When AI arrives at work, it can feel as if the decision has already been made somewhere else: by a software supplier, a senior manager, a procurement team or an enthusiastic colleague who has found a new tool. The people expected to use it every day may only meet it once the training invite lands.
A new report from the Institute for Public Policy Research, backed by the TUC, argues that this is the wrong way round. It says workers need more power to shape how AI is introduced, because the same technology can produce very different results depending on how it is used.
That makes this more than a trade union story. It is a practical workplace story for anyone whose job now involves AI-written drafts, automated summaries, ranking systems, scheduling tools, monitoring software, customer-service bots, coding assistants, HR screening or “productivity” dashboards.
The useful question is not simply whether AI is good or bad. It is: what is this particular tool being asked to do, who benefits, who carries the risk, and what choices do workers have if it changes their job?
What the report says
The Guardian reports that the IPPR is calling for a package of measures to give employees more influence over workplace AI adoption. The thinktank describes this as a “pivotal moment” for work, because AI can be used to support people, reduce the quality of work, or replace tasks and roles altogether.
The report draws a useful distinction between three outcomes. AI can augment work, meaning it helps people do tasks better or faster. It can degrade work, for example by increasing surveillance, speeding up targets or narrowing human judgement. It can also displace work, where tasks or jobs are removed.
Those categories are simple, but they are a good starting point for ordinary workers. A tool that drafts a first version of a routine email may feel helpful. A system that scores call-centre staff minute by minute may feel very different. A chatbot that helps a team answer common questions is not the same as a plan to cut that team without proper support.
The IPPR’s recommendations include a duty on employers to consult workers about AI adoption, stronger routes for worker representation, and a portable benefits idea that could support people with training, representation or protection as jobs change.
Why this matters in real workplaces
AI is often introduced as if it is mainly a software upgrade. In practice, it can change who makes decisions, how work is measured, what counts as good performance, and which mistakes are blamed on people rather than systems.
For many UK workers, the first effect may be subtle. A manager may ask for more output because AI is “saving time”. A team may be told to use a tool that records more information than expected. A role may quietly shift from doing a task to checking machine-generated work. A junior colleague may lose the chance to learn by doing because AI now produces the first draft.
That is why consultation matters. It gives people a chance to ask about consequences before a system becomes normal. It also gives employers better information. The people closest to the work are often the first to spot where an AI tool is useful, where it is unreliable, and where it creates pressure that does not show up in a sales demo.
ManyHands has covered this wider jobs debate before, including why the AI jobs panic needs a reality check. The point is similar here: the future of work will not be decided by one dramatic headline. It will be shaped by thousands of smaller decisions about how tools are bought, tested, managed and challenged.
Five questions to ask before a workplace AI tool arrives
What problem is it meant to solve? A clear answer matters. “Using AI” is not a purpose. Is the tool supposed to reduce admin, improve accuracy, speed up customer replies, spot risks, summarise meetings, train staff or cut costs? If the goal is vague, judging success becomes much harder.
What data will it see? Workers should know whether the system can access emails, calls, chats, files, customer records, HR data, performance metrics or private notes. This is especially important where tools handle personal data, sensitive work, trade secrets or client information.
Will it monitor people? Some workplace AI is genuinely assistive. Some is also a management tool. If it tracks speed, tone, clicks, keystrokes, conversations, attendance, customer ratings or error rates, staff should know what is being measured and how it will be used.
Who checks the output? AI systems can be fluent and wrong at the same time. If a tool drafts advice, classifies a customer, summarises a meeting or flags a worker for review, there should be a named human process for checking it. People also need to know whether they can challenge an AI-assisted decision.
What happens to the time saved? This is the question that often decides whether AI feels helpful or threatening. If a tool removes boring admin and creates space for better work, staff may welcome it. If every saved minute becomes a higher target, fewer colleagues or more surveillance, trust will drain away quickly.
What workers can do now
Most people will not be writing AI policy from scratch. But workers can still ask practical questions early, especially when a new tool is being trialled. Ask for the purpose, the data access, the limits, the review process and the plan for training. If your workplace has a union, staff forum or consultation group, raise it there too.
It is also worth keeping examples. If an AI tool saves time, note how. If it makes mistakes, note the pattern. If it changes your workload, note what has changed. Concrete examples are more useful than general anxiety, and they help separate good tools from bad rollouts.
For people using AI directly, the everyday checks still apply. We have written before about what to consider before giving an AI assistant access to your computer. At work, the same basic idea becomes a shared responsibility: access, permissions and oversight should be understood before the system is trusted with important tasks.
What employers should not ignore
Employers have a strong reason to involve staff beyond fairness. AI projects often fail when they misunderstand real workflows. A tool may look impressive in a demo but struggle with messy handovers, local judgement, unusual cases, older systems or the quiet knowledge that experienced staff use every day.
Bringing workers in early can make adoption more honest. It can show where automation is sensible, where human judgement must stay central, and where extra training is needed. It can also reduce the chance that staff quietly work around a tool because they do not trust it.
There is a difference between asking workers for feedback after a decision and giving them a meaningful say before the decision hardens. The IPPR report is pushing for the second version. Whether or not its recommendations become law, that is a useful standard for any workplace trying to introduce AI responsibly.
The practical takeaway
AI at work should not be treated as a magic productivity layer that can simply be dropped on top of people. It changes tasks, responsibilities, privacy expectations and sometimes job security. That means workers need clear information, time to respond and routes to challenge poor use.
The calmest way to approach workplace AI is neither panic nor blind enthusiasm. Ask what the tool is for, what it can see, how it will be checked, whether it monitors people, and what happens if it changes the shape of the job.
If AI is genuinely there to help, those questions should make the rollout better. If nobody can answer them, that is the warning sign.
Sources: The Guardian and IPPR.
