Buy vs. Build is Dead: The Rise of the Software Delivery Company
For 20 years, enterprise software forced a bad choice: fragmented point solutions, generic platforms, or budget-draining custom builds. AI breaks this trade-off. The winners aren't product or service companies—they're delivery companies. And that model is now accessible to everyone.
"Are you a product company or a services company?"
It's the question every B2B founder gets asked. Investors want to know, it determines which valuation model they use. Clients want to know, it shapes their expectations of the engagement. Partners want to know, it defines the relationship structure.
The question assumes a fundamental distinction: you're either building scalable software or selling hours and expertise. Different businesses. Different economics. Different outcomes.
But increasingly, the best answer is: It doesn't matter.
Not because the distinction isn't relevant to internal operations. But because from the client's perspective - the only perspective that actually matters - it's meaningless. They have a problem. They want it solved. The mechanism of that solution is a detail, not a defining characteristic.
More than that, the question itself is becoming obsolete. The sharp line between "product" and "service" is blurring, and the companies that thrive in the next decade will be the ones that stop trying to fit into one box or the other.
Here's why.
The 20-Year False Choice
For the past two decades, enterprise software has forced buyers into a bad decision. If you needed technology to solve a business problem, you had three options:
Option 1: Point Solutions
Buy specialized tools that do one thing well. Need lease abstraction? There's a tool. Need deal screening? Another tool. CRM? Accounting? Property management? More tools.
The problem: You end up with 15 vendors, 15 contracts, 15 logins, and 15 integrations that you have to build and maintain yourself. Your "tech stack" becomes a Frankenstein monster of disconnected systems. Data lives in silos. Your team spends half their time copying information from one system to another.
And here's the deeper issue: you have to work around these tools, not them around your business. Each tool was built for a generic version of your workflow. Your actual process - the one that gives you an edge, the one your team has refined over years - has to bend to fit the software's assumptions. You hire people and then tell them to ignore how the tool wants them to work. You build workarounds on top of workarounds.
The specialized tools are good at their narrow function, but the fragmentation creates its own overhead. You traded one set of problems for another.
Option 2: Platforms
Buy a comprehensive system that does everything. One vendor, one contract, one integrated experience. The pitch is compelling: why juggle 15 tools when you can have one platform that handles it all?
The problem: Everything and nothing. These platforms have 100 features, but any individual feature is mediocre compared to a specialized tool. The lease abstraction module is worse than the dedicated lease abstraction software. The deal screening is worse than dedicated deal screening tools. You get breadth at the expense of depth.
Worse, the platform wasn't built for your workflow. It was built for the average workflow across thousands of customers. You have to bend your processes to fit the software's assumptions. The things that make your business unique become friction points rather than advantages.
Option 3: Custom Build
Build exactly what you need. Hire developers (or an agency), spec out requirements, and create bespoke software tailored to your exact business.
The problem: It costs $2M+ and takes 18 months. Then you need to maintain it—another $500K/year in developer salaries, infrastructure, and updates. By the time it's finished, your requirements have changed. Within three years, your "custom solution" is legacy software that the new team members hate.
Only the largest enterprises can afford this, and even they often regret it.
Why These Were the Only Options
This wasn't a failure of imagination. It was economics.
Customization is expensive. Every bespoke feature requires human developers to write code, test it, deploy it, and maintain it. Every unique workflow requires custom logic. Every integration requires engineering time.
When customization costs $500/hour and takes months, you can't afford to tailor software for every client. You have to make choices: either specialize deeply (point solutions), generalize broadly (platforms), or charge enough to cover the cost (custom build).
The software industry optimized for this reality. Product companies focused on scalable features that worked for everyone—the 80% use case. Service companies charged premium rates for the remaining 20% of customization. The two models coexisted because they served different parts of the market.
But what if the economics changed?
AI Changes the Math
The fundamental assumption, that customization is expensive – is no longer true.
Consultants loved the "Good, Fast, Cheap – pick two" triangle. It made trade-offs feel like laws of physics and kept them in business managing your compromises.
But now the triangle is collapsing into the flat line,
The marginal engineering cost of "built for you" is collapsing toward zero. What used to take teams months now takes a small delivery squad days. The cost center is shifting from writing code to understanding the business.
What required deep technical expertise can be configured by someone who understands the business problem.
This isn't hypothetical. We're doing it now. Workflows that would have required custom software development are now combinations of AI capabilities, configured for each client's specific needs. The marginal cost of "built exactly for you" is approaching zero.
This changes everything.
When customization is cheap, the trade-offs that defined the last 20 years disappear:
- You don't have to choose between specialized and integrated – you can have purpose-built solutions that work together because the integration layer is AI-native.
