Uhlík vs Křemík · Czech podcast about artificial intelligence

Slevomat CEO Tomáš Braverman: we adopt AI step by step, through ambassadors and hackathons

Episode 7 · 22 April 2026 · 46:57 · Guest: Tomáš Braverman

Listen on: YouTube · Spotify · Apple Podcasts

Tomáš Braverman, CEO of Slevomat and previously head of Heuréka for ten years, talked on the Uhlík vs Křemík podcast about how his company is adopting artificial intelligence. The conversation was recorded at the Fenomén conference in Zlín, where he was one of the main speakers. His message to digital businesses is that anyone who is not fast enough with AI will be washed away by the wave.

Respect for the impact on society

Braverman says he is not entirely optimistic about AI. From Harari's book Nexus he took away that AI in the hands of authoritarians is an extremely dangerous tool. The second risk he sees is that AI replaces far more people than earlier revolutions did, essentially all white-collar jobs. In history, he says, unemployment and economic crisis led to radicalisation and war. He sees regulation as the only defence, even though he knows it usually arrives with a long delay. As a fan of capitalism, he rejects wealth redistribution.

As the head of a company, however, he takes AI as a given. He compares it to a big wave, a small tsunami: "And if you are innovative and fast, you ride it, but if not, it washes you away and you die."

Gradual change instead of a big bang

According to Braverman, companies have two paths. The first is a big bang: team processes are redesigned from scratch as AI first, so a team of 50 people can become five people plus AI. The second is a gradual transformation of both processes and people. Slevomat chose the second. The first path carries risks and temporarily double costs, because you have to keep the old processes running until the new ones work. The second reason is cultural: at board level they decided they want to keep employing people. Slevomat has 360 people today, and the radical path could, according to Braverman, leave it with perhaps 50. The company will still optimise its headcount.

Training has three levels. The basic level is mandatory for everyone, and the company pays for the tools. Slevomat is Google first, so mainly Gemini, with Claude for some parts. The second level is knowledge and tools for each team. The third, which Braverman considers the most interesting, is a group of AI ambassadors, currently 17 people and soon 19, at least one from every team. They get the best training, and their main job is delivery: automating manual work and speeding up specific use cases.

A full day instead of two hours on Friday

The ambassadors meet twice a month for a full-day hackathon, attended by outside companies they turn to for advice. It used to be different: a Friday afternoon or two hours were set aside for AI, and the ambassadors were supposed to devote 20% of their time to it. It did not work: in such a short time they never got into the work, their workload did not shrink, and they ended up working on AI in the evenings. The company is still fine-tuning how the whole process is run. The use cases should be chosen before the hackathon, and team leaders, not just the ambassadors, should take part in choosing them. An MVP is supposed to come out of the hackathon, and the team's leadership then tests and implements it. The company is now implementing 10 to 20 specific use cases.

Asked how much time people should set aside for learning, he points to working with the calendar: fixed slots for focus time and learning that need to be kept, much like a morning run. Everyone at Slevomat has time for AI training in their calendar, and managers are supposed to make sure their people actually get it.

A website with an AI guide and a brand for language models

Some customers, Braverman says, want to stay in ChatGPT or Gemini, so Slevomat is implementing MCP and has built an app for ChatGPT that is awaiting approval from OpenAI. He sees the winning model, however, in combining the website with an AI guide that advises the customer and immediately adjusts the website, for example by filtering offers. The website also has things a chatbot cannot deliver yet, such as gamification.

Search is changing too. In his view, SEO is shifting towards optimisation for language models: technical SEO, domain authority and good content still matter, but brand strength and mentions on social media are gaining weight. "AI strongly favours, benefits well-known brands. I don't know how they do it, but that's how it is." He does not think websites will disappear, just as e-commerce did not replace brick-and-mortar shops and omnichannel emerged instead.

You can hear the full conversation on the Uhlík vs Křemík podcast.

The episode is in Czech. Czech page with chapters, quotes and the full transcript