New Guide Shows How to Make AI at Work Actually Help You
Most workplace AI fails because of people, not tech — here's the fix
A team of researchers has published a 44-page guide on how to set up workplaces where people and AI actually work well together. It isn't a new gadget or app — it's a blueprint. The framework asks organisations to think about five things at once: the person, the AI system, the task itself, the company, and the wider world it operates in. In plain terms, it says that buying clever software is the easy part. Deciding how people will live with it day to day is the hard part.
Why should you care? Because AI is arriving at your job whether you asked for it or not. A tool can be technically brilliant and still make your work worse — if it interrupts you constantly, hides how it reached a decision, or leaves you unsure whether you're allowed to overrule it. The guide argues those are design choices, not accidents. Someone has to decide how much the AI decides alone, how clearly it explains itself, and who is accountable when it gets something wrong. Get those wrong and staff quietly stop using the tool.
To make it concrete, the authors built an example around a factory robot — a 'cobot', short for collaborative robot, meaning one designed to share a workspace with humans rather than sit behind a safety cage. They show how a vague wish like 'the robot should help the worker' gets translated into specific features: who does which step, how the robot signals what it's about to do, and how the pair recovers when something goes wrong. They then tested it in a case study to see whether the design held up.
The honest limitation: this is advice, not a product. There's no software to install, no price tag, no regulator enforcing it. Its value depends entirely on whether managers read it and act. Still, it gives you useful questions to ask if AI lands in your workplace: Who is checking its output? Can I override it? What happens when it's wrong? Those questions are worth asking now — before the system is bought and the decisions are locked in.
- The guide says successful workplace AI depends as much on people's experience as on the technology itself
- It covers five areas at once: the human, the AI, the task, the organisation, and society — not just the software
- A factory cobot (a robot designed to work alongside people) shows how vague goals turn into real design choices
Why It Matters
It gives you the questions to ask before AI arrives at your job — and before the bad decisions get locked in.