Many leaders start with the simplest thing. A subscription to a consumer tool, and everyone uses it. It is a good starting point, and I will never say otherwise. But there comes a moment when it gets stuck, and you might as well see it coming rather than discover it the hard way.
The public tool, perfect up to a point
A consumer tool is excellent for everything that does not touch your sensitive data. Rephrasing, translating, exploring an idea, learning, clearing the ground on a subject. As long as you entrust it with nothing confidential, the balance between benefit and risk is unbeatable. Start there, with no hesitation, and let your teams build the habit.
The problem does not appear straight away. It sets in as the use becomes serious. One day, someone pastes in a contract to summarise. Another day, a customer list to sort. A third, a sales strategy to challenge. With no bad intent, out of simple efficiency, your teams start handing an outside service information you would never email to a stranger. That is exactly how the trouble starts, not with a hack, with a habit.
The two tipping points
Two signals show it is time to move to an internal AI.
The first is confidentiality. As soon as you catch yourself wanting to entrust the tool with things you would not want to see surface elsewhere, the public tool has reached its limit. It is no longer a matter of comfort, it is a matter of security, and the extra cost of an internal solution justifies itself.
The second is relevance. A public tool does not know your company. It answers well in general, and worse and worse as your needs get sharper. The day you spend more time correcting its generic answers than using them, you need an AI that knows your trade, your documents, your reality.
How to decide without getting it wrong
My recommendation comes down to a trajectory, not a camp. Start with the public, to learn, test and identify the uses that really matter. It is the best way to discover, without investing, what you actually need. Many companies rush to a costly internal solution before knowing what they want to do with it, and end up with a fine tool no one uses.
Then switch to internal when one of the two signals lights up, confidentiality or relevance. At that stage, you know exactly which uses you want to serve, and the investment becomes targeted instead of a bet.
There is a middle path many forget, and it deserves to be known: you can perfectly well keep the public tool for harmless uses, and reserve an internal AI for sensitive content and business needs. The two are not mutually exclusive. The good leader does not choose once and for all, they put the right tool on the right use.
The real risk is neither the public nor the internal. It is not having drawn the line, and letting valuable information drift, by default, towards a public service without even thinking about it. Drawing that line clearly, telling your teams what goes where, is already half the decision. The rest is only a matter of timing.