AI moves into the workplace while rules and bills grow

AI is no longer a future project for business. It is becoming part of everyday life, and this week's news shows both how fast it is happening and how many questions are still open.

Work is changing now

A new test shows that the newest AI models are faster and more accurate than junior accountants at closing the books, and that they are far cheaper. However, they cannot replace the entire accountant's job. Pay is also affected: new AI services can help employers see who is paid too much or too little.

Microsoft's head of Switzerland, Catrin Hinkel, believes that the value arises especially when you change the collaboration between people and AI, rather than simply automating old routines. Investors believe in that track: Jefferies is raising its price target on Microsoft to 575 dollars after new Copilot features for businesses.

But AI can also be wrong with major consequences. Chatbots invent fake court cases that lawyers use without checking them, and new tools try to catch the errors before they reach a judge. The lesson is simple: the answer can sound convincing without being true.

The rules pull in different directions

In the US, the picture is split. Donald Trump would rather have a voluntary agreement with the AI companies than new laws, and according to him the agreement is "morally binding" but not legally binding. California is going the opposite way and has passed 13 AI laws, including one requiring that a human always be involved when employees are disciplined or fired. If the rules are tightened, according to Simply Wall St, it could give a tailwind to companies that sell compliance software. That is, however, an assessment and no guarantee.

The race requires enormous sums

Behind it all lies a battle for computing power. Leaked IPO documents show that Anthropic has committed to at least 518 billion dollars for computing power in the coming years. Revenue is growing rapidly, but so are the losses. At the same time, the company is betting 100 million dollars on an academy that will train 10,000 engineers to make AI work in companies.

In Europe, Mistral is aiming for 1 gigawatt of computing power by 2030 and is opening a hub in Munich focused on industrial AI. At the same time, the US lead over China is shrinking after DeepSeek's progress, according to Bloomberg Intelligence. And data can now be processed locally: Claude can run with data within India's borders.

What you can do

Keep an eye on two things. First: is it a human who makes the decisive decision in your company, and do you check the AI's answers when it matters? Second: where is your data, and which rules apply? Legislation is not moving at the same pace and in the same direction everywhere, so it is an advantage to have it under control before it becomes a requirement.