← Back to writing

AI 'workslop': Your team is using AI, but are they using it well?

On this page

What I keep seeing

I’ve used AI almost every day for about three and a half years to run my own businesses, and I now teach teams how to do the same. Over that time, the question at work has moved on from whether you should use AI. Your staff are probably using it already.

What worries me is what they’re producing with it, long reports that sound really good at first glance, but often don’t get to the point. The meaning isn’t clear, so whoever reads them wastes time working out what they mean, and there are mistakes in them too. If the person sending the report doesn’t have the expertise to judge it, or isn’t motivated to check it carefully, they’ll likely think it’s fine and just send it.

I haven’t found good UK-specific data on how widespread this issue actually is, but I can show you the US research alongside what I’ve seen myself, and then set out the method I use to avoid it creeping into my own work.

The rise of AI ‘workslop’

In September 2025, researchers from BetterUp Labs and the Stanford Social Media Lab published their research on this in Harvard Business Review. They surveyed desk workers in the US about AI-generated work that looks polished but is incomplete or missing important context, and they called it ‘workslop’.

The study is US-based, so we can’t necessarily assume that the UK looks the same, but it fits with what I see. I come across the same pattern all the time in LinkedIn posts and blog articles that sound polished but say very little. No matter what kind of writing it is, the problem is the same, and it’s worse in a long report than in a LinkedIn post, because it wastes far more of the reader’s valuable time.

Why it keeps happening

If your first experience of AI was using a free chatbot back in 2022, then you probably learned to type a quick question and take whatever came back. Unless someone has shown you a better way since, that’s likely still how you use it at work, and it’s probably how your staff use it too.

The problem with workslop is that it moves the effort from the person who creates it to the person who receives it. The sender might save themselves some time, but their colleague then ends up spending longer fixing the result. If your organisation rewards people for working fast and producing a lot, and nobody sends poor work back, anyone who finds themselves under pressure may use AI to produce workslop, because it’s the quickest way to look productive.

That’s why I don’t think training on its own is enough to fix the problem. You can teach someone a good method, but if it feels like more effort than the shortcut, they may well go back to the faster way.

The sender pays a price too, and your staff may not know it. When a colleague realises they’ve been sent workslop, it’s likely to change how they see the person who sent it.

What good use looks like

When you use AI tools effectively, they will absolutely save you a lot of time. For example, one system I built for a language school cut the time it takes them to send outreach emails from days to just eight minutes.

I follow five steps in my own work and teach the same ones in my training, which I call Decide, Think, Feed, Judge and Reuse.

1. Decide

Does the task need AI at all? If the work is fixed and repeatable, a template or a simple rule does it better. When I built an outreach system for a client, only the greeting changed from one email to the next, so I used templates for everything else. AI earns its place when the input changes each time and hence the response has to change with it.

2. Think

Work out the problem at hand before you open the tool. Bring your own idea, opinion or plan first. If you ask AI to give you ideas first, you risk getting the same average ideas that it would give anyone else.

I rarely ask AI to come up with ideas. The only time I do is at the very start of a problem, when I have no idea where to begin. I also use it as a sounding board, but I know my field well enough to judge which of its suggestions make sense. Most of the time, I voice dictate my rough thinking about the problem and use the model to make it clear and well organised.

3. Feed

The model only knows what you give it, so tell it who the work is for and give it the background and the documents it should draw on.

I talk to it with voice dictation rather than typing (I use Granola). Unless you’re a fluent touch typist, typing adds friction, so you tend to keep things short. With voice dictation, you can talk around the problem for as long as you need to, and the model handles a long, unstructured voice dump easily.

As well as your own thinking, you can ‘feed’ the AI tool with examples of what good work looks like. When I work outside my own field, I find an example of excellent work first and give it to the model as a reference. For example, before I made my first Facebook ad, I collected high-performing ads to use this way.

4. Judge

Hold the output to the standards of your field. You can only spot weak work in an area you know well, so your own expertise matters more as the tools improve. The rule I give teams is that if you can’t judge it, you shouldn’t send it.

5. Reuse

When you do a task often, I recommend setting up a Claude Project or Custom GPT that holds the context, examples and standards, so you don’t start from nothing each time.

I have loads of these, but one of my favourites is for my LinkedIn posts. It knows my writing style and preferences, and I can just dictate each new idea straight into it and have it write the post in my own voice.

What you can change as a manager

Your team can learn these five steps in a single training session, but they’re much more likely to keep using them if your organisation backs them. People pay attention to what gets rewarded and checked, so if you praise fast output and never question what’s in a report, the shortcut starts to look like the sensible choice, whatever they learned in training. As a manager, you decide both of these, so you’re in the best position to change them.

The first thing I’d do is make the good method as quick as the shortcut. Set up shared Projects for the tasks your team does most, with the context, examples and standards already loaded, so even someone in a hurry gets decent output without having to supply all of that themselves. When the careful way takes no extra effort, there’s much less reason to skip it.

I’d also make it clear that whoever sends a piece of work owns it. If a report reaches you full of errors, send it back to the person who wrote it, because if you fix it yourself, the shortcut has worked for them. After this happens a few times, people start checking their work before they send it.

It’s worth talking to your team about this too. They may not realise that colleagues notice workslop and think less of the people who send it, and knowing that gives them their own reason to take more care.

Finally, judge work by what it achieves rather than how much of it there is. A two-page report someone has thought about in depth is typically worth more to you than a twenty-page one that nobody checked. Once your team sees that you value the shorter, clearer version, they’ll have far less reason to pad things out.

Skills and incentives together

To get your employees using AI effectively, you need a method they can follow and a workplace environment that rewards careful work rather than fast work. The two very much depend on each other. A good method can feel like extra effort until it becomes a habit, and that habit forms much faster when the careful way is set up to be quick and poor work gets sent back to the person who wrote it. If you give your team the method without changing the incentives, you’re likely to keep getting workslop.

When both are in place, AI starts saving your team time rather than just handing a bunch of extra work to whoever has to read what it produces.

This is what I teach in my team training. In my sessions, your team uses the five steps on their own real tasks, so they leave knowing how to apply them to their actual work. If you want your team working this way, book a free 30-minute call and we’ll talk through where it would help most.

Sources