In depth
Why clear rules make AI faster
AI saves time when a company's rules grow along with it. That sounds like a brake, and it's actually the opposite.
Pace · Ground rules · Leadership · A case
Pace
Good brakes make you fast
In motor racing they say it's the brakes that make a car fast. If you can rely on your brakes, you brake later and get through the corner faster. AI works in a similar way: a company can get as fast as it's able to check and, if in doubt, turn back.
This pace is different for every company. It grows when your rules grow along with it. What makes that happen is a fixed date each quarter, when your leadership team decides on new features and adjusts its rules to what the vendors have changed in the meantime.
Imagine two companies. In the first, the AI policy has been on the managing director's desk since March. They want to see first where things are heading. That's quite sensible, really.
In sales, though, half the people now write their quotes with a private tool, because customers expect the answer the next morning. The head of sales knows this and says nothing, because she needs the numbers.
In the second company, a system has been pre-sorting job applications since March. The process includes an approval by HR. The HR specialist in charge has a little over a minute per application, and in six months he hasn't changed the ranking once. He works conscientiously. The time is simply only enough to sign off.
Both companies act sensibly, and both pay for it. In the first, a lot happens in the company's name that nobody decided. In the second, everyone relies on a check that nobody is able to do any more.
The institute METR measures how fast the tools change. It records how long a person needs for software and research tasks that the best models solve half the time. According to its analysis of January 29, 2026, this duration doubles about every 131 days. A target picture from January often no longer fits in December.
Ground rules
Checking time is the most visible part
The Savings Calculator shows a single point in a process where AI meets your company's rules. The same ground rules decide at many points where no AI is involved: who may decide what, when a trial becomes a rule that applies, and who can change it again.
These rules include what's written, such as rules of procedure, approval rules, job descriptions and goals. They also include what has grown up beside them, from the shortcut to the list next to the system. The gaps are often between these pieces, because each covers its own part and none covers how they work together.
AI makes such gaps visible sooner, because a program speeds up every process, including the one nobody is responsible for. So the work often starts with a single approval and then reaches as far as the question of how your organization decides and steers as a whole.
When things get stuck in your company, it's rarely down to the people. Usually the organization itself produces the behavior everyone is annoyed about. That's actually good news, because an organization is something you can change.