Automatic ticket routing in Zendesk

Product

LATEST NEWS

A quick recap from our latest Contekst event

A recap from our latest Contekst event.

The iconic Wintercircus in Ghent transformed into a SOLD OUT buzzing hub of ideas, innovation, and inspiration. Contekst, hosted by Tekst, brought together some of the brightest minds in Belgium's largest companies for a full day of learning, sharing, and connecting. From established enterprise leaders to emerging AI innovators, the event bridged the gap between vision and application—making one thing clear: AI is no longer a futuristic concept; it's a real-world driver of change.

Big Ideas on a Big Stage

The day kicked off with a powerful keynote by none other than Peter Hinssen, thought leader and futurist, who challenged the audience to rethink how organizations adapt in an age of exponential technology. His message? The companies that win with AI aren’t necessarily the ones with the most data—but the ones with the most agility.

This set the tone for a vibrant fireside chat featuring industry veterans Elke Vandepitte (Colruyt Group), Christophe Boucquet (Lecot), and Pieterjan Bouten (ex-Showpad, now at Encourage). Together, they shared their unique perspectives on navigating digital transformation, the human side of automation, and what it really takes to embed AI into the DNA of an enterprise.

A Community of Curious and Bold Thinkers

What made Contekst truly special wasn’t just the speakers or the tech—it was the community. Over 100 professionals from both industry and government came together to learn from each other and explore how AI can solve real, gritty problems. We were thrilled to welcome companies like Colruyt Group, Securex, Milcobel, D’Ieteren, BD, Lecot, Acerta, Daikin Europe, KBC Bank, Johnson & Johnson, Roularta, SD Worx, Mainfreight, Just Russel, and many more to the event. The networking moments were just as powerful as the talks—connections were made, use cases shared, and bold ideas born.

Missed It? Watch the Aftermovie

Couldn’t make it to Contekst? Don’t worry—we captured the energy, the moments, and the magic.
Watch the aftermovie and feel the AI-mazing vibes for yourself.

Thank You — and see you at Contekst 2025!

To everyone who joined us, contributed to the conversation, and made Contekst 2024 such a standout success: thank you. Your curiosity, energy, and openness to rethink what's possible with AI are what make this community so powerful. We’re already looking forward to Contekst 2025, happening on May 22 at the stunning Valke Vleug vineyard. Expect fresh air, fresh perspectives, excellent wines and even deeper dives into the future of enterprise AI. See you there!

‍

Tekst at Hannover Messe 2025!

Tekst Showcases Flemish AI Innovation at Hannover Messe 2025

Last week, the team at Tekst had the opportunity to showcase their cutting-edge AI solution at HANNOVER MESSE 2025, one of the world’s leading industrial technology trade fairs. Representing Flanders at the Startup Flanders pavilion, CEO and co-founder Wouter Janssen pitched Tekst's vision to global industry leaders—and even to political figures like Flemish Minister Matthias Diependaele.

“Well, I could use this in my cabinet,” the Minister remarked after Tekst’s demo—a strong sign that our solution resonates far beyond just private sector applications.

Revolutionizing the Inbox With AI

Tekst builds AI specifically designed to help companies—and public institutions—automate email and aftersales workflows. The goal? To eliminate repetitive, manual back-office tasks like email triage, forwarding, classification, and data re-entry.

Whether it's an overloaded customer service inbox or post-sales processes that drain resources, Tekst ensures teams can focus on what truly matters by:

Automating repetitive emails
Smartly routing messages to the right people
Reading and extracting data from PDFs and attachments
Providing clear insights into communication flows

As Wouter put it:

“We think every Flemish institution could benefit from a little Tekst magic.”

From Flanders to the World

HANNOVER MESSE is known for being a hotspot for innovation in Industry 4.0, AI, and digital transformation. With over 130,000 visitors and thousands of exhibitors, the event connects tech innovators with industrial leaders seeking to future-proof their operations.

For Tekst, participating wasn't just about visibility—it was a strategic step in scaling internationally. The company’s focus on low-disruption integration and high-impact automation makes it a compelling partner for enterprises and governments alike.

And the best part? You didn’t have to be a Member of Parliament to get the demo—attendees at the Startup.Flanders pavilion were treated to the same live pitch that impressed Minister Diependaele.

What’s Next for Tekst?

With growing interest from both industrial and public players, Tekst is expanding its AI capabilities and integration portfolio. From Ghent to Germany and beyond, they’re on a mission to make messy inboxes a thing of the past.

‍

OUR PRODUCT

No items found.

How it works

Tekst simplifies your business operations like never before. Our custom AI model interprets every incoming email, ticket, and chat, understanding the context and intent.

‍

Customer stories

Explore how Tekst powers next-gen operations with real use cases from real customers.

‍

Solutions

LATEST NEWS

New at Tekst: Build Your Own Inbox Automations with Custom Actions

Say Hello to Custom Actions: Built for Your Unique Processes

At Tekst, we understand that automation isn't one-size-fits-all. Many businesses have their own quirky, nuanced workflows — especially when managing shared inboxes. That’s why we’re excited to introduce a new feature we’ve been testing internally: Custom Actions.

