AI is changing your shopping

Voice agents connect real-time audio with tool calls.

Conversations become actions. AI can answer calls, respond to questions and use connected systems. You call the hairdresser – and an AI takes your appointment request. If it’s connected to the calendar, it can suggest free times. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Coding Agents edit code and run automated tests.

From a task to a code change. Describe a change. A coding agent can edit files and test the result. This can save work. Someone still needs to check whether the change is right. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Neural translation processes spoken language.

Automatic translation of conversations. An app can translate spoken sentences into another language. This helps when travelling. Check names and important details. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Language models generate context-aware reply drafts.

AI suggests email replies. An assistant can suggest a reply to an email. Read it through and decide for yourself what to send. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Semantic image retrieval indexes visual content.

Find photos by their content. You can search for photos with words such as ‘dog on the beach’. This saves time if your photo app supports the feature. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

OCR extracts text from scanned documents.

Scans become searchable. Software can recognise writing on a scanned page. You can copy the text. Check numbers in particular for mistakes. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Automated event extraction structures messages.

Add appointments from emails. A program can pick out the date and time from a message. Check the entry before relying on the reminder. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Speech recognition generates live transcriptions.

Captions for video meetings. Spoken sentences appear as captions during a conversation. This makes it easier to follow along. Names can be transcribed incorrectly. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Generative search systems synthesise search results.

Search engines summarise sources. A search engine can show you a summarised answer. For important questions, open the linked original sources too. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

On-device models execute inference locally.

AI runs on the device. Some AI features run directly on your phone. Whether data leaves the device depends on the specific feature. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Language models simplify complex sentence structures.

Make long texts easier to read. An assistant can rewrite a complicated paragraph in simpler words. Compare both versions so you do not lose an important point. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Object recognition supports visual inventory analysis.

Identify objects in photos. An assistant can try to name the objects in a photo. Similar objects can be confused. A picture is not always enough. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Recommendation systems analyse individual listening patterns.

Music suggestions adapt. A music app suggests songs based on what you listen to. You can discover something new. Your past preferences shape the selection. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Adaptive learning systems vary task difficulty.

Exercises adapt to your answers. A learning program can make exercises easier or harder. You can practise more effectively. Explanations may still be wrong. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Dialogue systems automate standardised service requests.

Chat programs answer routine questions. A chat program can answer simple questions about an order. For a complicated problem you will often need further help. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Image description generates alternative textual representations.

Software describes image content. A program can describe what might be visible in an image. This can make information more accessible. Important details may be missing. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Automated summarisation extracts key conversation points.

Create notes from meetings. Software can turn a conversation into a short summary. Everyone involved must agree to the recording. Check the notes. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Nowcasting models short-term precipitation patterns.

Short-term rain forecasts. A weather app estimates whether rain is about to reach your location. This helps with planning. Even short-term forecasts can be wrong. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.

Generative systems create ingredient-based recipe suggestions.

Cooking ideas from available ingredients. An assistant can suggest a dish based on your list of ingredients. Check quantities and preparation steps. Take responsibility for checking allergens. The specific task, the data used and the limits of the method matter. A feature description alone does not show how reliably a system works in everyday use.