Stop Bitching about AI (and do something about it)

The "Stop Bitching" Part

A lot of people are against AI. Blaming the tool instead of the people and entities who own, control and run it is the easy option, and it changes nothing. It directs anger at the thing you can see rather than at those making the decisions.

Those who benefit at society's expense have no reason to correct you. They stay out of view, collect the returns, and let the public argue with a technology that cannot be held to account.

That anger does not trouble those with media reach, political access and money. It does cost you something, though. It crowds out the uses of AI that would help people, and it leaves the decisions that actually matter to those who turn up to make them.

When people worry about AI and jobs, they usually mean something immediate: whether their employer will still need them, whether their children will find a first job, whether their pay will cover the bills, and whether learning another skill will be enough to maintain and improve their standard of living.

Those are reasonable concerns. The International Labour Organization's 2025 assessment found that one in four jobs worldwide is potentially exposed to generative AI. That does not mean one in four jobs will disappear. It means the technology could perform some of the tasks within them. The researchers expect changes to jobs to be more common than their complete replacement.

That distinction matters, but it does not pay the mortgage of someone whose job does disappear. Nor does keeping a job help much if the result is less pay, no future prospects and worsening employment conditions.

There is also the question of who gets the money. The World Inequality Report 2026 puts three-quarters of global wealth in the hands of the richest tenth, while the poorest half holds just 2 per cent. AI did not create that imbalance. But we should be asking whether the way it is introduced will make it worse.

Problems with automated management already exist. Research published in the Columbia Law Review describes how some platform companies use workers' data to set personalised, variable and unpredictable pay (somewhat similar to dynamic pricing in retail). These systems show how technology can give employers more control while leaving workers less able to understand or challenge the terms.

AI is a tool, but its design and the targets it is given matter. A company can use it to help employees do their work, or to cut staff and push the people left behind harder. The company chooses what it is trying to achieve. Government decides which practices are allowed, how profits are taxed, and what support exists for people who lose out.

Those decisions are made by people. Some of them are people you can vote for; others can be constrained by the rules elected governments make.

An Alternative Scenario

AI, like the inventions that came before it, will change the way work is done. I explored that history in The AI Revolution — History Repeating Itself — On Economics. What matters here is that the benefits and costs should be shared across society, rather than leaving the people whose jobs change to carry the burden alone.

We have changed the rules around work before. Britain's 1833 Factory Act restricted child labour in textile factories and created an inspectorate. Four inspectors could not adequately cover thousands of mills, and the law was widely evaded. Writing down a right was only the beginning; enforcing it mattered too.

The eight-hour working day was a major demand of the trade union movement before it became the basis of the ILO's first convention, covering industry, in 1919. It did not arrive because machinery decided people deserved more time at home. Workers organised, governments negotiated, and rules changed.

Workers do not have to agree on every detail of an economic system to ask for a decent living, a say in their working conditions, and a fair share of the improvements they help produce.

What follows is the kind of programme an elected representative could support. If I left something out, add it and send the list to your candidates as your own.

Sharing the gains

The richest people and the companies benefiting from AI should pay a fair share of tax. That means closing unjustified loopholes, enforcing the law against evasion, and reviewing how income from investments is taxed compared with income from work. Taxes on very large fortunes and inheritances should be part of that discussion, with clear thresholds that distinguish them from ordinary family savings. The purpose is to help fund the services and support on which society depends.
Tax policy should not make replacing a person artificially cheaper. A special tax on AI sounds straightforward, but it could discourage useful innovation and be difficult to administer. The IMF argues against a blanket AI tax, while recommending stronger taxation of capital income and a review of tax breaks that encourage companies to replace workers. The people and businesses gaining from the change should help pay for the transition.
Better productivity should improve people's lives. If AI lets a team produce more in the same time, the gains should be part of negotiations over pay, working hours, paid leave and training. Profit-sharing and employee ownership are other options. A shorter working week should mean less time at work, with pay protected, rather than the same workload squeezed into fewer days or a reduction in wages and employment benefits.
Basic necessities should remain within everyone's reach. Clean air and water, a secure home, decent healthcare, education, training and internet access should be treated as foundations of a functioning society. Losing a job should not mean losing the means to find another one, or leaving a family without essential care.

Supporting workers and their families

Work matters for more than the money it brings in. It can provide structure, confidence, relationships and a sense of contributing. In Magnifica Humanitas, paragraph 149, Pope Leo XIV describes work as a path to development and personal fulfilment, and argues that financial assistance alone is not a sufficient response to poverty.

That does not make an unemployed person less valuable. It means support should give people choices and opportunities, rather than simply enough money to get by.

