We've spent years teaching machines how to learn. Remember that people have only had a few years to learn how to work with them.
By Hayley Watson, PHD – Head of AI at Corndel
Chances are you're already feeling the effects of AI workslop.
We've all seen it. It’s about 15% of the work you receive. The report that uses thousands of words to say almost nothing. The polished presentation that lacks any real insight. The email response that sounds impressive but leaves you with more questions than answers.
Broadly speaking, AI workslop is technically complete work that lacks the substance to move something forward. But, it can be difficult to define precisely, because workslop is something you feel. It's that sense of the absence of a human in the room.
Research from BetterUp suggests employees spend almost two hours effectively re-doing each instance of workslop; twenty minutes more than if the work was done properly to begin with. This productivity cost often goes under the radar, but it can run into millions of pounds for large organisations.
There are human consequences too. Employees report feeling annoyed, frustrated, confused and even offended when they receive workslop. That's understandable; it can signal a lack of respect for time, expertise and so on. This is where relationships and collaboration start to unravel.
But what about the cause of AI workslop? It's mostly treated as a technology problem.
I see it as a human one.
Most people are still learning how to work effectively with AI. The tools are impressive, but there is a big gap between generating output and applying the judgement to know whether it should be shared, trusted or acted upon.
That level of quality, which businesses should treat as a non-negotiable, can only come from human thought.
The Pursuit of Productivity
Many organisations name 'productivity' as one of their main goals when it comes to AI adoption.
Faster processes and greater output are valid ambitions and AI absolutely can help achieve them when it's used right.
But productivity is not the only outcome that matters. Not everything can be made into a number, and the more qualitative opportunities are getting lost in the process.
Organisations should also be asking:
- How do we improve quality?
- How do we create new capabilities?
- How do we improve experiences?
- How do we drive innovation and growth?
Faster is not always better. The pressure to move quickly can get in the way of people developing the judgement they need to use these tools well in the long run.
AI shouldn't come between your people
One of the most overlooked impacts of poor AI use is on workplace relationships.
When you receive a response that feels completely machine-generated, with little evidence of human reflection or effort, you can feel... cheated.
It's not the fact that AI was used. Most of us are frequently and openly using AI tools today (though more of that openness and honesty sometimes wouldn't go amiss). The issue causing the frustration and offence is often when it feels like the other person has checked out.
Working relationships that you trusted and valued can get awkwardly out of sync. Managers find themselves questioning judgement. Colleagues are unclear on what's really been reviewed, challenged or endorsed.
The disconnect slows everyone down.
Addressing workslop, and the deeper faultlines it points to, is not just a matter of guidelines and governance. It requires people to talk to each other.
Have the courageous conversations. They can sometimes be hard, but they're worth it to rebuild the human collaboration that makes us great at what we do in the first place.
Leaders need to model what good looks like. Managers need to be comfortable having honest discussions about quality. Teams need the psychological safety to experiment, learn and sometimes get things wrong with AI, without feeling the need to cover it up. Empathy is a crucial tool on the road to regaining quality.
Patience from the top
Organisations need to recognise that people are still learning.
We've spent decades developing AI technologies. Most employees have only been working alongside them for a few years now (if that!). Yet we act like everyone should know exactly how to use them. They don't, and that's fine. It should be expected.
If your objective with AI is simply to go faster, problems like workslop will spread under the pressure. Instead, create the conditions for people to learn effectively, build confidence and develop good judgement over time.
The antidote to AI workslop isn't to get 'better AI'. It's patience, empathy and conversations that help people learn. It’s investment in skills training.
Set the right pace for evolution, for learning and development, have empathy for individual learning journeys, and be open to discussions about what we consider good use of AI.
These are the steps that will help your business avoid the deeper erosion of trust, quality and collaboration that comes with AI workslop. Not only will they help you avoid the productivity downslide, but you'll improve the quality of work and thus business outcomes, as well as quality of working life.
Hayley Watson, PhD