Personio wanted to stop its own AI costs from tripling and came across a business area. His customers should now benefit from the experience.
Berlin. In the first quarter of this year, the start-up Personio was profitable for the first time. In the best case scenario, company boss Hanno Renner wants it to stay that way.
However, the rapidly increasing costs for the use of artificial intelligence (AI) have recently stood in the way of this goal. "Our AI costs have exploded this year. We are currently at six million euros instead of the planned two million euros," says Renner in an interview with Handelsblatt.
This is a surcharge that the personnel service provider for small and medium-sized companies felt strongly. In 2024, total sales were 196 million euros. There are no newer numbers.
Renner wanted to put a stop to the sharp increase in AI spending as quickly as possible and had various applications developed internally. This is initially to get an overview and then to get the costs under control.
The individual projects quickly created a so-called “AI transformation hub”, a central point that records and then optimizes the applications. According to Renner, the measures quickly led to success.
After a record in AI spending in July, spending has fallen significantly again, mainly thanks to an AI coach who advises employees on how to use language models, how to set a more favorable standard and how to invest in further training.
But it shouldn't stay that way. Because Renner has done a possible business deal in its own tool: “We will now make the AI transformation hub available to all of our 16,000 customers.”
According to Personio, Personio is currently building a new function that shows who in a company uses which agents, i.e. AI software systems that independently plan and execute tasks, and how efficient they are.
According to Renner, he had previously created an agent overview in-house, which shows which agents there are with which skills and who is responsible for these agents. “We have now decommissioned many agents and only kept the ones that really work,” he says. These measures have also reduced costs at Personio.
According to this, this best-seller is currently being tested in everyday life by selected customers. The focus is on an overview that shows which employees use which AI models, by whom, how often and intensively, and how their use affects productivity. At Personio itself, for example, more than half of Claude spending went to the top ten percent of users.
Furthermore, the AI expenditure can be compared to the personnel costs. Companies can also create surveys to find out how much their employees already know about AI. “You can also measure these results against benchmarks to understand where different areas of the organization currently stand,” says Renner.
For individual positions in a company, Personio also wants to provide information about how AI affects certain positions. “This is aimed at companies that maybe don’t use AI that much yet – to give them a feel for where they should look,” says Renner.
The entire offer could be of interest. According to a study by the digital industry association Bitkom, spending on generative AI, i.e. AI that independently generates new content, is expected to double to 11.5 billion euros in the current year. Generative AI, which includes ChatGPT, Claude von Anthropic, Microsoft Copilot and Google Gemini, now accounts for 40 percent of all AI spending.
In a Handelsblatt check, all start-ups valued at billions, also known as unicorns in investment parlance, reported significantly increased AI spending. Reactions to this were varied, ranging from restrictions on usage to more intelligent routing of requests.
Some companies have now discovered better control of AI costs as a business model for themselves. There are comparison portals for programming models such as DeepSWE. However, there are hardly any offers like Personio's that are integrated directly into your own software.
