We just wrapped a sprint, and something caught my attention that I haven’t been able to stop thinking about. We blew past our planned story points, not by a little, by a meaningful margin. The team had been leaning heavily on AI coding agents throughout the sprint, and the productivity difference was impossible to ignore. It got me thinking about what this kind of output means beyond our own projects, specifically what it means for the economics of enterprise software. Not in theory. In actual dollars.
I spend a lot of time working with companies on their software needs. At Superior Technologies, we provide development consulting teams that help organizations build and deliver the software that drives their business. So when I started pulling at this thread, it wasn’t purely academic. I was looking at it through the lens of what we see every day: companies spending enormous amounts of money on platforms they only partially use, while the cost of building purpose-fit alternatives keeps dropping.
What Enterprise Platforms Actually Cost
Let me put some numbers on the table, because this conversation tends to stay vague when it shouldn’t. For a 500-person company, annual licensing alone looks like this:
- ServiceNow (ITSM, HR, Workflows): $150K–$400K/year
- Salesforce (CRM): $200K–$600K/year
- Workday (HCM/Finance): $50K–$200K/year
- SAP S/4HANA Cloud (ERP): $300K–$800K/year
- Microsoft Dynamics 365 (ERP+CRM): $150K–$350K/year
Scale that to a 100,000-employee enterprise and you’re looking at $10M–$60M per platform, per year. Those aren’t outliers, those are standard contract ranges. And that’s before you spend a dollar implementing any of it.
The Hidden Cost Nobody Budgets For
Here’s the thing every CIO knows privately but few organizations account for properly upfront: the licensing fee is rarely the actual cost. Gartner’s widely cited rule of thumb is that for every $1 spent on SaaS licensing, organizations spend $3–5 on implementation, integration, and customization. Panorama Consulting Group’s annual ERP report found that implementation costs routinely run 2x–6x the license cost, with the average ERP project coming in at $3.7 million over 17+ months. McKinsey’s 2020 research on large IT transformation projects found they run 45% over budget on average and deliver 56% less value than predicted. The people doing that configuration work? Primarily Big 4 firms like Accenture, Deloitte, PwC, and KPMG, billing somewhere between $200 and $500 per hour.
So the math isn’t: “We’re paying $1M a year for ServiceNow.” The math is closer to: “We paid $1M for licenses, another $2–4M to stand it up, and we’re recurring on that license fee every year with built-in escalators.” ServiceNow builds 7–9% annual price increases directly into multi-year contracts, that’s documented in their SEC filings. Salesforce raised prices an average of 9% in August 2023. Morgan Stanley’s 2023 CIO Survey found that CIOs expected software vendor price increases averaging 5–8% per year across the board. And once you’re three or more years into a platform, Bain & Company’s 2022 research found that switching probability drops below 10%. The lock-in isn’t accidental. It’s the model.
What You’re Actually Using
This is the part that should bother everyone more than it does. Pendo’s 2019 Feature Adoption Report found that only about 20% of software features get used regularly. The Standish Group’s CHAOS Report reached a similar conclusion, 20% of features are used “always or often,” while 45% are never used at all. Forrester estimates that 25–35% of enterprise SaaS spending goes toward underutilized licenses and features.
Think about what that means in practice. You’re paying for a massive, complex platform. You’re using roughly a fifth of it. You’re paying consultants to configure the fifth you use. And you’re doing this year after year, with pricing that goes up every year, on a platform you can’t realistically leave. I’m not here to say ServiceNow or Salesforce are bad products. They’re not. They’re enormous platforms with hundreds of capabilities, and for some organizations, especially those actually leveraging the breadth, the economics make sense. But most companies aren’t those companies. And from what I’ve seen working with clients across industries, most organizations know this. They just haven’t had a viable alternative.
The Calculation That Actually Changed
Here’s where this stops being a familiar rant and becomes something I think is genuinely new. The build-vs-buy argument has existed for decades. And for most of that time, “build” lost for a simple reason: upfront cost and time. You needed 15–20 developers, 18 months of runway, and the organizational stomach to absorb risk before the software delivered value. By the time you recouped costs, you’d spent years paying for something that wasn’t ready yet.
