👨‍🚀 From AI adoption to AI transformation in companies & teams

Hello AInauts,

Welcome to the latest edition of your favorite newsletter! We don't necessarily write only about what's happening, but rather what interests us.

And because we're interested in so many different things, you can expect a colorful mix today—live and unfiltered, straight from our sometimes chaotic minds.

From AI strategies for companies to business ideas for AI children's toys, everything is included.

Here's what we have in store for you:

• 🇪🇺 Mistral 3: New AI models from Europe

• 🔥 From AI adoption to AI transformation in companies and teams

• 🧸 Business idea: AI in toys?

Let's go!

🇪🇺 Mistral 3: New AI models from Europe

It's starting to annoy us too 😁... Almost every issue, we write about some new model that has just been released and is great.

We can also understand that this can sometimes be confusing. That's why we'll spare you the details about the new (powerful) update from Chinese provider DeepSeek.

But you won't get away completely unscathed, because we have to start briefly with this topic... Because we have updates from Europe 🇪🇺, more specifically from France—and this is something that will also make data protection fans happy.

The Mistral team presented the new Mistral 3 models this week.

The models, led by Mistral Large 3, are available in various sizes from small to large. Basically, they can do everything you would expect—just like all the other big ones.

The reason why nerds like us are excited:

• All new Mistral 3 models have been released under the Apache 2.0 license—in other words, they are available as open source!

• Those who did not want to rely on big tech companies from the US previously had to use Chinese open source models for top performance. Not everyone's cup of tea either...

• However, Mistral 3's performance is now just as good or even slightly better than that of its open-source competitors DeepSeek, Kimi, and others.

However, we find the small model variants most exciting—Mistral affectionately calls them Ministral.

They are also really, really smart. They are great with text and can also process images.

The kicker: They run completely offline, locally, without the internet—or even just in the browser. (Provided you have reasonably good hardware)

Why is this relevant to you?

In general, we believe it is extremely important that good AI models also come from Europe. With Mistral for language models and Flux for image models, we are already in a good position (and perhaps the Swiss will follow suit with Apertus).

Therefore, we should also support these companies.

The open source availability and compressed power of Ministral is super exciting if you don't want to do without the intelligence of AI but don't have internet access or want to operate completely locally with data protection.

Here's how you can try it out

Mistral 3 is generally available on all common platforms. You can also try Mistral's in-house ChatGPT Le Chat for free here.

If you want to try the local version, you can do so using Ollama, for example. We wrote about the setup here back in early 2025.

If there is sufficient interest, we will once again create an up-to-date guide on using local AI (just leave a comment).

Last but not least: if you want to test the example from the video above locally in your browser, you can do so here.

🏆 From AI adoption to AI transformation in companies and teams

Let's now move on from language models to AI strategies for businesses.

We have been exploring the question of how companies can deeply embed the new world of AI in people, processes, and products for quite some time.

The goal of this phase is clear: to raise the floor and improve the results for conventionally completed work by marginal steps (e.g., 10% efficiency).

AI adoption is essential. BUT: If companies really want to remain competitive and fully exploit the potential of AI, they need to go one step further...

After adoption comes transformation: the change that reinvents everything.

The next step, the AI transformation, is the actual organizational change.

It is no longer about doing the work a little better—it is about completely reinventing the way work is done (Re-Invent).

Transformation is the moment when you raise the ceiling and no longer measure improvements in 10% increments, but in multipliers (e.g., 10x).

The difference in brief:

AI transformation: The path to 10x impact

How can companies approach such an AI transformation?

AI-first companies such as Zapier approach the whole thing on four levels:

1. Leadership

Senior management must define the vision, urgency, and direction.

Role: Create dedicated roles with responsibility. Chief AI Officer. Encourage executives to develop bold AI bets within their teams. Allocate and implement the necessary budget.

