A manager once decided who worked the late shift, who needed extra training and whether an employee’s performance was satisfactory. Increasingly, some of those decisions begin with software. An algorithm can allocate tasks, set delivery routes, rank performance, recommend shifts or flag workers whose activity falls outside a target range.
- What algorithmic management looks like
- The appeal: efficiency and consistency
- When the target becomes the job
- The hidden assumptions inside a model
- Can a human manager really overrule the algorithm?
- The UK’s workplace monitoring consultation
- What a fair system would need
- Management by dashboard: a question of power
- The UK consultation has closed: what happens next?
- Sources and further reading
This is known as algorithmic management: the use of automated systems to organise, monitor or evaluate people at work. It is associated with delivery platforms and warehouses, but it is spreading into offices, retail, transport, customer service and remote employment. The software may not formally replace a manager. It can nevertheless shape the choices available to both managers and workers.

What algorithmic management looks like
The simplest examples are automated scheduling and task allocation. A system may use predicted demand, staff availability and historical performance to decide how many people are needed and when. In a delivery business, software can assign jobs, calculate routes and estimate how long each task should take. In a warehouse, handheld scanners and performance dashboards can record the pace and accuracy of work.
More advanced systems combine information from multiple sources. Attendance records, customer ratings, location data, task completion times and quality checks may feed into a performance profile. Artificial intelligence can then identify patterns, predict demand or recommend management action.
Not every automated system makes decisions about people. Some simply provide information. The distinction becomes important when an algorithm’s output affects pay, hours, access to work, disciplinary action or promotion. At that point, the system is no longer just helping to organise a process; it is influencing someone’s livelihood.
The appeal: efficiency and consistency
Employers have understandable reasons to automate parts of management. Software can coordinate thousands of shifts, respond quickly to changes in demand and identify operational problems that would be difficult to spot manually. It may reduce administrative work and help distribute tasks more consistently.
For workers, well-designed systems can sometimes offer benefits too. Scheduling software may make shifts available earlier, allow staff to state preferences or help allocate work more predictably. Data can reveal when teams are overloaded and support better staffing decisions.
But consistency is not necessarily fairness. A system can apply the same rule to everyone while disadvantaging people whose circumstances differ. A worker with a disability, caring responsibilities, a longer travel route or a different pattern of customer interactions may be judged against assumptions that do not fit their situation.
When the target becomes the job
A common danger is that a measurable target becomes a substitute for the actual purpose of work. If a warehouse system rewards the number of items processed per hour, workers may feel pressure to prioritise speed over careful handling. If a call-centre dashboard rewards short calls, complex customer problems can become an obstacle to a good score.
In platform work, workers may receive ratings or task offers through systems whose rules are difficult to see. A lower acceptance rate, customer complaint or missed target may affect access to future work. Even when a company describes the system as advisory, workers can experience its recommendations as instructions if ignoring them carries a penalty.
This can create a form of management without a visible manager. The worker receives an instruction from an app, a warning from a dashboard or a shift allocation generated by a model. There may be no obvious person to ask why the decision was made.
The hidden assumptions inside a model
Algorithms are built from data, objectives and design choices. Historical records may contain bias or reflect conditions that no longer apply. A model trained to identify high-performing workers may reproduce past management preferences, even if those preferences were unfair. A system predicting who is likely to leave may treat patterns associated with caring responsibilities or disability as a business risk.
Data can also be incomplete. A delivery worker may take longer because a building has difficult access. A call-centre adviser may spend extra time helping a vulnerable customer. A warehouse employee may be assigned heavier or more complicated tasks. If the system measures only time and volume, it may record the outcome without understanding the circumstances.
There is a further problem of feedback. If a worker is given fewer shifts because an algorithm predicts lower availability, the resulting lack of work may reduce their future performance data. The system can then treat the consequences of its own earlier decision as confirmation that the original prediction was correct.
Can a human manager really overrule the algorithm?
Employers often say that software supports decisions rather than making them. That distinction is meaningful only if the human decision-maker has enough information, time and authority to question the system. If managers routinely accept a score because it appears technical or because they are under pressure to follow it, human oversight can become little more than a signature.
