/ Article

Before we get into it, a quick note on where this paper is coming from.

zu builds custom software. Has for thirty years. That means we have a financial interest in the argument we’re about to make, and you should know that. What we also have is thirty years of watching these decisions play out in organizations where the stakes are high; utilities, mining operations, insurers, government bodies. The builds that compounded in value and the ones that became maintenance problems, the SaaS subscriptions that were exactly right and the ones that became the most expensive line item nobody looked at.

A Shift is Happening

For the first time in roughly twenty years, the assumption that subscription software is the right default answer for enterprise technology deserves a serious second look. That’s not a contrarian position for its own sake. The economics that made SaaS the obvious choice for two decades were real; faster deployment, lower upfront cost, someone else managing the infrastructure. Most organizations made that bet repeatedly, and for most of that period, it was the right call. 

Whats changed is that the three pillars holding that logic up are all under pressure at once.

The cost of building has fallen faster than most organizations have recalculated. AI-assisted development is delivering production-grade systems at timelines and costs that weren’t possible two years ago. The gap between a subscription and a build is narrowing, and most technology strategies haven’t caught up with that yet.

SaaS pricing has moved in the opposite direction. Vendors are bundling AI features customers didn’t request, restructuring contracts to reduce renewal leverage, and raising prices at nearly five times the general inflation rate. The software isn’t getting better at the same rate it’s getting more expensive.

Vendor stability is no longer a safe assumption. SaaS M&A hit a record in 2025, with private equity driving more than half of all acquisitions. The pattern is consistent: acquire a product your operations depend on, raise prices, reduce support, extract margin from customers who can’t easily leave. The organizations most exposed are the ones who never expected to need an exit.

This paper is for the leaders asking whether the assumptions behind their current stack still hold.

A Pattern Worth Recognizing

For decades, cable was the deal nobody loved but everyone accepted. You paid for 200 channels and watched 12. The bundle was bloated, the pricing opaque, and the contract locked you in whether the service improved or not. Then streaming arrived and the logic seemed undeniable: pay only for what you watch, cancel anytime, no more subsidizing channels you never chose. Millions of households cut the cord. The model felt like a correction, a more rational, consumer-friendly way to pay for content.

Then something familiar happened. Netflix raised prices. Then again. HBO launched its own service. Then Disney. Then Paramount. Then Apple. Each new platform launched with content you couldn’t get anywhere else, which meant cancelling wasn’t really an option if you wanted to stay current. Before long, the average household was paying more per month than they ever paid for cable, spread across five or six platforms, spending ten minutes deciding what to watch before giving up and rewatching something they had already seen. The bundle is back, with more logins, and advertising.

The streaming story didn’t fail because the idea was wrong. It failed because the economics of content ownership eventually reasserted themselves. Convenience attracted customers. Lock-in extracted value from them. And the promise of simplicity gave way to a complexity that looked remarkably like what people had left behind, only more fragmented, more expensive, and harder to untangle.

Enterprise SaaS followed the same arc, on a longer timeline and with higher stakes. The early promise was genuine: faster deployment, lower upfront cost, someone else handling the infrastructure. Organizations subscribed. Then they subscribed again. And again. Vendors raised prices, bundled features nobody asked for, and structured contracts that made leaving painful. The average large organization now runs over 100 SaaS applications. Nearly 60 percent of those licenses go underused. And the integration debt between them has created exactly the kind of fragmented, expensive, hard-to-untangle complexity that SaaS was supposed to eliminate.

This is not a story about SaaS being a mistake. It is a story about a cycle that most industries eventually complete: the initial promise, the adoption, the accumulation of hidden costs, and the moment when enough has changed that the original assumptions deserve to be examined again. That moment, for enterprise software, is now.

The following sections walk through each dimension in turn: operational fit, financial investment, strategic differentiation, and risk, with the questions worth asking at each one.

Operational Fit

At some point, most organizations stop asking whether their software fits how they work and start building processes around what their software allows. Workarounds become procedures. Procedures become policy. And the gap between how the organization actually operates and how it was designed to operate widens until nobody remembers which came first.

SaaS platforms are built for the broadest possible market. For horizontal functions like email, messaging, document sharing, and video conferencing, the standard tool is the right tool. There’s no competitive reason to build what everyone else is already using.

