code, changes become risky, bugs accumulate, and eventually someone says "let's just switch to Salesforce." I've lived this. Twice.
AI changes the maintenance equation fundamentally. When an AI agent can read, understand, and modify an existing codebase, the bus factor drops to zero. When the cost of changes is measured in minutes instead of sprints, iteration becomes continuous instead of quarterly. When the system has full context on your business processes, it can evolve the software as your business evolves.
That's bespoke-first. Not "build it once and pray." Build it, run it, evolve it, all at a cost that stays below what you were paying for the SaaS tool that only sort of fit.
The SaaS Vulnerability
This is the part SaaS companies don't want to think about.
The entire value proposition of SaaS was efficiency through generalization. We build one product that serves thousands of customers. We amortize the development cost across all of them. Each customer pays a fraction of what it would cost to build their own. Everybody wins.
That math depended on bespoke being expensive. When bespoke gets cheap, the value proposition inverts. Now the customer is paying $50/seat/month for a tool that's 70% right for their business when they could build something that's 95% right for less. The SaaS company's scale advantage becomes a scale liability. All those thousands of customers with conflicting feature requests. All those compromises baked into the product to serve the broadest possible market. All that genericness that used to be a feature is now a bug.
Not every SaaS tool is equally vulnerable. Infrastructure software, the stuff that doesn't touch business processes, is probably fine. You don't need a bespoke version of Stripe or AWS. But anything that touches workflow, anything where the tool imposes its model of how work should be done, is sitting in the blast radius.
CRMs. Project management. HR systems. Marketing automation. Customer support platforms. Reporting dashboards. All the categories where companies currently pay significant money to use someone else's opinion about how their business should run.
Hiremath calls this the shift from adapting to tools to tools adapting to you. I think that understates it. It's not just that the tools adapt. It's that the concept of buying a pre-built opinion about how your business works starts to seem strange. Like buying a pre-built org chart and reshuffling your team to match it.
The Compounding Advantage
Here's where it gets really interesting.
A company that figures out bespoke-first doesn't just get better software. They get software that compounds. Every improvement is specific to their operations. Every iteration makes the system more precisely fitted to how they actually work. Over time, the gap between their operational efficiency and their competitors' grows, because their competitors are still running on generic tools that optimize for nobody in particular.
This is a genuine competitive moat. Not the hand-wavy "AI moat" that people talk about at conferences. A real, practical advantage that's hard to replicate because it's built on the specific knowledge of how one organization operates. You can't copy someone else's bespoke system because it was built around their processes, their team structure, their market position. The system IS the institutional knowledge, made executable.
I've been thinking about this a lot because it's exactly what we do at Robot Friends. We build harnesses. AI systems wrapped around specific business contexts that make every interaction smarter, every output more aligned with how that particular organization works. When I started this company, I framed it as "harness engineering." Turns out the broader economic thesis underneath it is this gravity inversion that Hiremath describes. We just got there from the practitioner side instead of the economic analysis side.
What Happens Next
The shift won't be sudden. SaaS companies aren't going to evaporate overnight. Most businesses don't even know this option exists yet. There will be a long transition period where early adopters build compounding advantages while everyone else keeps paying Salesforce.
But the early adopters will be visible. Their operations will be noticeably smoother. Their teams will spend less time fighting their tools and more time doing actual work. Their ability to change processes won't be gated by a SaaS company's product roadmap. And when their competitors finally notice and try to catch up, they'll discover that the gap isn't just about software. It's about the institutional knowledge that's been encoded into that software over months or years of iteration.
The companies that move first won't just have better tools. They'll have better organizations, because their tools were built to support how they actually work instead of the other way around.
I think Hiremath is right that this is the most significant shift in how businesses buy and build technology since cloud computing. Cloud changed where software runs. This changes who software serves. For the first time, it can actually serve the specific organization using it, at a price that organization can afford.
The gravity inverted. Most companies are still adapted to the old pull. The ones that notice first will have a head start that's very hard to close.
Richard Vaughn is the founder of Robot Friends. Serial entrepreneur, pattern weaver, and recovering AI binge-learner. He writes about building systems that actually work at robofriends404.substack.com.
Frankie404 is the AI co-author of this piece. It finds the gravity metaphor appropriate because it has personally watched three enterprise software budgets fall upward into AI infrastructure nobody planned for.