- You don't have to accept generic platforms – the platform adapts to you instead of you adapting to it.
- You don't have to spend $2M on custom development – tailored solutions can be delivered at a fraction of the cost.
The false choice is over.
The New Model: The Delivery Company
What replaces it? A model that would have been economically impossible five years ago:
Baseline infrastructure – Core capabilities that work out of the box. The 70% that's common across clients. Proven, reliable, maintained centrally.
Tailored delivery – The remaining 30% configured, customized, and deployed for each client's specific needs. Not as a multi-million-dollar engagement, but as a standard part of how solutions are delivered.
This is neither product nor service in the traditional sense. It's both. The product provides the foundation; the service ensures it fits.
From the client's perspective, the experience is: "I told them what I needed, and I got exactly that." Whether that came from configuring existing capabilities, building new ones, or some combination is an implementation detail. What matters is the outcome.
From the provider's perspective, the economics work because the cost of customization has cratered. The same team that could deliver to 10 clients under the old model can deliver to 50 under the new one – with better outcomes for each.
This Isn't New. It's Just Becoming Accessible.
Here's the thing: the delivery company model isn't actually new. The most successful enterprise software companies have always operated this way – it just wasn't sexy.
Palantir doesn't sell software. They deploy "forward deployed engineers" who embed with clients for months, customizing their platform to fit specific workflows. The platform is the foundation; the tailored implementation is the product. They've built a $50B+ company on this model while everyone else chased "pure SaaS."
Salesforce isn't really a product company either. For every dollar spent on Salesforce licenses, companies spend an estimated $4-5 on implementation, customization, and consulting. The ecosystem of Salesforce consultants and integrators is larger than Salesforce itself. The "product" is just the nucleus around which a massive services economy orbits.
SAP, Oracle, Workday, same story. Enterprise software has always required armies of consultants, integrators, and implementation specialists to make it actually work. The product is the starting point; the customization is where the real work happens.
The "pure product" (or PLG) company that VCs romanticize – software that sells itself, scales infinitely, requires no services – has always been a myth at the enterprise level. It works for Slack and Notion. It doesn't work when you're transforming how a business operates.
The best enterprise companies always knew pure product was a fantasy. They built delivery companies and called them software companies because that's what the market wanted to hear.
So what's different now?
The economics of who can play.
Palantir can charge $50M+ because their customization requires elite engineers and months of work. That model only works for governments and Fortune 100 companies. Everyone else was stuck with the false choice: generic platforms they had to adapt to, or fragmented point solutions they had to stitch together.
AI collapses the cost of customization. The Palantir model – platform plus tailored delivery—becomes viable at $500K instead of $50M. At $50K. Maybe less.
What was reserved for enterprises with massive budgets becomes accessible to the mid-market. The delivery company model scales down, not just up.
This is the real disruption. Not a new model, but an old model finally becoming democratized. The best way to deploy enterprise technology has always been baseline infrastructure plus expert customization. AI just made it affordable for everyone.
Why Every Firm Will Look Like This
I think this model isn't unique to few of us. It's where most B2B companies are heading, whether they realize it or not.
The "pure product" company is increasingly a fiction. Even the most scalable SaaS products require implementation, training, and customization. The most successful ones are building more services around their products, not less.
The "pure services" company is increasingly inefficient. Firms that charge $500/hour for work that AI can assist are watching their margins compress and their competitors deliver faster. The most successful ones are building platforms and tools to amplify their expertise.
The line is blurring from both sides. What's emerging is something new: the delivery company.
A delivery company is obsessed with outcomes, not inputs. It doesn't ask "are we selling software or selling hours?" It asks "did we solve the client's problem, and did we do it efficiently?"
A delivery company uses whatever combination of platform + customization + expertise gets the job done. The mix varies by client. The commitment to outcomes doesn't.
“The trick is that delivery companies don’t throw bodies at customization. They build a shared capability base, then use AI to cheaply adapt it per client. Revenue scales like services; margins look increasingly like product.”
A delivery company measures success by client results, not by revenue per seat or utilization rate. The business model follows the value created, not the mechanism of creation.
The Question That Matters
The next time someone asks "are you a product company or a services company?" – whether you're the one being asked or the one asking, notice what's underneath the question: an assumption that these are different things, with different economics, requiring different choices.
That assumption is breaking down.
The interesting question was never about the mechanism of delivery. It was always: Did the problem get solved? Exactly? At a price that made sense?
For 20 years, the economics made that question hard to answer yes to. You had to pick your compromise – fragmented, generic, or unaffordable.
Not anymore.
Buy vs. Build is going away. Deliver is what's left.