Time tracking

What is automatic time tracking, and how automatic is it really?

Updated 4 min read

Short answer

Automatic time tracking means software collects signals in the background, such as which apps you use, which files you change, your calendar and where you've been, and turns them into suggested time entries. The software doesn't decide: you confirm which time counts for which client. You forget less, and you still only log hours that are right.

Capturing signals is not the same as logging hours

A time entry says: this work, for this client, on this project, this long. A signal only says something happened: Word was open for forty minutes, a commit came in, you were at an address across town at ten. Between the two sits a decision.

Software can rarely make that decision on its own. A window titled ‘Proposal.docx’ could belong to three clients. A visit to an address could be a client, or the bakery. Good automatic time tracking collects as many signals as it can, groups them into sensible blocks and makes a suggestion. You turn the suggestion into an entry.

Rule of thumb: collecting is automatic, judging is yours. Software that books hours to clients without you makes mistakes you only notice on the invoice.

Which signals can software use?

The useful sources depend on the device. A computer mostly shows what you worked on, a phone mostly shows where you were. Together they cover most of a working day.

  • Apps and window titles. Which program was active and what the title bar said, such as a document name or an email subject.
  • Browser. Which websites and web apps you visited, through a browser extension.
  • Files. Changes in folders you choose, such as a client folder.
  • Git commits. For developers: commits in chosen repositories, with time and subject.
  • Calendar. Meetings from Google Calendar or Outlook, with title and attendees.
  • Location. Visits to addresses and car trips, recorded by your phone.
  • AI agent sessions. Work you did with an assistant like Claude Code, Codex or Cursor. More in Logging hours with AI.

Why you still confirm

Hours feed invoices, budgets and sometimes tax or grant claims. A wrong hour carries through. Confirming briefly every day or week keeps you in control of what reaches your client.

Confirming is also when you write a description a client understands. ‘Excel 1:12’ means nothing to anyone; ‘reworked the 2027 budget’ does. Software can suggest that text, but you know what actually happened.

  1. At the end of the day or week, open the suggestions.
  2. Merge adjacent blocks into one task.
  3. Pick client, project and activity, or adjust the suggestion.
  4. Write a short description and save. Whatever you skip stays private.

What does your manager see?

This is the key question with automatic tracking. There are roughly two models. In one, every signal goes to the employer, who can see which sites you visited and how long you were idle. In the other, the signals stay yours and your employer only sees the hours you register yourself.

The second model works better, because people log more honestly when their raw data isn't being read. It also keeps time tracking from turning into surveillance. Why that's better for productivity too is in Measuring productivity without surveillance.

Timer, automatic or filling in afterwards?

The three approaches don't exclude each other. Many people run a timer for long, clear-cut jobs and use automatic suggestions for the rest of the day. If you still work in a spreadsheet, there's a comparison in Tracking hours: spreadsheet or app.

Three ways to capture hours
Start-stop timerAutomatic (signals)Filling in afterwards
Effort while workingStart and stop at every switchNoneNone
Effort afterwardsLittleConfirm suggestionsA lot: reconstruct from memory
Small jobs and interruptionsOften forgottenCapturedMostly lost
AccuracyGood if you remember to stopGood, as long as you confirmVaries
PrivacyOnly what you enterDepends on who sees the signalsOnly what you enter

How Half5 does it

Half5 keeps signals in a personal “Timeline”. The desktop app for Mac and Windows records active apps and window titles, browser pages through a Chrome and Edge extension, file changes in chosen “Folders”, commits from chosen “Git folders” and sessions from Claude Code, Codex and Cursor. The iPhone and Android app adds location visits and car trips; your calendar comes in through Google Calendar or Outlook. You filter by “Computer work”, “Calendar”, “Locations” and “AI sessions”.

The timeline is end-to-end encrypted: “Only you can read your timeline. Half5 cannot access your timeline.” At the top you see “Still to register” with “Suggestions”; only when you confirm a suggestion does it become an entry your organisation sees. Location visits follow the same rule: “Only you see these visits until you confirm them.” “Automatic Client Recognition” and “AI-generated Description” are part of Premium.

Frequently asked questions

Does automatic time tracking log my hours entirely by itself?

No, not when it's set up well. The software collects signals and makes suggestions; you confirm which time counts for which client or project.

Can my employer see which websites I visit?

That depends on the software. Monitoring tools show it; Half5 doesn't, because the timeline is end-to-end encrypted and only confirmed entries are shared with your organisation.

Does it work if I don't sit at a computer?

Partly. A phone can record visits to addresses and car trips, and your calendar fills in meetings. Work that leaves no digital trace you add yourself.

How often should I confirm suggestions?

Daily takes a few minutes and you still remember what you did. Weekly works too, but the longer you wait, the harder it gets to write good descriptions.

So is a timer pointless?

Not necessarily. A timer is handy for a long task when you know up front who it's for. Automatic suggestions catch the small jobs and interruptions a timer misses.

Someone cycles home along the canal after work

Close your laptop on time today.

Try Half5 free for 30 days, on your own or with your whole team.