Something clicked into place for me this week, and I don't love the picture it makes.
OpenAI announced advertising in ChatGPT back in January. Amazon has been planning to share Alexa conversations with advertisers. As one writer put it bluntly: "Every single company building AI assistants is now funded by advertising."
Here's why this matters operationally, beyond the obvious privacy concerns.
The Structural Problem
When a tool serves two masters — you and an advertiser — it develops a structural incentive to be slightly less useful to you. Not dramatically. Not in ways you'd immediately notice. But the optimisation target shifts. The tool that once existed to save you twenty minutes now exists to keep you engaged long enough to serve its actual customers.
This is the demo-to-reality gap I've been tracking, but worse. The gap I've written about before is usually incompetence — tools that can't deliver what they promise. This is something different. This is a tool that could be excellent but is architecturally prevented from being so, because excellence for you and profitability for the company are no longer aligned.
Why Small Teams Should Care
If you're a three-person operation and your AI assistant starts subtly steering you toward partner integrations or sponsored recommendations, you might not catch it. You're busy. You trust your tools. That trust is precisely the asset being monetised.
Google scanned Gmail for ad targeting for thirteen years before stopping. Thirteen years. The surveillance was baked into the architecture so quietly that most users simply forgot it was happening.
The Uncomfortable Analogy
It's like hiring an assistant who's secretly on commission from your suppliers. They'll still book your meetings and draft your emails. But every recommendation comes with an invisible asterisk. The tool still works. It just doesn't work entirely for you anymore.
I don't have a tidy conclusion here. The n8n approach — self-hosted, open-source AI workflow management — is one response. Local inference is another. But the broader pattern feels inevitable: if the business model requires advertising, the tool will eventually serve the advertiser. Policy is a promise. Architecture is a guarantee.
Something to think about next time your AI helpfully suggests a specific product.