Research shows that AI assistants can help workers complete writing-heavy office tasks faster. But the best results come from using the tool as a draft generator and keeping the final check with a human. The examples below can be used with any general-purpose assistant.
- “From these notes, write a polite email apologising for the delay and proposing a new deadline.”
- “Cut this long email to three sentences and put the main request first.”
- “Make the tone of this reply more formal but warm.”
Meetings
You can give the assistant meeting notes or a transcript (recorded with participants’ consent) and say “list the decisions taken, with owners and dates”. Always check names and dates before sending the summary.
Spreadsheets and data
- “Write and explain an Excel formula that totals sales by month based on the dates in column B.”
- “How do I find duplicate customer records in this list?”
- “Pull out three key trends from this table summary as bullet points.”
Test formulas on a small sample first; AI sometimes suggests the wrong column or function.
Presentations and reports
- “Create a 6-slide presentation outline for managers from this report, with a title and three bullets per slide.”
- “Prepare two sentences of speaker notes for each slide.”
Checklist
- Are the figures and calculations right?
- Are names, titles and dates correct?
- Is the tone right for the recipient and company culture?
- Does it contain confidential information?
Safe use
Use customer data, contracts and internal documents only in enterprise tools approved by your company that come with a data protection agreement. You remain responsible for what you produce.
What does research say about productivity?
The first comprehensive studies measuring the impact of generative AI at work produced striking results. In an experiment by MIT's Shakked Noy and Whitney Zhang, published in Science in 2023, participants using AI for writing-heavy professional tasks finished the work about 40 per cent faster, and evaluators rated the quality of the output higher. In a study by Erik Brynjolfsson and colleagues at a customer support centre, the number of issues resolved per hour rose by 14 per cent on average; the largest gains were seen among less experienced workers.
These results apply to specific tasks and conditions. The same research also reminds us that blindly trusting the tool on subjects it does not know or gets wrong can lead to errors.
Set shared rules within the team
When AI moves from being an individual habit to part of teamwork, a few shared rules prevent confusion:
- Which tools are approved? Everyone should know the list of tools and account types the company allows.
- Which data is not shared? Clear limits should be set for customer information, personal data, contracts and financial figures.
- Who checks? The person who does the final read of texts going outside should be clearly defined.
- When is it disclosed? Whether AI assistance is noted in reports should be agreed in advance.
AI tools’ features, plans and prices change often. Always check the tool’s official page for current information.
Sources and further reading
- Erik Brynjolfsson, Danielle Li and Lindsey Raymond, “Generative AI at Work”, NBER Working Paper 31161, 2023.
- Shakked Noy and Whitney Zhang, “Experimental evidence on the productivity effects of generative artificial intelligence”, Science, 2023.
- Microsoft and Google help pages on office AI features.
- Noy, S. and Zhang, W. (2023), “Experimental evidence on the productivity effects of generative artificial intelligence”, Science
- Brynjolfsson, E., Li, D. and Raymond, L. (2023), “Generative AI at Work”, NBER Working Paper 31161
Leave a comment
Your email address will not be published. Required fields are marked *