The calculation changes when the function is specific to how you operate. The closer your workflows are to the median, the better SaaS fits. The further they are, the more you’re paying to approximate something that a system built around your operations would do without compromise.

The organizations that feel this most acutely are rarely the ones that made bad decisions. They’re the ones that made reasonable ones, repeatedly, over a long period of time. Each workaround made sense when it was built. Each process that bent to accommodate the software seemed like a small compromise at the time. The compounding is invisible until a failed implementation, a key person leaving, or a new regulatory requirement the software can’t accommodate forces the question. By that point the gap between how the organization works and how it was designed to work has been accumulating for years. A new platform won’t close it. The work is organizational now.

THE QUESTIONS TO ASK:

• How well does the software fit your specific workflows, or are your processes standard enough that an off-the-shelf platform serves you well?
• How many of your current operational processes exist because that is genuinely the best way to run your organization? How many exist because that is what your software allows?

Strategic Differentiation

Every organization in your industry has access to the same platforms, the same vendors, the same feature sets. When your technology is a subscription, your competitors can match it by the end of the quarter.

The organizations pulling away in specialized markets made a different bet, often without framing it as a strategic decision at the time. They stopped treating technology as infrastructure to be managed and started treating it as capability to be built. Not all at once. Usually one system, in one part of the operation, where the fit between what the software could do and what the organization actually needed had degraded past the point of workarounds. They built something instead, learned from it, and made it better on their own roadmap. And then they looked at the rest of their stack and started asking the same question everywhere else.

What accumulates over that process isn’t just better software. It’s organizational knowledge encoded into systems that reflect how the organization actually operates at its best. The exception handling that took years to get right. The workflow built around a regulatory requirement specific to your jurisdiction. The decision that gets made consistently under pressure because the system was built around how your organization actually thinks. None of it appears in a vendor’s feature list. It lives in what they built. A competitor can subscribe to the same platform you’re on. They can’t subscribe to ten years of operational intelligence encoded into something you own.

THE QUESTIONS TO ASK:

• How important is it that your technology reflects what makes your organization distinctive, or is the software a utility that doesn’t need to differentiate you?
• Is the way you serve your customers something your competitors could replicate by subscribing to the same platform?

Financial Investment

The reason most organizations didn’t build was simple: it was too expensive, timelines were long, and the execution was high-risk. In the last two years those variables have changed. The front-loaded investment that made custom development economically unviable has come down. The subscription costs that made SaaS the obvious alternative have gone up. And the mechanisms driving those increases are structural enough that they won’t reverse on their own.

The subscription model that made SaaS so appealing was built on a simple promise: predictable costs, software that kept getting better, and vendors who needed to earn your renewal. That’s not the deal most organizations have anymore.

The Changing Economics of SaaS

Price inflation is the most visible problem. SaaS pricing is increasing at nearly five times the general inflation rate of G7 countries, and switching is painful enough that most customers absorb the increases rather than fight them. Microsoft’s July 2026 update makes the point concretely: E3 and E5 list prices rise 5 to 8 percent, but the removal of Enterprise Agreement volume discounts in November 2025 pushes the effective increase for large enterprises closer to 20 percent. Same vendor, same product, materially higher bill.

AI bundling is making it harder to see. According to Vertice’s 2025 SaaS Inflation Index, 60 percent of vendors deliberately mask rising prices through AI bundling. Features (A.I specific or not) get added on the vendor’s timeline, whether the customer asked for them or not, and the price goes up to match.

Credit systems are the newest mechanism, and difficult to track. As AI agents take on work that previously required human seats, vendors are repricing toward consumption and outcome-based models. They sell subscriptions with a set number of credits, then reserve the right to change how many credits each action costs. No price increase notice, no contract amendment. Just an erosion in what the spend actually buys.

For organizations that chose SaaS partly because the costs were foreseeable, that’s a materially different deal than the one they signed.

THE QUESTIONS TO ASK:

• Are your SaaS costs rising faster than the value they deliver?
• Do you know how your vendors are pricing AI features, and whether you’re paying for capabilities you aren’t using?
• If you projected your current subscription costs forward five years, would the cumulative spend justify a purpose-built alternative?

Risk and Control

Every software vendor relationship carries some level of dependency. The question is how much, and what happens if the terms change. For most of the SaaS era, that risk was manageable. Vendors competed for customers, pricing was relatively stable, and switching cost was a negotiating chip rather than a trap. For many platforms that is still true.