Key actions: Make it visible. Include progress on AI transformation in monthly reports and company goals. Identify core workflows with leverage and improve them with AI. Be a role model in AI usage.

2. Talent & Culture

This area promotes AI fluency and a willingness to experiment. This is crucial for acceptance throughout the company.

Focus on AI fluency: Create programs for employees to build AI skills. BUT: Make sure to actually measure them!

Zapier, for example, requires 100% of its employees to be "AI-fluent" and classifies employees according to their level of proficiency:

• Unacceptable

• Capable

• Adoptive

• Transformative

The highest level (Transformative) is achieved when AI fundamentally changes the way work is done and this new way of working influences others in the team, in projects, or in the organization.

This is how Zapier measures it, for example, depending on the area:

3. Tools

Teams must be equipped with the right technology stack, data, and orchestration.

• Data infrastructure: It is important to strengthen the data foundations and standardize the data infrastructure and preferred tools.

• Procurement and access: A process for procuring AI tools, including a "fast lane" so that tools can be tested quickly and the right ones can be found.

• Support role: AI Automation Engineers convert prototypes into production-ready systems and ensure fast, high-quality delivery of automated core workflows.

For us, a key element is a number of new positions, such as AI automation engineer.

We have already placed four such engineers in companies. Above is an excerpt from the job description.

If you are looking for something similar, here is a link to our template as a DOC file that you can copy.

4. Governance (supervision/control)

The last of the pillars is one of the most important. This factor provides guardrails and ensures secure and scalable acceptance.

What do we mean by that?

The whole thing was a sneak peek into a topic we are currently working on with several companies. And for which we want to raise awareness.

In 2025, the main focus was on companies successfully getting their existing teams up to speed on AI.

Advancing AI adoption.

Many are still in the midst of it, and it will remain relevant and important next year as well.

However, we can see that the top companies will take AI transformation one step further in 2026 and reinvent many things. Exciting times, dear AInauts!

🧸 Business idea: AI in toys?

To conclude this rather lengthy newsletter, we would like to briefly address a rather controversial topic: AI in children's toys

This topic has been on our minds for quite some time. We believe that there are some really interesting opportunities for business ideas here.

At the same time, there is also a lot of movement in the market right now. Both positive and negative.

Let's start with some negative headlines. The company FoloToy and their AI Teddy Kumma got into quite a bit of trouble because the cute teddy bear gave testers tips on where to find "dangerous things" in the house and even worse... Click here for the full report.

Sure, similar to ChatGPT, these types of toys are extremely difficult to control and highly questionable.

However, we also stumbled upon a positive example: Stickerbox

It is a small box to which you submit an idea. Based on this, an image is generated and printed as a sticker. Children can color it in, stick it on, and collect it.

Somehow cool. Somehow, we can't get it out of our heads.

Business ideas in the AI toy sector

If you're bored and find business in this area exciting, here are a few ideas.

1. AI Story Box: Children narrate the beginning of a story, the AI generates continuations, and at the end, it prints out a personalized mini-book/comic (thermal paper booklet). Recurring revenue through paper refill packs.

2. AI music toy: Child hums or sings a melody → AI turns it into a complete song with instruments.

3. AI building instructions generator: Compatible with common building blocks: Child says, "I want to build a dragon" → AI generates step-by-step instructions based on the available blocks (via camera scan).

4. AI pen pal printer: Children dictate a letter to a fictional AI friend (pirate, astronaut, dinosaur) → after a few days, they receive dynamically varying "handwritten" reply letters printed out. Episodic adventures as a subscription.

That's the first piece of input if you're planning a side business for 2026 😁 …

You made it to the end—thanks for reading! We’ll be back soon with even more updates.

Reto & Fabian from the AInauts

P.S.: Follow us on social media - it motivates us to keep going 😁!
X, LinkedIn, Facebook, Insta, YouTube, TikTok

Your feedback means the world to us. We read every comment and message—just hit reply and tell us what you think!