This powerful update is part of our automation module, and it lets you build your own tailor-made actions to slot directly into any automation flow. Whether it’s managing internal requests, categorizing support emails, or routing inquiries to the right person — you can now create automation steps that match your exact use case.

How It Works: A Real-World Example

To show how easy and useful this is, our Head of Product, Servaas, walks us through a real automation scenario:

A shared inbox is connected to an Outlook trigger.
When a new email arrives, it's processed by Tekst’s custom AI.
If the email is about an admin-related question, it automatically forwards the message to the responsible person — in this case, Servaas.

A simple and very basic automation flow, so far… But what if you want even more control?

Servaas explains: “Lunch is holy for me.”

So what if you don’t want to be interrupted by admin emails between 12:00 and 13:00?

Here’s where Custom Actions shine:

Servaas adds a custom step to his automation flow.
He clicks “Create Custom Action,” which opens Tekst’s app builder.
No coding? No problem — there’s a prompt-based UI. He simply writes:
“Write an action that detects if I’m at lunch (between 12 and 1), and wait until lunch is over before continuing.”

Tekst’s AI takes over from there:

It generates the logic and code automatically.
You can customize the code or simply add this new action to your automation.
Now, admin emails only forward after lunch is over — so you can eat in peace.

Why This Matters

Custom Actions give you unmatched flexibility — without needing to code (though you can if you want).
Perfect for teams who want smarter inbox workflows.
Save time. Avoid disruptions. Automate the boring stuff.

Whether your process is industry-specific or just a bit... personal (like protecting your lunch break), you now have the tools to automate your way.

‍

Curious how it works in practice? Watch Servaas walk through it in detail:

‍

USE CASES

Clutter removal and email filtering

Proposed answer generation

Smart summaries

Automated complaint handling

Automatic order intake

Automate inventory management

INDUSTRIES

Customer Stories

LATEST NEWS

New at Tekst: Build Your Own Inbox Automations with Custom Actions

Say Hello to Custom Actions: Built for Your Unique Processes

How It Works: A Real-World Example

To show how easy and useful this is, our Head of Product, Servaas, walks us through a real automation scenario:

A shared inbox is connected to an Outlook trigger.
When a new email arrives, it's processed by Tekst’s custom AI.
If the email is about an admin-related question, it automatically forwards the message to the responsible person — in this case, Servaas.

A simple and very basic automation flow, so far… But what if you want even more control?

Servaas explains: “Lunch is holy for me.”

So what if you don’t want to be interrupted by admin emails between 12:00 and 13:00?

Here’s where Custom Actions shine:

Servaas adds a custom step to his automation flow.
He clicks “Create Custom Action,” which opens Tekst’s app builder.
No coding? No problem — there’s a prompt-based UI. He simply writes:
“Write an action that detects if I’m at lunch (between 12 and 1), and wait until lunch is over before continuing.”

Tekst’s AI takes over from there:

It generates the logic and code automatically.
You can customize the code or simply add this new action to your automation.
Now, admin emails only forward after lunch is over — so you can eat in peace.

Why This Matters

Custom Actions give you unmatched flexibility — without needing to code (though you can if you want).
Perfect for teams who want smarter inbox workflows.
Save time. Avoid disruptions. Automate the boring stuff.

Whether your process is industry-specific or just a bit... personal (like protecting your lunch break), you now have the tools to automate your way.

‍

Curious how it works in practice? Watch Servaas walk through it in detail:

‍

BY INDUSTRY

BY CUSTOMER

About

LATEST NEWS

Meta's Release of LLama 4

Meta has launched Llama 4, the latest in its series of large language models. This release marks a major step forward in the AI space, combining advanced capabilities with Meta's open-model philosophy. Llama 4 builds on its predecessors by offering substantial improvements in performance, scale, and multimodal capabilities.

Key Features and Model Variants

Llama 4 is available in several versions, with notable models being Llama 4 Scout (17B active parameters, 109B total) and Llama 4 Maverick (17B active, 400B total). A larger Llama 4 Behemoth (2 trillion parameters) is in development but not yet released. The introduction of Mixture-of-Experts (MoE) architecture allows for more efficient use of parameters, activating only a fraction of them for each query, which results in high performance while managing compute requirements effectively.

Performance and Benchmark Improvements

Llama 4 significantly outperforms previous models like Llama 3, and even competes with the latest models from OpenAI and Google. It handles complex reasoning and tasks more effectively than Llama 3 and delivers better results on key benchmarks. One of the standout features of Llama 4 is its extended context window, which can accommodate up to 10 million tokens for the Scout model—much larger than previous versions. This allows Llama 4 to manage longer conversations or analyze larger documents without losing context, a game changer for many industries requiring deep, long-form analysis. A sidenote: Although Llama 4 models are capable of handling longer context windows, the technical paper indicates that they were only trained with up to 265,000 tokens, not 10 million. Beyond 256,000 tokens, performance may vary.