Vendor dependency has always been a consideration. What has changed is the scale and speed of it. SaaS M&A hit a record in 2025, with more than 2,500 transactions and private equity driving more than 57 percent of all acquisitions. The playbook is consistent: acquire a product your operations depend on, raise prices, reduce support, and extract margin from those who have nowhere else to go.

At the same time, the economics of SaaS itself are shifting in ways that directly affect the organizations running on it. In the first week of February 2026, over $1 trillion in market capitalization was erased from software stocks in seven days as markets priced in the impact of AI agents on seat-based business models. Vendors under pressure to restructure cut support, accelerate product changes, and make decisions that serve their transition, not their customers. The disruption is no longer isolated to acquisition targets and underfunded startups, it is happening across the industry simultaneously.

Operational Exposure

Support is usually the first thing that gets cut. After Broadcom acquired VMware, it reduced the combined workforce by more than 22,000 people, including senior architects from core platforms. Customers reported slower response times, less experienced support engineers, and tickets being closed without resolution. AT&T eventually sued Broadcom over support continuity, claiming it was being forced to pay for software it didn’t need just to maintain access to support it had already paid for. What had been a vendor relationship became a legal dispute.

Roadmap divergence follows close behind. The platform you bought is rarely the platform you end up on. Vendors rebuild around AI, reposition toward larger markets, and retire the features that smaller customer segments depend on — all without your input. End of life announcements and feature deprecations happen on the vendor’s timeline, not yours. And the roadmap itself is shaped by the customers willing to spend the most. For a mid-sized insurer, a government ministry, or a regional mining operation, that is never going to be you. The features that get built are the ones the largest banks and the largest enterprises asked for. Everyone else gets what they get.

There is also the gap between the sales promise and the product reality. According to a Software Finder survey of enterprise software users, only 28 percent said their software exceeded what was promised during the sales process. The demo runs on clean data in a controlled environment. The reference customers are the top implementations, not the median. And the roadmap features that closed the deal have a way of arriving late, arriving differently, or not arriving at all.

Sovereignty Exposure

For Canadian public sector and regulated industry organizations, data sovereignty has moved from a compliance consideration to an active geopolitical risk. The US CLOUD Act permits American authorities to demand access to data held by US-headquartered companies regardless of where that data is physically stored, a structural incompatibility that a Computer Weekly investigation confirmed no major hyperscaler can technically resolve.

The US government is now using trade policy to compound that pressure. The 2026 National Trade Estimate Report on Foreign Trade Barriers explicitly targets Canada’s Sovereign Cloud Initiative, signaling that efforts to move Canadian data beyond American legal reach will face active resistance. Most Canadian data requirements were written before any of this was true. Revisiting them against the actual architecture of tools currently in use is, for most organizations, overdue.

Strategic Exposure

The concentration risk extends beyond any single vendor relationship. In July 2024, a single CrowdStrike software update took down 8.5 million Windows devices simultaneously, grounding airlines, shutting down hospitals, and disrupting financial institutions worldwide. No organization could prevent it, control it, or set the timeline for recovery. Attackers and system failures go where the scale is. Two decades of consolidation have placed critical infrastructure in the hands of a small number of vendors, and custom software doesn’t carry that exposure.

Data portability is worth understanding before you need it. Which means before a shutdown is announced or a price increase makes leaving the most practical option. Most organizations sign SaaS contracts without fully assessing how their data is stored, what format it lives in, or what a migration would actually require. Proprietary formats, years of accumulated configuration, and complex integrations make switching more expensive and time-consuming than the original evaluation accounted for.

THE QUESTIONS TO ASK:

• Do you lack access to your own data, customization, or performance control?
• Would losing access to this platform disrupt your operations?
• How easily could you migrate your data and move to an alternative if you needed to?

Where to Start

The build-vs-buy question has never been a single decision. It’s a portfolio of decisions, made at different times, under different conditions, by different people, and rarely revisited with the same rigour that went into making them the first time. The organizations that get this right aren’t the ones with the most sophisticated technology strategies. They’re the ones who treat the question as an ongoing one, who know what they’re paying, what they own, and which platforms their operations couldn’t survive without.

That clarity is rarer than it should be. If you want to know where your stack stands, we built something for that. [zu.com/assessments/saas-vs-custom]