Too Big To Care
This week at work, I was deeply amused when a request to evaluate a random SaaS app (as in, spin up a proof of concept before formally reviewing the app) included the plan to hook it up to our live data in Dynamics 365. Because "we won't get a good test otherwise".
Needless to say, I had...opinions. Thoughts. Even qualms.
That's a nice business process you got there. Be a shame if something happened to it...
Anyway, rather than expressing my feelings in some antisocial and career-threatening fashion, I started looking into the app in question. The first thing I saw on their landing page caused my eyebrows to make a valiant effort to reach my hairline (a tall order, given how much no-man's-land of forehead has to be crossed in order to do so).
Their call to action spake thusly:
D365 keeps changing. Your business still needs to work.
Think about that sentence for a moment. Really savor it. Compare it to, say, advertising copy for disaster insurance. Then ask yourself why we're pretending that D365 changes are as uncontrollable as the weather.
Microsoft itself says that after each Dynamics 365 update, customers must ensure their solutions still work as expected, and recommends automated regression testing to reduce the effort involved.
And then there's the Microsoft Regression Suite Automation Tool (RSAT), Microsoft's own tool for automating some of that regression testing. It is now marked for deprecation, with support ending May 15, 2027.
So customers who invested in Microsoft's tool to manage the risk created when Microsoft changes Microsoft must now manage the change created when Microsoft changes the tool for managing Microsoft changes.
You couldn't make the example much more recursive if you tried.
Microsoft 365 reflects the same philosophy at vastly greater scale. Microsoft's own documentation describes an environment of "continuous change" in which IT administrators must continually assess impact, prepare users, test workflows, monitor results, and adjust controls and guidance.
Microsoft is even changing how customers manage that change. Its new release model uses Frontier, Standard, and Deferred audiences, with Targeted Release being retired. Deferred Release provides some control, but for eligible features it generally means postponing their arrival for about 30 days, not deciding whether to accept the change at all.
And that distinction matters. A lot. There's a world of difference between controlling change and merely being given time to prepare for it. One of those is a steering wheel; the other is noticing the turn signal on the truck you're being towed by.
We have normalized the idea that a customer may need to buy another SaaS product to protect itself from changes made by its SaaS vendor. Not to enhance or augment. Just to try to detect failures in our processes caused by changes we didn't approve.
There's an entire industry built on this problem. And, in my not-so-humble opinion, the fact that it needs to exist is the real problem.
Have I got a deal for you...
SaaS was supposed to free customers from maintaining software. PaaS and IaaS extended the same idea further down the technology stack. And, for the most part, the bargain has been enormously successful.
But somewhere along the way, another part of that bargain changed. We didn't just outsource operating the technology. Increasingly, we surrendered control over when and how the technology changes.
Squint a little, and the modern cloud bargain looks something like this:
- The vendor owns the platform.
- The vendor controls the roadmap.
- The vendor decides when it changes.
- The customer owns the consequences.
Now, before anyone fires up the comments section (which I don't have, so...good luck with that): my beef isn't that cloud platforms change. Of course they do. They have to. Vulnerabilities need patching, capabilities need to evolve, and nobody seriously wants Microsoft 365 frozen in amber at its 2018 feature set.
The real question is who carries the cost and risk of all that continuous improvement. And today, the answer is increasingly "the customer". The vendor gets continuous delivery; the customer gets continuous change management. Lucky us.
Too big to care
That asymmetry only gets more important as the dependency compounds. Microsoft 365 starts innocently enough: email and Office. Then collaboration. Then identity. Device management. Security. Automation. Business applications. Data. AI. And if you're really committed, Azure can extend that dependency right down to the infrastructure your products run on too. It's turtles all the way down, and every turtle sends an invoice.
Granted, any one of those decisions may be perfectly rational on its own. Microsoft frequently offers an excellent product, compelling integration, and economies of scale that nobody else can touch. But collectively? Something changes.
"Don't like it? Use something else" is not meaningful recourse when switching means unwinding identity, endpoints, collaboration, business processes, integrations, security controls, compliance evidence, institutional knowledge, and years of accumulated engineering. That's not a migration; that's a full-body transplant. At that point, a "strategic partnership" starts looking uncomfortably like the absence of any credible alternative.
And that's where too big to fail becomes too big to care.*
To be clear, I'm not suggesting that Microsoft is twirling its mustache and deliberately mistreating its customers. It doesn't need to. It merely needs to become large enough, and its customers dependent enough, that the disruption imposed on any one customer matters less than Microsoft's incentive to keep the platform moving. No malice required...just math. (Cory Doctorow has a much more colorful name for where that math eventually leads: enshittification).
So what's the rational response? It isn't to abandon Big Cloud, buy a bunch of PowerEdge servers off eBay, and start racking them in a closet. It's to stop treating further consolidation as automatically desirable.
The next time a Microsoft product promises tighter integration, lower administrative overhead, or One More Pane of Glass™, maybe the discussion shouldn't end with whether Microsoft can provide it. We should also ask whether concentrating yet another critical function with the same vendor increases our ability to control our own technology environment...or decreases it.
And when someone inevitably asks "What else can we put in their cloud?", somebody needs to ask right back:
Would we still sign up for this if we could actually leave?
Because if the honest answer is "no", that's not a partnership anymore. That's Stockholm syndrome with an enterprise agreement. (No offense, Stockholm. Some of my favorite people are there.)
In fairness to Redmond
Before anyone sharpens their Copilot-logoed pitchforks: Microsoft is the example here because it's the one I live in all day, not because it's uniquely guilty. It's not even the worst offender. Not by a long shot.
Broadcom bought VMware and promptly rewrote the licensing rules for a whole lot of data centers. Atlassian retired its Server products and pointed everyone toward its cloud or a pricier Data Center license. Google has retired so many products that tracking the bodies has become a hobby for some folks. Pick a giant, and you'll find customers who woke up one morning to discover their roadmap had been rewritten for them.
And I'm hardly immune. This very site runs on Cloudflare Pages, it gets built and deployed by GitLab, and my homelab leans on 1Password for every secret it touches. I've made the same bargain, just at a much smaller scale (and with much lower stakes if one of them decides to change the rules on me). The difference is that I could still walk away from any one of them in an afternoon...well, a long weekend. Okay, a long weekend and a lot of coffee.
So no, this isn't really a Microsoft problem. It's a scale problem. The bigger the provider and the deeper the dependency, the less any single customer's pain registers. The questions above apply to every vendor you can't easily leave; Microsoft just happens to be the basket where many of us keep the bulk of our eggs.
Oh, and that "proof of concept" on live data? Still not important right now. Ask me again after it's been through a proper review. #TestDataOrGTFO
* Full disclosure: I didn't coin "too big to care", and I had no idea it was already in use until I started writing this post. Back in 2020, EU Commissioner Thierry Breton said there was a feeling in Brussels that online platforms had become "too big to care". Then-FTC chair Lina Khan picked it up on The Daily Show in 2024 as a telltale sign of monopoly power, and last year she aimed it squarely at Microsoft after a Game Pass price hike. I'm just late to the party, as usual.