AI Time Tracking – what works, what doesn’t

The mere thought that AI time tracking could mean faster, easier, fully automated time tracking, or even just make time tracking less of a chore is very attractive and worth exploring.

AI can certainly help save people time in a range of tasks, and time tracking is no exception. However, with many companies scrambling to ‘hop on the band wagon’, it is important to see past the gimmicks as many approaches actually end up taking users longer and can overcomplicate the process. Anyone who has used Alexa, Siri, or Google knows that AI occasionally gets the wrong end of the stick and can waste time.

Timewatch has been researching the use of AI for many years now and has so far trialled a number of approaches to AI time tracking and AI timesheet entry including voice and chat, and have experience of what works, and what doesn’t.

AI timesheets that don’t actually work and why

Our first forays into AI were using Voice and Chat to track time. In testing, research showed it takes much longer to speak to a phone or use chat than to enter the data manually! The reason for this is that AI was just providing a different way of entering data, and it turns out that using a keyboard, even on a phone, and just typing in a customer, project and time is much faster than talking to a phone or chatting. Chat was actually the slowest form of AI time tracking as typing “add 2 hours to customer ABC, project 12345 on the 25th of April 2024” takes longer than choosing a date and entering “3, ABC and 12345” into the appropriate timesheet fields in a system built for the job.

Compounding this are the inevitable amusing but time-wasting cases where AI misunderstands what we say or type. Not surprisingly, we abandoned these forms of AI time tracking, so it brings a smile to our faces when we see other companies releasing AI timesheets that use voice or chat as their AI timesheet.

We wanted to take a different approach, one that would take advantage of the self-learning capabilities of AI; we also needed a yardstick against which to measure how effective any AI timesheet is and if it is really helping.

The best way to implement AI

In considering the best way to implement AI time tracking, Timewatch refuses to create a product for novelty value or to jump on a bandwagon or trend. We also require that AI tools be deployed in existing applications should be a choice. This has led us to create eight golden rules for the effective use of AI in any time tracking apps.

1) AI must improve the quality of life for users

The adoption of AI apps can automate repetitive tasks, freeing up time for more creative work or a better work-life balance.

2) AI must create less - not more - work for users

Just as AI can remove the burden of repetitive tasks for your team, inaccuracies could create more work for someone else. For example, automated timesheets will save employees time, but inaccuracies will cause more work for admins and managers to monitor for and correct, which must be avoided.

3) Users must always remain in control of AI

Many of us are familiar with Alexa, Siri and other AI apps getting things wrong. It’s not as bad as HAL 9000 from the Arthur C. Clarke novel / Stanley Kubrick film 2001: A Space Odyssey, but users must feel in control. When things go wrong, we need to be able to correct issues or control automation and be confident that it is effective. We don’t want to get into arguments with an AI app.

4) Any AI functionality should not change the way people work

When enhancing existing software, progressive users should be able to adopt new AI features, but it should also be possible for employees not to use it, and those that want to try it to use it in parallel with existing systems without it inconveniencing them.

5) Don’t just jump on a AI bandwagon

AI shouldn’t be implemented just for marketing reasons, it needs be effective. Jumping on the bandwagon fails every one of our golden rules.

6) Understand the limitations of AI

AI is a powerful tool but it’s not perfect. It’s important to understand what AI can and can’t do before you start using it. When working on a critical task no one wants to hear, “I’m sorry Dave, I’m afraid I can’t do that.”

7) Use quality data and use it appropriately

The quality of data has a big impact on any AI apps performance.  Make sure your data is accurate, and relevant to your goal.

8) Don’t be creepy

A common issue people have with Voice systems is that it listens to them. AI needs user data to be able to help them, so be realistic, think about what data is needed, how it is stored, used and secured or if it needs to be stored. Address user concerns up front.

Sadly, to date every “AI timesheet” we’ve seen fails most if not all of these rules. That doesn’t mean AI can’t be used for effective time tracking, it can and it will, and Timewatch customers will very soon.

How AI time tracking can be faster and easier

All time tracking apps need to know the date and time of the work done as well as who or what the work was for. AI can’t read our minds to fill in our timesheets for us, but it can look for patterns and use existing data to work out some of this for use to save us time entering timesheets.

One thing all time tracking software systems have in common is that the user needs to specify who or what the work is for: a customer, a project, or a task. AI can help users do this if there is existing data to work out what customer or project work is for (see “Use quality data and use it appropriately” above).

Can AI automate time tracking?

Short answer: yes!

Timewatch already works with calendars, voice call records, meeting invitations, and other details to help people recall the work they have done to help more accurately complete their timesheets.

