How Much Time Dictation Saves at Work

Dictation saves time only on the words you would otherwise type, and on those words it is much faster. In one controlled study, English entered by speech on an iPhone reached 153 words per minute, against 52 on the built-in keyboard ^{1 2}. A working week saves less than that ratio suggests, because editing the dictated draft takes time, and deciding what to say takes the same time either way.

For someone who types 10 hours of finished text a week, the arithmetic is short. Speech at the measured speeds needs about 3.4 hours for the same words, which saves 6.6 hours before any editing ^{3 4}. If correcting takes a quarter of the speaking time, the net saving is about 5.8 hours ⁵. With the correction load measured in a 2001 study, the saving shrinks to about 0.3 hours ⁶. The rest of this page explains where your own numbers go, so you can run the same calculation on your own week. For the wider subject, see our overview of dictation for work and study.

Dictation saves time only on the writing part of a job, so measure that part first

Dictation changes how fast you turn words in your head into text. It does not shorten meetings, reading, research or the work of deciding what to say. So the first input is how many hours you spend writing, and that varies widely by job.

In a 2018 survey of working professionals in several fields, run by North Carolina State University, the highest reported share of the work week spent writing was 42 percent, for marketing professionals ⁷. Programmers reported the lowest, at 26 percent ⁸. The other fields averaged between 30 and 40 percent ⁹. These are averages of what respondents said about their own work ¹⁰, not logged time. Use them to set expectations, then measure your own week.

Heavy writers spend far more time at it. In a 2016 survey of 547 business people who write at least two hours a week for work, not counting email, respondents averaged 20.4 hours of writing a week ¹¹. That sample is self-selected, so it describes people who write for a large part of their job, not the typical office worker.

Turning shares into hours is simple. A programmer on a 40-hour week who writes 26 percent of it has about 10.4 hours of writing ¹². A marketing professional on the same 40-hour week, at 42 percent, has about 16.8 hours ¹³. Count only the hours that go to composing or revising text. Leave out time spent on data entry, spreadsheets or code, because those involve typing but are not prose. Those writing hours are the input the formula starts from.

Speaking is about three times faster than typing ¹, but correction time narrows the gap

The gap between speech and typing is real, and in the controlled comparison below, it is large. In the iPhone study, speech entered English text 2.93 times as fast as the keyboard, and the speed counted the finished text, including the time spent fixing errors ¹. The keyboard figure of 52 words per minute is close to the 51.56 average in a very large online typing test ¹⁴. That test drew self-selected volunteers, and its average is higher than the population figure.

Three things narrow the gap.

The first is the task. The phone study had participants enter short messages, which is transcription of text that already existed. Composing a document while you speak is slower, and the thinking section below covers it.

The second is errors. Speech produced fewer errors during entry than the keyboard, but it left slightly more errors in the final text: 1.30 percent uncorrected against 0.79 percent for the keyboard ¹⁵. A newer engine has been tested since then. In a 2025 test of one 7-minute-31-second episode, Apple's speech API had a word error rate of 10.3 percent ¹⁶, which is about one wrong word in ten. That is one test of one recording, not a rate for every speaker, but it is enough to put an editing budget into the calculation. Word error rate is the standard measure for this, and our glossary entry on word error rate explains how it is counted.

The third is the correction load from earlier systems. In a preliminary 2001 study with a small sample, users spent over 65 percent of their time correcting errors and under 35 percent dictating ¹⁷. That was an early recognizer, and the study predates current engines. Read 65 percent as a sign of how large the correction share can be, not as a current average. It still shows what is at stake: when correction dominates, the speed gain disappears.

The formula: hours writing per week, times the speed gain, minus editing time

Write W for hours a week of keyboard writing, T for your typing speed in words per minute, S for your speaking speed in words per minute, and E for hours spent editing dictated text. The net hours saved are W minus the speaking time (W times T divided by S) minus E. Every input can be measured in one week, which is the point of the method.

Using the iPhone study's speeds, a typing speed of 52 ² and a speaking speed of 153 ¹, the speaking term is 3.4 hours for 10 typed hours ³. The table below shows what happens with three editing assumptions.

The inputs need to be measured, not assumed:

• W, writing hours. Log them for five working days, as described in the last section.
• T, typing speed. Measure it on the keyboard you use most. The online average was 51.56 words per minute, but that test drew self-selected volunteers, so it is a reference point, not a norm ¹⁴.
• S, speaking speed. Measure how fast you read your own draft aloud. The 153 figure ¹ is a reference for entering known text, not a promise for every speaker.
• E, editing time. Record every correction, every re-dictation and every trip back to the start of a sentence during the test week. This is the input most people underestimate.

The formula assumes you already know what you want to say. When you compose while you speak, the speed falls sharply. Earlier reports of composing with speech software gave 8 to 15 words per minute ¹⁸, far below normal speaking speed. Those systems were older, but the lesson holds. Dictation is fast when the words already exist in your head, and slow when you are still working out what to say. For a useful version of the method, outline the section and then dictate it.

Managers, support agents, sales reps and developers write different things and save different amounts

The same formula gives different answers by role, because the writing itself differs in length, structure and how much of it is code or data.

