Colour theme
Book a consultation

Lessons from the 2024 AI-at-work report

Use the 2024 Work Trend Index as a starting point for staff discussions, the AI tools you allow, practical training and a measured business pilot.

A finger pointing at a line graph on a printed page of bar and line charts lying on a wooden desk.

A workplace trend report can help you ask better questions about AI. It can’t tell you which tasks your staff should automate or how much time your business will save.

Microsoft and LinkedIn’s 2024 Work Trend Index combined a survey of 31,000 full-time knowledge workers in 31 countries, Canada among them, with other research and platform data. It’s a 2024 snapshot, and it doesn’t describe Canadian small businesses today.

Here are five ways to turn that discussion into decisions you can check locally.

Find out how staff are already using AI

The report found that 75% of the knowledge workers surveyed used generative AI at work, and 78% of those users brought their own tools. Some of your staff may be doing the same.

Ask which tools people use, for which tasks and what information they enter. Make it easy to describe a workaround without presenting the conversation as a performance assessment. You need an accurate picture before choosing controls.

Sort personal subscriptions from accounts the business has set up. Record whether a tool receives customer records, internal documents or only public information. Don’t put sensitive examples into another AI tool to conduct the review.

Teach review skills alongside prompting

Only 39% of the people using AI at work told the survey their employer had trained them.

Training should include checking facts, recognizing missing context and deciding when a task needs human judgement. Ask staff to practise with fictional material and compare the result against known references.

A polished answer can still be wrong. Make the reviewer responsible for confirming names, figures, instructions and any statement that becomes a business commitment.

Choose training for the work people do. Someone drafting an internal outline needs different checks from someone preparing information for customers.

Test a complete workflow

Pick one bounded task, such as preparing an outline from material that’s already public. Record the time and quality of the current process before running the trial.

Count preparation, review and correction alongside generation time. Test awkward examples as well as easy ones. Staff reports of convenience are useful feedback, but they won’t tell you whether the business saves time overall.

Write down what would make the pilot successful and what would stop it. Include an ordinary way to complete the task if the service isn’t available.

Tell staff which tools they can use

A short policy should name the tools staff can use, permitted data, prohibited uses and who to ask when the task falls outside those rules. Explain where work is stored and whether it can be used for model training under the selected service terms.

Review those terms for the account and plan you’ll use. Versions sold under the same product name can have different privacy settings.

Keep accountability with the business

Canada’s privacy regulators’ generative AI principles address necessity, openness, safeguards and accountability. Consider the people affected by the use, especially when outputs influence decisions about them.

Keep sensitive decisions outside an initial productivity trial. Assign a business owner who can examine errors, hear staff concerns and change the process.

After the pilot, compare the evidence with the original goal. Expand only the tasks that met the agreed checks, and revisit the decision when the tool or data use changes.


Featured Image Credit