Our most popular system is our Outlook and Google timesheet app. This uses the date and time information from calendars for tracking and lets users add customer / project to turn appointments into timesheets. These systems automate some of the timesheet entry process, but customer / project entry is manual. With AI we can make this even better.

Our forthcoming update will use data from Outlook appointment Subject and notes to work out which customer, project and task an appointment is for. For example, an appointment with a Subject of “Meeting with Kevin from Arcadia” has all the information for automated time tracking. Our AI timesheet can correctly find the correct allocation of Customer: Arcadia, Project: Arcadia Extension, Task: Meeting in this case because Kevin is a contact on this project.

Will AI time tracking be accurate?

This is the big question, as errors in time tracking software can cause serious problems, especially if time is used for billing customers, it must be 100% accurate.

Unfortunately, at this point in AI evolution, AI is not 100% accurate. We took the approach that it makes sense to assume AI will never be 100% accurate and build in self checks and error feedback and correction. Users will be able to set a level of accuracy that is needed for AI to automate time tracking, but as per rule 3 – users must always be in control – users can override automated entries. Overriding an automated entry decreases accuracy, creating a self-correcting feedback loop learning from the customer / project suggestions the user chooses to re-learn until the required level of accuracy for automation is again met.

Will correcting errors create more work for users?

What happens when there are errors from an AI app? Does the time taken to correct errors undo any benefits AI gives? In facing these questions, our research concluded that there needs to be some yardsticks against which the value of AI choices can be judged, but that automation can always be undone with a single click – rule 3!

Using AI in speech to help in time tracking

Our first foray into the value of AI time tracking apps was using Siri and Google to speak to a timesheet. It was all very clever and interesting and – it worked! We could talk to our timesheet and it would automatically track time via AI, but after the novelty wore off, we realized its use was pointless. It took longer to enter timesheets by voice than it did manually typing, especially when AI misheard the smallest thing. If you use Siri, Google or Alexa you know that sometimes they get it very wrong.

Using AI chat to help in time tracking

Interestly this is an approach many organizations are using, but we discounted this approach as being viable many years ago. Again, it’s all very interesting, but once the novelty wore off – AI timesheet entry via chat wasted more time than voice! It was faster to type data into a timesheet that having a conversation with chat and AI made more errors / misunderstandings than a user using manual entry. Our (not-so) favorite AI timesheet model we have seen is an AI chat screen showing a back and forth with an AI bot demonstrating it would be faster to open a mobile timesheet app and enter the date, time, and project than have a debate with an AI chat bot.

The perfect answer to AI time tracking

It turns out that the perfect implementation of AI in timesheets is to let AI do what AI is really good at – using existing data to save people time and automate time tracking, not use AI chat or voice to make the process harder. This is not to say AI timesheet entry via voice or chat won’t improve, but even if it were perfect there won’t ever be a huge saving in time. Imaging typing into a Chat bot “Add Customer: Arcadia, Project: Arcadia Extension, Task: Meeting, Date: 24th of December 2024, Start time:9:00am End time: 11:am, or Duration: 2 hours”! This is always going to take longer than choosing the correct customer, project, date and entering the time spent! And then there is voice – saying this is always going to have issues in misunderstandings.

The best use of AI that achieves the biggest time saving is to use data that already exists to save employees time.

A good example is our Outlook timesheet App, which turns appointments into timesheets. Using Outlook appointment data saves users having to enter the same date all over again, but users still need to enter customer / project information. However, if AI can use the appointment Subject and/or Notes to work out the customer / project, Outlook appointments can be accurately and automatically turned into timesheets. This approach really would save employees time. It would actually save all the time users would normally spend entering timesheets or time tracking and it would be accurate. Interestingly, it would also save all the time spent using AI voice or AI chat!

Who is Timewatch?

Timewatch are specialists in professional services timesheet solutions as well as resource scheduling and professional services automation software.  Our unique SP is that our solutions are designed to be customized to each organization’s specific requirements, but retain a single codebase so that customers can always upgrade to new releases.

Timewatch provides the perfect blend of a bespoke / customized solution with all the benefits of an off-the-shelf solution.  Timewatch is ISO 9001 and 27001 certified, we provide public cloud, private cloud and on-premises solutions.

Contact us

If you’d like to learn more about our AI time tracking software, timesheet reports and analytics available, see our systems in action or find out how it could integrate your timesheet system  with Outlook or is AI time tracking, please contact us.

Book a free call or live product walk through with a product specialist using the form to the right.

Speak with a specialist

Learn more about AI Timesheets

OutlookTime Inline

Timewatch do not share my information with any third party. In clicking submit you agree to our terms of service andprivacy policy


Other posts you may be interested in: