The AI features you didn't sign up for
Every SaaS tool you pay for is quietly shipping AI features by default. You didn't procure them, you can't cleanly turn them off, and your team is now building workflows around them.
The AI features you didn't sign up for
Google added Gemini across Workspace last quarter. Microsoft rolled Copilot into Office. Notion, HubSpot, Slack, Zoom — every tool in your stack now ships an AI helper by default. Your team is building workflows around them. You didn't procure any of it.
The quiet contract
You bought a tool for a specific job. The vendor bolted a model on top and shipped it as a feature. Nothing about your contract changed. Nothing about your data classification changed. But the workflow your team relies on is now passing through a system you didn't evaluate.
That's a dependency. And most founders can't tell you the shape of it.
Ask three questions of any AI-enabled feature in a tool you already pay for:
- Which model is behind it? Not "which vendor". Which model.
- Where does the prompt travel, and what's logged along the way?
- If the feature disappears next quarter, what breaks in your workflow?
If you can't answer these for the tools your team uses daily, you're not managing your technology. It's managing you.
Two things happen next
The first is price. The AI feature that shipped free at launch will get its own tier the moment usage justifies it. Google, Microsoft, Zoom have all done it. Your team has built habits around a free feature; the invoice arrives; you pay.
The second is drift. The vendor tunes the model. The output changes. The behaviour your team relied on last month isn't the behaviour this month. There is no changelog for tone.
Neither of these is catastrophic on its own. Across a portfolio of tools, they mean your business is running on infrastructure you don't control and can't reason about.
What to do about it
Start with an inventory. For every SaaS tool your business pays for, note whether it now has AI features enabled by default. Most do. Then flag the workflows that depend on them. Some will be trivial. Some will be load-bearing.
For the load-bearing ones, decide deliberately. Keep the dependency and know you have it, or route around it. Both are valid. Ignorance is not.
The businesses that will handle the next five years well know exactly where their AI is, who provides it, and what they'd do if it went away. Most don't.
Not sure which parts of your stack quietly turned into AI infrastructure? Book a conversation and we'll map it with you.
Robin Carswell
More on
'AI-first' is a paint job, not an architecture
Most tools claiming to be AI-native haven't changed their architecture at all. Here's the test that separates genuine AI-native systems from legacy software with a chatbot bolted on.
The three workflows every business should automate first
Most automation projects fail because the starting point is wrong. Here are the three workflows that almost always justify the effort — and how to pick which one to tackle first.
What AI-native actually means for a business that doesn't write code
Most businesses calling themselves AI-native are at level one of four. The gap between levels is the difference between accumulating tools and building something that compounds.