AI seen as a necessary tool for press agencies

With the rise of Generative AI like ChatGPT, how do you see its impact on media outlets such as Vietnam Economic Times / VnEconomy?

The arrival of Generative AI has truly been a turning point. At Vietnam Economic Times / VnEconomy, we have constantly sought solutions to meet readers’ demand for rapid, smart access to information. While we tried various technologies, including traditional keyword-based searches, none fully met our needs. They were resource-heavy and didn’t yield the expected results.

With Generative AI, everything changed. We recognized that AI not only optimizes structured data processing but, more importantly, addresses the unstructured data problem; a valuable resource long overlooked in journalism. Previously, accessing this data required complex, manual categorization. Now, Generative AI processes vast amounts of information quickly, accurately, and at a lower cost.

Over the past 30 years, Vietnam Economic Times / VnEconomy has built a valuable archive of resources in the form of articles. Prior to Generative AI, we didn’t fully appreciate its worth. However, we had already started digitizing content in collaboration with the Central Press Distribution Company.

This process led to the idea of an AI tool. While we had vague thoughts of something like today’s AI chatbots, it wasn’t until ChatGPT emerged that everything clicked. Almost immediately, our Askonomy chatbot was born. The development of Askonomy happened “as fast as lightning” - it really did.

Could you tell us more about that moment?

The story began one evening at around 9pm. A staff member from the Central Press Distribution Company (under Vietnam Post) messaged me with the idea of creating a chatbot for Vietnam Economic Times / VnEconomy. I immediately called a quick-response team, and within an hour, we decided to move forward with the project.

We saw clear market demand. Traditional search engines like Google weren’t able to deliver direct, fast answers as expected. Askonomy, however, could provide instant answers in an interactive Q&A format, meeting user needs more effectively.

Our partner, Actable AI, mentioned that the project only required a small investment, so just a month later, Askonomy was launched at the 2023 Spring Press Festival using ChatGPT’s API.

I later invited Dr. Trung Huynh, Founder and CTO of Actable AI, who earned a PhD in AI and was based in the UK, to Vietnam for further discussions. He was confident, and I told him, “If you can do it, try to build a unique model.” He replied, “I think I can.” From that moment, we started developing Askonomy.

Before that, we tested ChatGPT. When we asked, “What do you know about Vietnam Economic Times / VnEconomy?” it responded, “I don’t know anything.” This was disappointing, as we are one of Vietnam’s largest economic media companies. After rephrasing the question, ChatGPT responded correctly, but this revealed the core issue: anyone can teach ChatGPT, but its answers are only as reliable as the data provided.

This highlighted the need for our own AI model.

Additionally, using ChatGPT’s token-based system would lead to high costs as user numbers grew, which made it unsustainable.

By 2024, we had launched a new version of Askonomy based on Mistral’s open-source models, and today, on June 2, 2025, we are introducing the full Asko Platform, with smarter second-generation models and more tools.

The naming process for Askonomy was quick and fun. We initially thought of a Q&A tool, but the name “Question and Answer” felt too common. I wanted a name starting with “A,” as it signifies top-tier quality. Combining “Ask” and “Economy”, we landed on Askonomy.

What was the biggest motivation behind your and Vietnam Economic Times / VnEconomy’s decision to invest in Askonomy?

It’s quite simple. The main motivation came from the challenges traditional media faces in the digital era, especially content distribution. If we continued publishing in the old way, journalism would have no future. That’s why, from the moment we transitioned from Vietnam Economic Times newspaper to Vietnam Economic Times magazine, we built a digital distribution system in the PDF format.

When ChatGPT launched, we saw it as the perfect opportunity to shift towards smarter, more proactive content delivery. Askonomy isn’t just a distribution tool, it empowers readers to engage with news actively. Instead of passively consuming content, users can now ask questions and receive instant answers.

Another breakthrough lies in Vietnam Economic Times / VnEconomy’s content management system (CMS), developed by our strategic tech partner, Hemera. This CMS is the key to managing content more efficiently while enhancing the user interface and experience through advanced technology.

Hemera also partnered with us to build the Asko Platform, which includes Askonomy and a suite of AI-powered tools designed to optimize media-related tasks. This system will allow us to bring cutting-edge AI technologies closer to both our readers and future enterprise users.

How is Askonomy currently being used in Vietnam Economic Times / VnEconomy’s content production process? How will the platform be customized to serve newsrooms across the country?

At Vietnam Economic Times / VnEconomy, Askonomy has become a smart chat tool powered by AI that enhances content distribution and encourages active reader engagement. Built on the Askonomy platform, we have developed a suite of applications within the broader Asko Platform ecosystem to support our reporters and editors in daily tasks, including article writing, content summarization, translation, voice-to-text, and text-to-speech conversion.

These tools will be officially launched on June 2, which marks the 32nd anniversary of the first issue of Vietnam Economic Times, published in partnership with Switzerland’s Ringier Group, and also kicks off the month celebrating the 100th anniversary of Vietnam Revolutionary Press Day.

A key feature of the platform is the use of specialized language models, also known as domain-specific language models (DLMs), tailored to each newsroom’s focus. For example, a legal publication would be provided with models trained on legal terminology and context. Economic media outlets, on the other hand, would receive models deeply trained in financial and economic language. With a large, well-processed dataset, these DLMs enable each newsroom to build an AI-powered ecosystem uniquely aligned with their editorial workflows and audience needs.

Given the high costs and lack of clear monetization in AI, how can a newsroom afford to invest in building its own chatbot or AI system, even if it’s considered a “small investment”?

