David Novák (Kogi): AI without hundreds of millions — fast results without a grand transformation

In this episode of the internal Kogi no-holds-barred series, we talk with David Novák about putting companies in motion in the age of artificial intelligence. David Novák explains why AI isn't just another technology project, but a shift that, within a few years, will rewrite most roles' job descriptions and split companies into AI-native ones and those that stick with traditional tools.

The biggest brake, though, often isn't the technology or the infrastructure budget, but organizational paralysis and people's fear of even using AI at all. And when adoption only happens as an unmanaged shadow AI economy, it does save individual time, but it doesn't produce any systemic change in performance, decision-making speed, or the quality of internal communication.

AI paralysis at large organizations

At larger organizations, David Novák describes a recurring pattern: an AI agenda gets created (a center of excellence, an AI department, an AI officer), a complex strategy gets built — and the rest of the company passively waits to see what comes out of it. This is often backed by a narrative: get the data in order first, then move on to AI use cases — which, in practice, leads to years of paralysis and gives people an alibi for not changing teams, habits, or ways of working.

Data and governance aren't pointless. Without them, a lot of hard use cases in risk, compliance, or automation can't be handled safely. The problem arises when the data-first approach becomes a universal excuse that stops even fast pilots in places where you could start immediately, with minimal investment.

The value of AI lies in the organization's soft layer

The biggest value of AI may not be in robotizing processes, but in the organization's soft layer: communication, interpretation, sharing strategy, meetings, notes, emails, Teams/SharePoint, and culture. That's exactly where an enormous amount of value gets lost through noise, misunderstandings, and differing interpretations — and AI can quickly help by analyzing the company's language and detecting drift from the strategy.

At the same time, if the strategy isn't real (missing trade-offs, just a wish list), AI will only speed up the production of “nice-looking text” that doesn't actually decide anything.

The practical first step: basic order

The practical first step toward managed adoption, then, isn't necessarily a giant data platform, but basic order in where the company talks and stores knowledge: where meeting notes are, strategic documents, meeting outputs, how Teams channels and SharePoint are structured. Without that, adoption slides into a “shadow AI economy” (people using AI secretly, unsystematically), which saves individuals time but doesn't move the company forward as a whole.

As a fast pilot, something like sales coaching from meeting transcripts makes sense: AI prepares the notes along with concrete feedback, complete with quotes and recommendations, possibly with metrics — those are often necessary for leadership to be able to justify the change, even though they carry the risk of “gaming the score.”

What else will you find answered in the podcast?

• Why does David Novák claim today's managers won't live to see a bigger technological revolution than this one — and what will that do to roles at companies?
• How and why does AI paralysis happen at large organizations (a center of excellence, “data first,” years of waiting)?
• What is the “shadow AI economy,” and why does it often fail to deliver systemic value to a company?
• Why is it worth starting in the organization's “soft layer”: language, meetings, documents, interpretations of strategy?
• What “minimal order” do you need in Teams/SharePoint so AI stops happening only as scattered, ad hoc use?
• What does a practical AI use case for sales coaching from meeting transcripts actually look like?
• Why can you sometimes not justify a change without harder metrics — and when, on the other hand, do metrics destroy the whole point of development?
• Why, even if the AI bubble eventually “bursts,” will the technology stay, and the pressure to change how we work won't let up?

If you want to approach AI in your organization as a managed change (not as an endless strategic project, nor as unmanaged “shadow” use), start with a specific goal and one pilot where you can quickly demonstrate value — and only then scale.