A couple of years ago, using an AI assistant at work felt like a bit of a novelty. Now they’re built into email clients, spreadsheets, meeting software and pretty much every tool with a monthly subscription. The marketing promises huge time savings across the board, but if you’ve actually used one for a few months, you’ll know it’s more mixed than that. Some jobs genuinely get quicker. Others take longer once you factor in checking and fixing what the AI produced.
So where do they actually help, and where are they mostly just getting in the way?
What the research says so far
There’s now a decent amount of real-world research, and it paints a fairly consistent picture. One of the most widely quoted studies looked at over 5,000 customer support agents at a software company who were given an AI assistant. The NBER working paper “Generative AI at Work” found that on average they resolved around 14% more issues per hour. The interesting detail is who benefitted. Newer and less experienced agents saw the biggest gains, while the most experienced staff saw very little improvement at all.
Another study, run with Boston Consulting Group, gave consultants access to AI for a range of realistic tasks. On tasks the AI was good at, they finished more work, faster and to a higher standard. But on a task deliberately chosen to be just outside what the AI could handle well, the consultants using it were noticeably more likely to get the answer wrong than those working without it. The researchers described this as a “jagged frontier”, and their full paper is worth a read if you manage a team.
Put simply, AI can make people better at some things and worse at others, and it isn’t always obvious which is which.
Where AI assistants tend to save real time
First drafts
Getting from a blank page to something you can edit is where most people feel the benefit first. Emails, job adverts, product descriptions, meeting agendas and internal announcements all fall into this bucket. The draft is rarely perfect, but editing is quicker than starting from scratch.
Summarising long material
Long email threads, meeting transcripts, reports and policy documents can be boiled down to the key points in seconds. This is especially useful when you’re catching up after time off or joining a project halfway through.
Reformatting and tidying
Turning notes into a table, rewriting something in a more formal tone, converting a list into a proper paragraph, or cleaning up messy data. These are the dull, fiddly jobs that AI is genuinely good at and that most people are happy to hand over.
Getting unstuck
Asking an AI to explain a spreadsheet formula, suggest a few ways to phrase a tricky message, or talk through a problem can save a lot of time spent searching or waiting for a colleague to be free.
Where they don’t save as much time as you’d think
Anything where accuracy really matters
AI assistants can sound completely confident while being completely wrong. They make up figures, misread documents and occasionally invent sources that don’t exist. For contracts, financial figures, technical specs or anything going to a client, you still need to check every detail, and that checking can eat up most of the time you thought you’d saved.
Work that depends on context the AI doesn’t have
An AI doesn’t know about the conversation you had with a client last week, the internal politics around a project or the reason a decision was made two years ago. It’ll fill those gaps with reasonable-sounding guesses, which can be worse than no answer at all.
Things you’re already very good at
If you can write a particular kind of email in two minutes, getting an AI to draft it and then rewriting it to sound like you might actually take longer. This matches what the customer support study found. Experienced people often gain the least.
Tasks that build your own skills
There’s a growing concern, especially for people early in their careers, that handing over too much too soon means you never properly learn how to do the work yourself. That’s hard to measure, but it’s worth thinking about before you let AI write every report.
A few practical tips for using AI at work
- Check your company’s policy first. Many businesses have rules about what data you can paste into AI tools, particularly anything involving customers or personal information.
- Treat the output as a draft, not a finished piece. Read it properly before sending or publishing anything.
- Give it more context. Vague requests get vague answers. Explaining who the audience is and what you need the result for makes a big difference.
- Verify facts and figures against the original source, every time.
- Notice where it’s actually helping. After a few weeks, you’ll have a good sense of which tasks it speeds up and which ones you’re better off doing yourself.
So, is it worth it?
For most office jobs, AI assistants are useful in the same way a good calculator or spellchecker is useful. They take the edge off repetitive work and help you get started quicker. Where things go wrong is when people treat them as an expert rather than a fast, slightly unreliable assistant. Use them for drafts, summaries and the boring bits, keep your own judgement for anything that matters, and you’ll probably find they save you a resonable chunk of time each week. Just don’t expect them to do the job for you, and dont skip the checking.
WordPress Theme: Lavoie