The UK Information Commissioner’s Office has emphasised that where monitoring tools support decisions about workers, human involvement must be meaningful. A manager should actively examine the recommendation, consider other evidence and have the competence and authority to disagree. Solely automated decisions with legal or similarly significant effects can trigger additional UK GDPR safeguards.
That means an employee should not be left with a vague statement that “the system decided”. There should be an explanation of the relevant data, the criteria applied and a practical way to challenge mistakes. If a decision affects pay, hours or continued employment, the ability to obtain meaningful human review becomes especially important.
The UK’s workplace monitoring consultation
The UK government’s 2026 consultation on workplace monitoring technologies recognised that modern monitoring extends beyond observation. It includes systems that collect information about workers and use it to inform or make decisions, including automated decision-making and algorithmic management. The consultation asked whether transparency, worker engagement and accountability need to be strengthened.
The consultation closed on 30 September 2026. It is a policy process, not evidence that a new statutory code or additional legal duty has already taken effect. Existing data-protection and employment obligations continue to apply, and employers should not wait for a future reform before considering whether their systems are fair and proportionate.
International developments also show regulators paying closer attention to workplace data. A November 2026 employment-law update reported enforcement actions in several European countries involving unlawful location tracking, in-vehicle camera surveillance and access to employee email accounts. The details differ by jurisdiction, but the direction is clear: workplace technology is increasingly being assessed through the lens of worker rights, not just business efficiency.
What a fair system would need
A responsible algorithmic-management system should have a clearly defined purpose and collect only the information needed for that purpose. Workers should be told what data is used, how recommendations are generated at a meaningful level, who can see the results and what decisions may follow.
Employers should test systems for accuracy and unequal effects, review them when working conditions change and keep a record of significant decisions. Workers need a route to correct inaccurate information, explain exceptional circumstances and appeal consequential outcomes to a person who can genuinely change the result.
Targets should also be examined for what they encourage. A metric that improves throughput while increasing injuries, mistakes or customer complaints is not a successful measure of performance. A system should be judged against the real quality and safety of work, not merely the quantity of data it can produce.
Management by dashboard: a question of power
Algorithmic management is not inherently harmful. Used carefully, it can make complex operations more organised and help identify problems that managers might otherwise miss. But the technology changes the relationship between employer and employee when decisions become difficult to see, explain or challenge.
The central risk is not simply that a machine becomes a boss. It is that organisations can hide choices about priorities, acceptable performance and workplace discipline behind a system that appears neutral. An algorithm does not remove human judgement; it encodes judgement in its design and in the way its outputs are used.
As software takes on more managerial functions, accountability must become clearer rather than more obscure. Workers should know when an algorithm is influencing their working lives, understand the basis of important decisions and be able to challenge them. Efficiency may be the sales pitch. Fairness, transparency and human responsibility are the tests that matter.
The UK consultation has closed: what happens next?
The Department for Business and Trade’s Make Work Pay consultation on workplace monitoring technologies was issued on 8 July 2026 and closed on 30 September. It considered possible safeguards for monitoring systems, including principles for responsible use and the role of worker voice. The consultation document was explicitly marked as not government policy and subject to consultation: it should not be read as a new set of rules already in force. In its response, Acas welcomed a principles-based approach but called for more evidence about the scale and effects of monitoring, meaningful employee consultation and further consideration of enforceable safeguards. Its submission also warned that algorithmic systems can intensify work and amplify bias if introduced without proper worker involvement. The next stage is policy development, not the arrival of a completed statutory code.
Sources and further reading
- Department for Business and Trade — Workplace Monitoring Technologies Consultation.
- ICO — Automated processes in worker monitoring.
- Ius Laboris / Legal 500 — Workplace Data Privacy and AI Update, October 2026.
- TechTarget — Ethical AI surveillance, 1 September 2026.
- GOV.UK — Monitoring staff at work.
- Acas — Response to the Make Work Pay workplace monitoring technologies consultation, 30 September 2026.
