‘Labour’s workplace AI rules need a worker challenge right’

Artificial intelligence is moving into British workplaces through the side door. It arrives as scheduling software, productivity scoring, recruitment filters, absence monitoring, call analysis and automated recommendations to managers. Each system may look like a technical upgrade. Together, they can reshape power at work before employees know what has changed.

Labour’s workplace monitoring consultation is therefore well timed. The government is asking how monitoring can be fair, transparent and responsible. But transparency alone will not protect working people. An employee can be told that an algorithm influenced a decision and still have no practical way to correct bad data, challenge a misleading inference or stop the consequences while a dispute is reviewed.

Labour should turn the consultation into a worker challenge right that applies whenever automated monitoring or decision support can affect pay, hours, hiring, promotion, discipline, dismissal or access to work.

That right should start with notice in plain English. Workers and unions need to know what a system measures, which decisions it can influence, where its data comes from and who is accountable for the result. A generic privacy policy is not enough. Notice must arrive before deployment, not after harm.

READ MORE: ‘Britain must control the value created from our own data – or someone else will’

Second, employers should preserve the evidence behind disputed decisions. When a worker challenges a score or recommendation, the relevant inputs, outputs, rules and human interventions should be frozen long enough for meaningful review. Without an evidence trail, appeals become arguments against a disappearing screen.

Third, every system needs a named human owner with authority to change the outcome. “The computer says so” cannot become a modern version of managerial impunity. The responsible person should be able to correct data, reverse a decision and explain what will change to prevent repetition.

Fourth, workers need temporary protection while serious disputes are investigated. An automated flag should not trigger dismissal, lost shifts or withheld pay before a human review. The greater the potential harm, the stronger the presumption that the decision pauses.

Fifth, individual appeals must connect to collective voice. A recurring error affecting many workers is not merely a customer-service problem. It is evidence about job design, discrimination, workload and bargaining power. Unions and elected worker representatives should receive anonymised summaries of challenges, reversals and unresolved patterns.

Finally, corrections must travel downstream. If an inaccurate score has already influenced scheduling, performance records or future opportunities, fixing the original entry is insufficient. Employers should identify every system and decision that relied on it, repair those consequences and record the lesson for future procurement.

Become a friend of LabourList and join our community. Our friends support our vital non-factional work and get access to exclusive content and events. 

This proposal fits the argument LabourList has documented around guaranteed hours: rights lose their force when implementation leaves workers exposed to one-sided flexibility. The same principle applies to workplace AI. A nominal right to information is weak if the employer still controls the evidence, the timing and the remedy.

It also addresses the danger of zero-rights employment in new technological form. Freelancers, agency staff and workers managed through platforms are often the people most exposed to automated allocation and monitoring, yet least able to challenge it. Labour’s rules should follow the practical relationship of control, not permit companies to outsource accountability through contractual labels.

Training matters too. The government’s AI upskilling evidence review rightly focuses on helping people use new tools. But workers also need the skills and protected time to question them. Training should include how to identify a consequential AI decision, preserve evidence, raise a concern and participate in a review without retaliation.

Labour can test this quickly. Require large employers and public bodies to run a 30-day challenge pilot before scaling any consequential monitoring system. Publish the number of challenges, reversals, unresolved cases and average time to remedy. Bring unions and worker representatives into the evaluation. Procurement should stop if the organisation cannot explain or correct the system’s decisions.

The political choice is not between innovation and workers’ rights. It is between innovation that earns consent and automation imposed without recourse. Labour’s historic purpose is to give working people power over the conditions that shape their lives. In the age of workplace AI, that must include the power to challenge the machine and make the correction stick.

Subscribe here to our daily newsletter roundup of Labour news, analysis and comment– and follow us on TikTok, Bluesky, WhatsApp, X and Facebook. You can also write to our editor to share your thoughts on our stories and share your own. The best letters are published every Sunday.

• • SHARE: If you have anything to share that we should be looking into or publishing about this story – or any other topic involving Labour– contact us (strictly anonymously if you wish) at [email protected].
  • SUBSCRIBE: Sign up to LabourList’s morning email here for the best briefing on everything Labour, every weekday morning.
  • BECOME A FRIEND: If you enjoyed this, why not consider becoming a Friend of LabourList? Help sustain our journalism, and of course Friends do get benefits…
  • PARTNER: If you or your organisation might be interested in partnering with us on sponsored events or projects, email [email protected].
  • ADVERTISE: If your organisation would like to advertise or run sponsored pieces on LabourList‘s daily newsletter or website, contact our exclusive ad partners Total Politics at [email protected].