The world already changed. Most of the software running your business didn’t.
Microsoft just handed us the clearest evidence yet that the enterprise software industry is splitting into two camps: those building AI into the foundation, and those charging you extra to staple it on top. And the staple camp is starting to look expensive, both financially and strategically.
The numbers tell the story before anyone has to editorialize. After three years of aggressive Copilot promotion, fewer than 4.5 percent of Microsoft 365 customers pay for it, and only about 1 percent use it weekly. That’s not a marketing problem. That’s a product-market fit problem wrapped in a pricing strategy that UK regulators are now formally investigating for potentially misleading customers into more expensive tiers they didn’t ask for.
Taking a step back, I think what we’re watching is the enterprise software equivalent of the moment when mobile apps stopped being “websites that also work on phones” and became native mobile experiences. The companies that rebuilt from the ground up won. The companies that bolted mobile onto desktop architecture spent years playing catch-up, and most never caught up at all.
The pattern is showing up everywhere, and it’s not subtle.
Microsoft raised prices as much as 43 percent on some SKUs, largely to bundle in AI features that most customers demonstrably don’t use. Total cost per user can now clear $35 monthly before you open a document. Google took the opposite approach—discontinued the standalone Gemini add-on entirely and folded AI into the base Workspace plan at $14 per seat. No toggle, no negotiation at renewal, no “Classic” tier hiding in the cancellation flow. One company treats AI as something you rent after buying the product. The other treats it as the product itself.
The Adobe-Figma saga makes the same point, except regulators did the analysis for us. Adobe couldn’t build what Figma built fast enough, so it tried to spend $20 billion to eliminate the competition. European and UK authorities blocked it explicitly because the deal “threatens to significantly affect competition in the market for interactive product design.” Adobe had its own competitor, XD, which it quietly shelved after announcing the Figma deal. A forty-year-old creative software giant looked at the fastest-growing product in its category and decided acquisition was faster than innovation.
Here’s what matters about Figma’s AI versus Adobe’s: remove the AI from Figma and the product fundamentally changes. Remove Firefly from Photoshop and you get the same Photoshop that shipped in 2019. That’s the whole test. If the AI is decoration, it can be removed without breaking the core workflow. If it’s the engine, the product stops functioning without it.
CRM is where this gets most expensive, fastest.
The SaaStr conference framework is blunt: the CRM market is being rebuilt from scratch, not incrementally. Their advice to buyers is “Follow the Agents”—pick the platform where AI agents actually live and do work, because every agent you deploy deepens switching costs. At ten agents deep, changing platforms becomes prohibitively expensive.
That framework cuts against both HubSpot and Salesforce equally, and it should. Both Breeze and Agentforce are AI reasoning layers sitting on database schemas designed fifteen to twenty-five years ago, before anyone imagined training models on that data structure. An AI wrapper reading a synced copy of your CRM data is fundamentally different from a system whose data model was built for native AI interaction. That sounds academic until you’re the one waiting for sync jobs to complete before AI can act on what just happened in your pipeline.
New entrants are already raising serious capital around this thesis. Monaco launched with $35 million from Founders Fund, built around AI agents supervised by salespeople rather than chatbots pasted onto pipeline views. Another AI-native CRM raised $30 million in a seed round that closed in 28 hours, positioned explicitly to replace the fifteen-tool sales stack most growth companies have accumulated.
I don’t believe legacy CRM disappears overnight. Large enterprises with deep Salesforce deployments and growing Agentforce agent libraries have real switching costs, and “Follow the Agents” actually favors Salesforce if that’s where you’ve already committed. But for companies choosing their stack today, the honest question isn’t which logo is more familiar. It’s whether the AI reads your data live or reads a copy three steps removed from where work happens.
What’s striking is how much of this mirrors patterns I’ve seen in other industries. A logo on a jersey is decoration. A trading card that is the experience is the product. The same distinction applies to enterprise software. Is the AI woven into the workflow, or is it standing next to the workflow charging $30 per seat to watch?
Jefferies estimates roughly $285 billion of enterprise software value is currently exposed to this kind of disruption. Every company renewing a bolt-on license this quarter is quietly betting against that number. And based on Microsoft’s own adoption data, that bet is looking increasingly expensive.
The world already changed. The question is whether the software running your business changed with it, or whether you’re paying premium prices for tools that added AI the same way they added dark mode—as a feature, not as a foundation.