Managers and other professionals write a mix of email, reports and feedback. Dictation works well for long messages you can say in full sentences. It gains less on short replies, because a short message has less text for each second you save. Check the share of your log that is email, since email is often where most of the writing is. For a method, see how to dictate emails faster than you type them. Microsoft says that the Dictate option in Outlook is available only to Microsoft 365 subscribers ¹⁹, so check the license before you plan a rollout around it.

Developers are the clearest case where speech is not a simple win. In a 2006 evaluation of a speech-based Java system, most expert developers had little trouble learning the system, but they found it slower than typing ²⁰. The system was an early research tool, so this does not settle the question for current software. It does suggest that dictating code syntax is a poor use of speech. Prose dictates well: tickets, design notes, pull request descriptions and documentation. A programmer working a 40-hour week at the survey's 26 percent writes about 10.4 hours a week ¹², and only part of that is prose. For more on this, read how developers use dictation.

Sales and support staff write short, frequent notes: call summaries, follow-up emails and case updates. We did not find an independent study that measures how much of a sales rep's or support agent's day goes to typing notes after calls. Measure your own number instead. Count the minutes spent on notes and follow-ups for a week, then run the formula on that. Short notes are where the start-and-stop cost of dictation is largest, so the gains per item are smaller than for long documents. See how sales and support teams use dictation for the workflows people use.

Heavy writers sit at the top of the range. The 20.4 hours a week in the 2016 survey ¹¹ is the largest pool of writing time in these figures, so even a modest net share of it adds up. The same formula applies: keep the editing load low, or the saving falls toward the 2001 result. For a related look at long-form work, see writing a book by dictation.

Thinking time does not shrink: dictation speeds up getting words down, not deciding what to say

Planning, not typing, takes most of the time spent composing a letter. In a 1978 IBM study of letter writing, about two-thirds of composing time was planning, whether the letter was written, dictated or spoken ²¹. Dictation changes the remaining third, the time spent getting the words down. The study measured one-page business letters, and the share may differ for long documents, so for a proposal or a report that has to build an argument, the total gain may be smaller than the typing ratio suggests.

Experience changes the result too. In the same study, people with years of dictation experience composed letters between 0 and 25 percent faster than novices, depending on how complex the letter was ²². A longitudinal study of hands-free dictation found the same pattern: users became more efficient at dictation, navigation and error correction as they gained experience ²³. A trial that lasts a single day measures the learning curve as well as the method, so run it for a week and keep the log going.

The practical rule follows. Plan the argument before you dictate, either as a short outline in your own words or as bullet points. Then dictate each section in one pass, and edit afterwards. This matches what the 2001 correction numbers warn about: if you dictate half-formed sentences and then fix them, the editing time eats the gain.

Team rollouts depend on habit and workspace as much as on the software

Dictation in a shared office has a cost the formula does not show, which is the people around you. In a 2005 field study, 99 percent of employees surveyed said their concentration was impaired by components of office noise, especially ringing phones and background talk ²⁴. The authors found no evidence that people get used to these sounds ²⁵. A 2019 laboratory study of open-plan offices writes that speech noise is generally accepted as the main source of discomfort for employees ²⁶.

A rollout should therefore decide where dictation happens. The options are a private room, a meeting pod, or a headset with a close microphone, and dictation blocks scheduled so that colleagues do not have to listen to a stream of speech all day. For the full answer, see whether you can use dictation in an open-plan office and our guide to choosing a dictation microphone.

Habit is the other half. A few days may not be enough to see the gain, because experience changes the result, as the section on thinking time shows. Pick one document type to start with, such as email replies or meeting notes, and log the results before you widen it. Keep a short list of names and terms that the software gets wrong, and add to it as you go. A one-page checklist for the first two weeks is worth more than a second software trial.

Measure your own gain for one week before paying for anything

Run this procedure before you buy anything. It takes one working week and gives you the four inputs of the formula.

1. Log your writing hours for five working days. Use a timer and mark each block as email, documents, reports, notes, code or other.
2. Measure your typing speed on a short timed test, using the method in our guide to how typing speed is measured.
3. Dictate for one week with a free tool first. Our overview of how to choose a dictation app explains what to compare. If the built-in tool on your computer is enough, use it: for Mac, see voice to text on Mac, and for Windows, see our Windows page.
4. Count editing minutes separately. Every correction, navigation step and re-dictation goes in the editing column, not the writing column.
5. Run the formula with your numbers, then compare the result with your cost.

Put a value on the result only after the numbers are in. At the median pay for software developers, $135,980 a year ²⁷, the hourly equivalent is about $65.38 ²⁸. At that wage, the 5.75 net hours saved per 10 typed hours in the 25 percent case are worth about $376 of time ²⁹. That is time valued at a wage, not cash. It only becomes money if the saved hours go to work you would otherwise pay someone to do or would have delayed.

Compare that value with the software price on the vendor's own pricing page, and check the free plan's limits too, since a usage cap can matter in a heavy writing week. Our app, Yap, which we make, runs on-device on Mac and Windows, and its plans are on our pricing page. It is one option among several, and the same week-long test applies to any of them. If the saved minutes do not cover the cost at your own pay rate, the built-in tool or no tool may be the right answer.

Related guides

• Dictation for work and study, the overview of this subject
• How to dictate emails faster than you type them
• Writing a book by dictation
• Speaking versus typing speed, the speed comparison in more detail
• Dictation versus typing
• The cost of AI dictation