AI Partnerships Need Product Visibility, Not Just Distribution
2026-05-17 · 4 min read · Janaina Maia
The reported tension between OpenAI and Apple is not just a partnership story. It is a product visibility story.
The New York Times reported that OpenAI has considered legal action against Apple over how ChatGPT has been integrated into Apple devices. The issue, according to the report, is that OpenAI expected the Siri integration to make ChatGPT more prominent and drive more people toward ChatGPT subscriptions. Instead, OpenAI has reportedly been disappointed that the option is hard for some users to find.
I think this is a useful warning for every team building AI into someone else’s workflow: distribution is not the same as adoption.
An AI feature can be technically integrated and still invisible.
Platform partnerships look powerful on paper. One company has the model. Another company has the user relationship, the operating system, the enterprise suite, or the workflow surface. Put them together and the value should be obvious.
But users do not experience partnerships. They experience moments in a product.
If the AI capability is hidden behind unclear affordances, vague routing, or a fallback path the user barely notices, it will not build trust or habit. It may technically be present, but it is not meaningfully part of the experience.
The design problem is attribution and control.
When an AI assistant hands work to another model, the user needs to understand what is happening. Not in a legal-disclaimer way. In a practical product sense.
- Why is this AI being used now?
- What can it do that the default assistant cannot?
- What information is being shared?
- Can I choose another option?
- How do I know which system produced the answer?
Those questions matter because AI changes the trust contract. A normal software integration can be quiet. An AI integration that interprets intent, handles personal context, or answers on behalf of another assistant should not be too quiet. Users need enough visibility to feel oriented, not surprised.
Enterprise AI has the same problem, with higher stakes.
This is not only about consumer devices. The same pattern is coming to enterprise software: copilots calling specialist agents, agents invoking third-party tools, workflow systems routing tasks to external models, and assistants pulling from multiple knowledge sources.
Design leaders should push for a clear interaction model before the architecture becomes a black box. If a system uses different agents or models behind the scenes, the interface needs to make that complexity understandable without making the user manage it manually.
That is a delicate balance. Too much detail becomes noise. Too little detail becomes mistrust.
The practical rule is simple.
Do not treat AI integrations as plumbing only. Treat them as experience decisions.
If a partner model is important enough to put inside the product, it is important enough to design properly: clear entry points, transparent handoff moments, meaningful consent, useful attribution, and a reason for the user to care.
The future of AI product design will not be won only by the strongest model or the biggest distribution channel. It will be won by the teams that make complex AI ecosystems feel legible, trustworthy, and worth using.