AI is spreading in companies faster than governance

There is a common thread in this week's business and economy news: AI is digging deeper into companies' daily operations, while governance struggles to keep up.

One clear sign is that half of Danish companies, according to a new survey, don't know how employees actually use artificial intelligence day to day. At the same time, an American survey shows that one in four adults now uses AI daily, and that many let AI agents act without asking first. That picture fits well with the fact that IT managers now have to scrutinize their cloud agreements, because AI usage makes the bill from AWS, Azure and Google hard to predict.

Oversight and money heading in different directions

The money flows in the industry point in several directions at once. OpenAI expects a huge deficit of up to $278 billion through 2030 due to investments in computing power, while Anthropic reportedly has annual revenue of more than $100 billion and is considering a new model ahead of an IPO. The French company Mistral has meanwhile raised 3 billion euros from investors. So there are both companies burning money on a large scale and others earning or raising capital at a similar pace.

On the oversight side, Anthropic and the consulting firm Accenture are investing two billion dollars in independent oversight of advanced AI models, but critics ask whether the arrangement can really be called independent when Anthropic itself is funding it. OpenAI is similarly trying to give companies better tools to measure how employees use ChatGPT, but both OpenAI itself and independent analysts warn that usage figures are not the same as proof of real value.

AI is woven into existing systems

Another trend is that AI companies are competing to get closest to companies' core systems. Salesforce has launched its own AI model, Koa, for sales and customer service, while Anthropic's Claude can now connect directly to Salesforce data with access to 37 sales tools. Google is doing something similar by letting its AI assistant Gemini pull data from both Salesforce and HubSpot. And in the pharmaceutical industry, Novo Nordisk is working with Anthropic to use AI to find new drugs faster.

There is also a geopolitical angle: the Chinese company DeepSeek is betting on Huawei chips to get around American export rules, which shows that the AI race is also about access to hardware, not just software.

For companies, the advice is to get a handle on who uses which AI tools for what, before the bill or the liability grows too large. Pay particular attention to whether the new oversight schemes and value measurements actually provide an overview, or whether they are mostly marketing.