That calculus has changed. Not because software got easier to build in the abstract, it didn’t. But because AI coding agents have fundamentally changed how fast a skilled team can move. What I saw in that sprint isn’t anecdote. It’s the shape of what development looks like now when you have the right people with the right tools. A team of experienced developers equipped with AI agents can realistically deliver the specific features a company actually uses from a platform like ServiceNow, in less time than a year, for less than what they’re currently paying in annual licensing. Not the entire ServiceNow catalog. Not every feature of a Fortune 500’s Salesforce instance. The subset of features the client actually uses. The 20%.
And at the end of it, they own it. No annual escalators. No lock-in. No paying for features they’ll never touch. The decision about what to spend going forward is theirs: keep the full team to expand the platform, scale down to one developer for maintenance, or anything in between. That kind of flexibility simply doesn’t exist in the enterprise licensing model.
The Objections That Are Worth Taking Seriously
Of course this isn’t a clean argument with no counterpoints. Let me name the real ones.
Small subscriptions don’t justify a development team. True. If a company is paying $30K/year for a handful of Salesforce seats, no development team makes sense at that scale. This argument lives at meaningful spend levels, call it $500K+ annually before it starts to make economic sense. Enterprise software comes with enterprise support, compliance, and reliability guarantees. Also true. Building in-house means owning uptime, security, compliance certifications, and disaster recovery. That’s real overhead, and it’s not free. Organizational knowledge and change management are hard. Switching off a platform your team has used for years isn’t just a technical problem. It’s a people problem. Bain’s research on switching probability reflects this. It’s not just vendor lock-in technically, it’s organizational inertia.
These are legitimate. I’m not dismissing them. But I think they’ve been used to close off a conversation that deserves to stay open, especially now. The right development partner can help navigate these challenges, the compliance requirements, the migration planning, the change management, so that an organization isn’t taking on all of that risk alone.
The AI Add-On Problem Makes the Timing Urgent
There’s one more layer to this, and it’s accelerating. Every major platform vendor is now bolting AI capabilities onto their existing products and pricing them separately, aggressively, on top of already significant license costs. Salesforce’s Einstein 1 editions carry a $150–200/user/month premium over non-AI equivalents. ServiceNow’s Now Assist reportedly adds an estimated $30–50/user/month, and CEO Bill McDermott has stated on earnings calls that it adds 20–30% to contract values. Microsoft 365 Copilot costs $30/user/month on top of existing licenses, an 83% increase over a base E3 subscription.
Here’s the other side of the coin: the AI capabilities being sold as premium add-ons by these platforms are, in most cases, available to a development team for a fraction of the cost through direct API access. You’re not paying $150/user/month for something magical, you’re paying for the wrapper, the brand, and the convenience of it being inside the platform you’re already locked into. The irony is that the AI revolution is simultaneously making it more expensive to stay on these platforms and dramatically more feasible to build alternatives to them.
A Question Worth Asking
I’m not arguing that every company should immediately rip out their enterprise platforms and build replacements. That’s not realistic, and it’s not what I’m saying. What I am saying is that the conversation has changed, and most organizations are still having the old version of it. The off-the-shelf vs. build-in-house debate used to be a straightforward risk calculation that almost always favored buying. That’s no longer obviously true, particularly for companies spending significant money on platforms they’re partially using, with AI add-on costs stacking on top, and annual escalators baked into contracts.
The question worth asking, seriously, is whether the 20% of a platform you actually use could be built and owned for what you’re paying to license the other 80%. For some companies, the answer is still no. But for more companies than are currently asking the question, I think the answer is yes. And if you’re a company sitting on a seven-figure annual platform spend wondering if there’s a better way, it’s a conversation worth having. We have it with companies all the time.
Am I missing something here? What would it take for companies to actually start making this shift? I’d genuinely like to hear from people who’ve been on both sides of this decision.
