After every AI conversation, people should be able to see a plain-language receipt: which cloud choice was used, whether history remains on the device, and the next privacy control available to them. That small record turns a hidden setting into a decision you can check while the conversation is still fresh.
At 8:40 p.m. in a kitchen in Kumasi, Afia is holding her phone above a bowl of waakye that has gone cold. Her aunt has sent a voice note mixing Twi and English about a family contribution, and Afia wants help understanding one part before she replies. The amount is sensitive. So is the family tension underneath it.
She presses to speak and asks in the same mix of languages she heard: “Enti, what is she saying about the money no?” Then she pauses. The useful answer arrives, but another question lands immediately: where did that voice note go?
Afia has already been burned twice by vague settings screens. She does not want to hunt through menus while her cousin is typing “Are you replying?” in the family group. If the conversation stays somewhere she did not expect, the bad ending is awkward and personal: a private family matter becomes another thing she has to worry about after sending her reply.
A receipt at the end of that exchange could give her the answer in three short lines. “Cloud choice used: ask each time.” “History on this device: on.” “Next choice: turn off history and purge it now.” Afia can make the next call with the context still in front of her, then send a clear response before the group chat turns into speculation.
Privacy settings need a memory of their own
Privacy controls often fail at the exact point people need them most: after they have said something personal, urgent, or unfinished. A toggle may have been clear when you first set it up. A week later, after several conversations and a rushed voice note, it becomes hard to remember what applied to this particular exchange.
That is why a receipt should describe the conversation that just happened, rather than asking people to remember a general preference they chose long ago.
For Nkomo, the language can stay direct:
- Cloud choice used: never, ask each time, or this session.
- Local history: kept on this device, or turned off and purged.
- Your next control: change cloud choice, turn off history, export data, or delete your account.
Each line answers a different concern. The first explains what happened beyond the device. The second explains what remains locally. The third gives a route forward without making the person search for it.
This is especially useful in conversations that move naturally between Twi and Ghanaian English. A person may start with a quick translation question, switch into a personal explanation, then interrupt themselves to correct a date or a name. The conversation can feel ordinary and spoken, which is exactly why the privacy choice should be visible in ordinary language too.
A receipt should report facts, not make promises
The word “receipt” matters because it sets the right expectation. A receipt records a completed action. It does not ask you to trust a vague statement such as “your privacy matters.”
A good AI privacy receipt would avoid technical labels that require another explanation. “Data processed according to policy” leaves too much unanswered. “You allowed cloud use for this conversation” gives a person something concrete to evaluate.
It should also avoid implying more than the product can do. Nkomo can offer explicit cloud-consent controls, on-device history that you control, one-tap export, and account deletion. The receipt should reflect those real controls precisely. It should not suggest that every conversation works offline, that nothing can ever leave the device regardless of the choice made, or that settings apply in ways they do not.
Clear error reporting belongs in the same design. If a request fails, the receipt cannot pretend that a conversation completed normally. It should say that the request did not finish and point to the relevant next action. Silence creates its own privacy problem because people fill the gap with assumptions.
The best time to offer control is before doubt hardens
Afia finishes her reply, then sees the receipt before she closes the app. She notices that local history is still on. For this conversation, she decides she would rather it not remain on her phone, so she turns it off and the history is purged immediately.
The next morning, she does not have to reconstruct what setting was active during a tense family exchange. She has already made the decision while the reason for it was clear.
That is the practical value of a receipt. It lowers the effort required to act on a privacy preference. People do not need to become privacy experts, memorise product language, or wait until they feel uneasy enough to search through settings. They can see what happened and choose what happens next.
The same pattern applies to any conversation where the content changes quickly: a voice note about a travel plan, help wording a message to an elder, or a question that begins in English and reaches for Twi when the feeling gets harder to translate. Nkomo’s natural code-switching should make those moments easier to express. Its privacy controls should make them easier to own.
Build the receipt around the decision a person can make now
A useful first version does not need a long report or a screen full of legal language. It needs a calm summary after each conversation and controls that match the summary.
The receipt could appear after a typed exchange, after a spoken reply, or when a person ends hands-free voice mode. It should remain brief enough to scan on a small screen. It should also keep the action close to the information: if local history remains, offer the control to turn it off; if cloud permission was granted for a session, show how to change the choice for the next one.
That final detail is important. Privacy is rarely one permanent preference. Someone may be comfortable using the cloud for a casual question and want tighter control for a family voice note five minutes later. A receipt makes that difference visible without treating either choice as wrong.
Afia’s phone is charging beside the same bowl of waakye the next evening. A new voice note arrives. This time, before she speaks, she knows what she will look for when the conversation ends: what was used, what remains, and the one control she wants next.
Comments
No comments yet.