Many people say that building proprietary AI models is extremely challenging, especially in terms of cost. And it’s easy to see why, when they point to giants like ChatGPT and other big tech platforms. These models require massive resources. ChatGPT, for example, has hundreds of billions of parameters, supports hundreds of languages, and pulls from datasets across all domains. On top of that, running these models consumes vast amounts of electricity, water, and computing power (like GPUs). I often refer to such models as “dinosaurs”.

That’s why, from the very beginning, I envisioned something different, a model I call “the ant”: compact, efficient, and focused on a specific task. Instead of building an all-knowing AI, we designed one centered on economic language and trained exclusively on Vietnam Economic Times / VnEconomy’s data with literally a millionth of the investment required for those mega-models. To further cut costs, we limit it to just two languages: Vietnamese and English, which are enough to meet our readers’ needs.

But this “ant” concept doesn’t stop there. What I really envision is a colony of ants - each business or newsroom owning a tailor-made AI model crafted for its specific needs. That’s the core idea behind the Marcom AI Platform - an ecosystem of small but powerful AI models purpose-built for newsrooms, corporate communication departments, government agencies, and industry associations.

Instead of pouring billions into infrastructure like the tech giants, we rent GPUs and scale based on actual demand. This makes the model more cost-effective and flexible, while also creating a self-sustaining cycle, from tech development to solution delivery.

To bring this vision to life, Vietnam Economic Times / VnEconomy partnered with Actable AI and Hemera to establish the Marcom AI Platform Consortium - a bold strategic move to build a specialized AI ecosystem for marketing and communication.

You just mentioned the Marcom AI Platform Consortium, and it has been noted that Vietnam Economic Times / VnEconomy will also lead the Data & AI Alliance (D.A Alliance) project. Could you share the vision and goals behind creating this new playing field for businesses and press agencies in Vietnam?

The idea behind the Marcom AI Platform and the D.A Alliance is rooted in the commitment to implementing Resolution No. 57-NQ/TW from the Politburo [on breakthroughs in science, technology, innovation, and national digital transformation in Vietnam] as well as the National Digital Transformation Strategy the Party, State, government, and society have been advancing in recent years.

Over the past four years of developing Askonomy, Vietnam Economic Times / VnEconomy has independently researched and invested in building AI applications tailored to its own needs and has secured positive results.

From that success, we recognized that Askonomy could serve as a practical model for other newsrooms to follow. With its proven efficiency and accessible cost, we believe other media outlets can absolutely pursue this path. The Marcom AI Platform Consortium is here to support that by offering small language models (SLMs) and sharing our implementation experience, enabling newsrooms to integrate AI quickly and effectively into their workflows.

AI thrives on large datasets, and the infrastructure to support it can be prohibitively expensive. If media organizations and businesses collaborate, sharing both data and infrastructure while also leveraging pre-built AI models, we can build a highly efficient and cost-effective ecosystem.

The Marcom AI Platform Consortium is designed to be an open playing field for Vietnamese businesses and media outlets to tap into shared SLM models, reduce investment costs, and optimize performance. This not only boosts productivity but also strengthens editorial credibility and reader engagement, laying the foundation for the sustainable development of Vietnam’s press industry.

Under the patronage of Vietnam Economic Times / VnEconomy, a respected economic press agency with more than 30 years of experience and a mission of supporting enterprises, the D.A Alliance aims to act as a collaborative bridge. It will foster partnerships to build a robust data economy and advance AI adoption across sectors, embedding Generative AI into the heart of every business in Vietnam.

Returning to the “AI ant colony” model you mentioned earlier, could you elaborate on the specific value that each AI “ant”, such as Askonomy, can bring to newsrooms and businesses?

As we’ve been developing the AI ecosystem at Vietnam Economic Times / VnEconomy, our goal has been to build individual AI models that can be optimized at each organization. For instance, if someone asks about a specific project, Askonomy can instantly summarize all related information within five seconds, without the need for digging through files. That’s the core value we aim to deliver: serving organizations with large volumes of data while ensuring data confidentiality. This is the fundamental direction we are pursuing.

I envision that millions of organizations and enterprises will need such solutions. Rather than relying on massive, generalized AI models like ChatGPT, which are trained on diverse datasets from countless domains and may not deliver precision for specialized needs, we are building an “ant colony” of lightweight AI models. These are fine-tuned using each organization’s proprietary data and fully safeguard that valuable dataset.

For example, Askonomy is trained solely on Vietnam Economic Times / VnEconomy’s data, ensuring that responses are accurate and contextually relevant without being diluted by unrelated sources. While ChatGPT can digest everything from reports to profiles, it doesn’t always produce results tailored to specific organizational requirements.

Building an AI “ant” sounds feasible for many organizations. But could you clarify the cost implications so they can better envision what’s involved?

When implementing such an ecosystem, two key elements matter most: data and infrastructure. Today, infrastructure is no longer a major barrier. We are constantly working to solve that puzzle. As technology advances, hardware costs are falling, AI models are becoming more efficient and require fewer GPUs to operate. Modern chips are increasingly optimized for AI workloads, significantly reducing deployment costs.

Looking ahead, I foresee AI workloads being processed directly on personal devices, such as smartphones, without needing large servers. This clearly unlocks tremendous potential, allowing us to harness the computing power of millions of individual users to collectively strengthen AI systems.

That’s why the “ant colony” model we’re building holds such vast promise. A model like Askonomy doesn’t require the immense number of parameters found in large language models (LLMs) - it’s lean, efficient, and focused only on what truly matters. This translates to lower costs and higher efficiency, especially for organizations seeking tailored AI without the bloat of unnecessary data or infrastructure.