People who lose work need income they can live on while finding their next job. Support should cover the period between jobs and include people whose earnings fall sharply. Basic income is one proposal in this debate, but any payment must be judged on what it provides and how it is funded. Monetary support should sit alongside access to services, training and decent work.
Retraining must be affordable and connected to real opportunities. Someone with rent, children and bills cannot always stop earning to attend a course. Paid learning time, help with childcare and travel, and training developed with employers should be available. Publicly funded programmes should be assessed on whether participants find suitable work and improve their earnings, rather than simply on how many certificates are handed out. Support must not quietly write off older workers. Someone made redundant at 55 is often the hardest to place and has the least time to recover financially. Programmes should be judged on how they directly affect the people they are designed to help.
Workers must have a say before changes are imposed. Employers introducing systems that affect jobs, pay, hours or monitoring should explain the likely effects and consult staff and their representatives beforehand. Workers must be able to join unions and bargain collectively without retaliation. Those discussions must cover how the technology is used as well as what people are paid.
An automated decision must be open to explanation and appeal. Nobody should lose shifts, pay or a job without being able to find out why and challenge an error. A named person must be accountable for important employment decisions. Monitoring must have clear limits, and targets must allow for breaks, safety and the fact that the person doing the work is human.
Labour authorities need the staff and powers to enforce the rules. They must be able to inspect the records behind automated employment decisions, investigate complaints and impose meaningful penalties. Workers must have a practical way to report abuse without risking their livelihood. The lesson from those first factory inspectors still applies: a right that cannot be enforced offers little protection.
Protection must follow the person, not how they are categorised. Some of the people most exposed to these systems have the least cover: gig and platform workers, people on zero-hours or on-call contracts, agency and temporary staff, and anyone labelled self-employed while working for a single company. These are also the people already managed by software that decides their hours and their pay. If rights covering automated decisions, monitoring and consultation apply only to employees, companies will reclassify people to escape them. The test should be whether someone's work is directed and measured by the company, not what the contract happens to call them.

Public money and public accountability

Grants, tax breaks and public contracts must come with conditions. Where companies promise jobs, apprenticeships or training in exchange for public support, the agreement must contain numbers, deadlines and published results. It should allow money to be recovered when commitments are broken. A press release promising jobs should not be enough to obtain a subsidy.
Essential technology must serve the public. Where control of vital digital infrastructure would leave public services dependent on a few companies, public ownership, a public stake or binding access conditions should be considered. The choice should depend on the service and the risks involved. Any public stake must also come with transparent management and accountability to citizens.
If a rescue is necessary, the public must be protected. Government support during a financial crisis should safeguard households, essential services and the wider economy. A rescue should come with independent oversight, restrictions on executive rewards and, where appropriate, an ownership stake or repayment terms that let the public share in a recovery. Where there is evidence of fraud, it must be investigated and those responsible held to account.
When the state uses automated systems, rules must apply. Automated systems already touch benefit claims, tax enquiries and eligibility for services. When they fail, people are told (for years) that the computer is right. Any automated decision affecting someone's income or access to a public service should be explained in plain language, have a named official accountable for it, and include an appeal that a person can actually handle. Government must publish where these systems are in use before they are switched on.
People must always be able to reach a person. Banks, utilities, insurers, healthcare providers and government services must offer a route to a human being within a reasonable time and at no extra cost. A chatbot is fine as a first step. It must not be the only option.

Countries working together

These proposals should not stop at a border. Countries should cooperate on corporate taxation, labour standards and the oversight of powerful technologies. Trade agreements should include enforceable protections against forced labour, child labour and other serious abuses. Products and services made using abusive labour should face additional taxes or levies to reduce the price advantage gained through exploitation and make them less attractive to consumers. Consumers should also be given enough information to make informed choices about the products and services they buy.

Tariffs and import restrictions need scrutiny too. As the World Trade Organization explains, tariffs can raise prices for consumers. Any measure should have a clear purpose, evidence behind it and a review of who will pay. Protecting workers should not become an excuse to protect a favoured company from competition.

Identical technology, different decisions

Picture an employer introducing AI under the conditions described here.

Staff hear about the change before it happens. Their representatives help negotiate how it will affect workloads, pay and hours. Some employees learn to use the new tools during paid working time. Someone whose role disappears receives support while moving into another job or suitable training. Apprentices still get a chance to learn. The company becomes more productive, pays its taxes and shares some of the gains with the people doing the work.

Compare this to the same change announced after the redundancies have been decided, with the savings benefiting shareholders and senior management while the costs fall on the families affected.

The technology would be the same in both cases. The difference is the decisions around it.

The "Do Something about it" Part

This is where it stops being an opinion and starts being a decision.

Tax, employment rights and social support are shaped by elected governments and legislatures. Councils and regional authorities may also influence local training, public contracts and incentives for employers.

Three things you can do before the next election:

Find out what is being proposed. Look for plans on employment rights, retraining, taxation and support for people who lose work. Check what major employers are receiving in public support and what they promised in return. Where the information is not published, ask your representatives for it. Take the list to a union, a community group or a public meeting and ask what is already being done.
Find out where your candidates stand. My piece Where Do They Actually Stand? offers prompts you can use with an AI chatbot to investigate a named candidate and a topic you care about. Switch web search on and ask for evidence: voting records, published policies and statements in context. Follow the links and check them yourself. If there is no evidence of a position, that is something to ask the candidate about. Then make up your own mind.
Ask one question and publish the answer. Pick something concrete: "Will you fund paid retraining for people whose jobs disappear?" "What conditions would you attach to employers receiving public money?" "Will workers be able to challenge automated decisions about their pay?" Ask at a public meeting, by email or on social media. Share the answer, or say that no answer was received. When a representative follows through, a like or a repost helps others notice it too.

Then vote. Choose the candidates whose records and proposals best match what you want, and keep asking what they have done after the election. Voting gives you a say in who holds power; organising and following up help you use that say between elections.

If this is useful, pass it on. Send it to a colleague worried about their job, someone looking for their first opportunity, a union representative or a community group. Send it to a candidate and ask which of these proposals they would support.

Arguing about AI online changes little by itself. Sharing something people can put to their representatives, and voting for someone who will act on it, is the part that counts.

Follow This, That and (Maybe), the Other:

Search This Blog

TT(O)M