Multimodal Capabilities

Unlike its predecessors, Llama 4 introduces multimodal abilities, enabling the model to process both text and image inputs. This multimodal functionality positions Llama 4 alongside models like GPT-4, allowing it to analyze and respond to visual information in addition to text, which broadens the range of applications, such as visual question answering or image-based reasoning tasks.

Currently, there is no audio modality, but it's expected that the open-source community will add this functionality in the coming weeks.

Integration with Meta’s Ecosystem

Llama 4 is integrated into Meta’s suite of apps, including WhatsApp, Facebook Messenger, and Instagram. This means users will directly benefit from improved AI assistants across Meta’s platforms. The model can assist with content generation, respond to user queries, and even analyze shared images. The widespread integration ensures that Llama 4 is actively used and refined within Meta’s vast ecosystem, enhancing its real-world utility. This is a trend observed among most model providers, who are expanding into consumer technology (e.g., SearchGPT, B2C licenses, etc.).

Safety and Open-Source Accessibility

Meta has continued its commitment to open-source AI with Llama 4. The model weights for both Scout and Maverick are available for developers to download and experiment with, encouraging community-driven innovation. However, the licensing comes with restrictions, such as prohibiting large-scale commercial use by companies with over 700 million monthly active users. Meta also places a strong emphasis on safety and alignment. Llama 4 comes with tools like Llama Guard to filter unsafe queries and Prompt Guard to ensure the model generates responsible content. These safety measures are designed to mitigate risks of misuse while allowing researchers and companies to harness the model's power. The main advantage is its accessibility, allowing users and developers to customize safeguards and align the model with their specific use cases.

Strategic Importance and Future Outlook

Llama 4 represents a strategic pivot for Meta in the AI race. By offering such powerful capabilities in an open-source format, Meta challenges closed models from competitors like OpenAI and Google. The model's multimodal abilities, longer context windows, and open access make it a strong contender in the rapidly advancing field of conversational AI. Looking ahead, Meta has plans to scale even further with the Behemoth model, and future versions of Llama (Llama 5, Llama 6) will likely push the boundaries of AI even more.

For developers and researchers, Llama 4 offers a flexible, powerful tool for building advanced AI applications across industries. It signals that Meta is positioning itself as a leader in AI development, pushing for openness and innovation at scale. As Llama 4 is integrated more deeply into Meta’s ecosystem, its influence will only grow and it will shape how AI is used across platforms and industries in the years to come.

‍

Why multilingual AI is the future: 6 reasons why

Multilingual AI is AI that is capable of speaking more than one language and its counterpart is Monolingual AI

Be aware tho, there is a lot of “Multilingual” AI that under the hood translates everything to English. At tekst.ai we don’t like this as this downgrades the performance of other languages. We believe in AI models that speak each language with the same performance and do a lot of R&D to also make SOTA AI available to everyone - regardless of each language.

Here are 6 reasons why multilingual AI is the future

1. Multilingual customers

The number one reason is to be able to also handle non-English speaking customers. In Europe alone, 200+ languages (some non-official) are used in our communication. By not serving these customers with AI we disadvantage them.

In the customer support world, this would mean that someone not speaking English will receive slower and less accurate support. We are striving to not let this happen.

2. Price

What do you think is more expensive: 1 model for +104 languages or 10 models for 10 languages?

Even though the latter only supports 10% of the languages of the first one it is almost 10 times as expensive.

We don’t think much more is needed ;)

3. Effort

Setting up a lot of monolingual models is also a tedious task and often repetitive. Wasn’t the whole goal of using models to avoid repetitive waste??

Setting up one multilingual model takes about the same time and effort as a monolingual model (and even effortless by using our tool *Kuch *Kuch) while having the advantages of multiple monolingual ones.

4. New markets

Europe has a multilingual characteristic. For global companies to break into Europe it is needed to deliver support in the language of the customer. US-based companies often struggle with the fact that their support needs more than one language and need to move away from a monolingual system. our advice is to start thinking multilingual as soon as possible, by also incorporating this into your tech and tools will enable future growth.

5. Rising trends

Studies and research in this area have been ongoing for a while now and it’s expected that new and highly functional multilingual AI models will be trending in the near future. We are riding these waves and achieving state-of-the-art results in bringing real multilingual AI to everyone - regardless of language.

6. Can’t stay behind

This one is maybe more a personal frustration but we - non-English people really can’t stay behind. We can’t let good AI be only useful for English and we really should do an effort to also have good AI in our language. less than 5% speak English and for the remaining 95%, good AI should also be useful. Otherwise, a lot of automation and intelligence will not be useful for them.

It really should be a no-brainer now to only use multilingual NLP - however, be aware of the fake AI vendors, and be safe out there!!

‍

ABOUT US

About us

Learn how Tekst is redefining enterprise operations with AI-driven innovation.

‍

Careers

We're hiring!

Join Tekst and help build the next generation of AI-powered solutions.

‍

Blog

Explore insights, trends, and ideas shaping the future of enterprise AI.