Abena’s queue is waiting. A missed sale could cost her customer.

For a market seller, the useful interface is the few seconds between one customer paying and the next customer asking a question. Voice matters because stopping to type can cost attention, a sale, or both.

At 10:40 in the morning, Abena has three people waiting at her stall in Kumasi. One woman holds up a cloth and asks for another colour. A man at the edge of the table is already counting notes. The third customer is watching Abena’s hands, waiting to see if she will find the smaller size.

Abena has just sold two items together. She wants to keep the details straight before the next rush blurs them: the colours, the quantities, the customer’s request to return later. Her thumb is sticky from handling coins. Opening a keyboard would pull her eyes away from the queue.

She speaks instead, in the mix of language that comes naturally: “The blue one and the smaller red one, wɔnkae sɛ ɔbɛsan aba for the other size.” Then the woman with the cloth asks a follow-up question.

The risk is small but real. If Abena loses the thread of that sale, the returning customer may come back to an argument about what was promised. If she loses the people standing in front of her, they may walk to the next stall. A queue does not wait politely while someone finishes typing.

Speech belongs where the work is happening

People often describe voice tools as convenient. At a stall, in a kitchen, in a taxi, or while carrying a child, speech can be the only input that fits the moment.

The work is already moving. A customer changes her mind. Someone asks for a price in English, then explains the exact item in Twi. A family member calls with a request halfway through another task. Requiring a person to select one language, type carefully, and wait for a response asks them to leave the actual situation in front of them.

Natural code-switching matters here. The point is not to make every sentence entirely Twi or entirely English. It is to let people speak the way they already explain things: the product name in English, the correction in Twi, the emphasis wherever it belongs.

That is why a language picker can create a pause at exactly the wrong time. The Language Picker Nkomo Removed, and What a Pause Can Cost looks at how a small extra decision can interrupt a conversation before it begins.

A voice interaction has to survive interruption

Back at Abena’s stall, the man with the notes asks whether she can reduce the price if he takes two. She turns to him. The woman with the cloth is still waiting. Abena cannot give a device her full attention for a long exchange.

A useful voice interaction leaves room for that. With Nkomo, she can hold to speak, hear a reply, and interrupt naturally when a customer needs her. She can type when the stall is quiet and speak when both hands and eyes are already occupied.

The same rule applies beyond the market. Think of the person preparing food while a relative sends a voice note, or someone walking home who needs to rehearse a message before sending it. The interface has to make space for the life around it.

Speech also needs to be honest when it fails. A missed response or an error should appear clearly, because the person has to decide what to do next. Silence leaves them guessing whether the message was heard, saved, sent somewhere, or lost.

Privacy decisions should fit the pace, too

A busy moment should not force a person into a vague bargain with their data. If Abena is speaking details aloud near customers, she deserves to know what stays on her device and what reaches the cloud.

Nkomo makes that choice explicit: never, ask each time, or this session. Its history is on-device, and turning history off purges it immediately. People can export their data or delete their account in one tap.

These controls matter because voice can feel casual even when the content is personal. A person may switch from discussing a sale to rehearsing a message for family, then back again. The language can change quickly. The sensitivity can change even faster.

For a plain-language look at those choices before using the microphone, read Before you tap the microphone.

The morning after should feel simpler

By midday, Abena has served the queue and remembered the customer who plans to return. She has not spent the whole morning looking down at a screen. Her attention stayed where the work was: the cloth in her hands, the questions in front of her, and the people deciding whether to wait.

That is the test for a conversational tool in a crowded day. It should meet people in the language they are already using, let them speak when their hands are busy, and make its privacy choices clear before the conversation gets personal.