Voice expense logging: when it actually wins and when it doesn't

Dictating an expense isn't always faster than typing it. It depends on the moment, and there are four where it wins by a mile and two where typing is better.

Knowing which is what makes the feature get used properly instead of tried once and abandoned.

The four moments where it wins

1. When your hands are full. Leaving the supermarket with two bags, walking, in a moving car. It's the most obvious case and the most frequent: it's exactly the moment the expense just happened and typing isn't an option.

2. When there are several expenses in a row. A day of travelling: breakfast, taxi, tickets, lunch. Typing four expenses with their four categories is tedious; dictating them one after another isn't.

3. When the expense has several fields. "Forty-two on groceries with the blue card" is three fields in one sentence. Typing it means opening three pickers.

4. When you're about to put it off. This is the highest-yield case and the least obvious. The real alternative is almost never "typing it properly": it's "I'll log it later", which means not logging it. Dictating it in five seconds beats not logging it, even if you have to fix the category afterwards.

The two moments where typing wins

When it's genuinely noisy. A crowded bar, the street at rush hour. The transcription will come out wrong and fixing it takes longer than typing would have.

When the expense is unusual and you want it exact. A large expense you'll need for a deduction, with a long note. There precision matters more than speed.

How to dictate so it comes out right the first time

Four things, and all four are habit rather than technique:

Say the amount first. "Thirty-two on coffee" works better than "coffee, thirty two". The number at the start is the most important field and the one that tolerates a transcription error least.

Use your category's name, not a synonym. If your category is called "Groceries", say groceries. Kesef sends the transcription service the names of your categories precisely so it can match — saying "supermarket" when your category is "Groceries" usually works, but saying the exact name always does.

One sentence per expense. If there are four expenses, four short sentences. One long sentence with four amounts is easier to split wrongly.

Check before confirming, not after. Fixing it in the moment costs one tap; fixing it tomorrow means remembering something was wrong.

What gets sent and what doesn't

Worth knowing, because it's the first reasonable question anyone asks when they hear "AI voice logging": the audio and the names of your categories are sent, so the service can work out which expense you're talking about. Your amounts, your history and your identity are not. It's detailed in the privacy policy and on the voice logging page.

What voice doesn't fix

It doesn't fix not having the habit. If the problem is that you never remember to log anything, voice lowers the friction but doesn't create the reflex — that comes from recurring expenses set up once (which log themselves) and from the monthly close.

In one line

Dictate when your hands are full, when there are several expenses in a row, when the expense has several fields, and above all when the real alternative is not logging it. Type when it's noisy or when the expense has to be exact.

Kesef is free, on iPhone and Android

Shared expenses with no cap, 60+ currencies, and the numbers explained.