Vietnam officially connected to the internet in November 1997, at a time when online applications were mostly limited to email and a handful of websites operated by major technology companies. In journalism, only a small number of news organizations began experimenting with online publishing between 1999 and 2000. Founded in 1998, Thoi bao Kinh te Viet Nam’s online version VnEconomy was one of only a few digital publications to operate as a genuine online newspaper, with content updated on a daily basis.
Over the 26 years since, digital publishing technology has evolved in ways that would have been difficult to imagine back then. The advantages from being an early pioneer enabled VnEconomy to accumulate valuable experience and develop the determination to adopt the most advanced technologies available in journalism.
Initial experiments
The technological foundations of online journalism at the time were vastly different from today’s. This was not merely a reflection of Vietnam’s technological limitations, as digital publishing technology was still in its infancy from a global standpoint.
Comprehensive content management systems (CMS) and database-driven publishing platforms did not yet exist. There was no Google Analytics to measure readership, nor were there cache servers or content delivery networks (CDNs). Theories on online page design and digital graphics were largely adaptations of print newspaper layouts. Serious research into eye-tracking behavior and reader interaction patterns on digital screens would not emerge until several years later.
In the beginning, VnEconomy relied on the technologies available at the time to publish and update news. The work of online reporters and editors was highly manual. Most daily tasks involved receiving content from the print edition, converting fonts, using static HTML editing software such as FrontPage or Dreamweaver, manually inserting hyperlinks, updating homepage links, and uploading HTML files to a server.
Reusing content from other publications, primarily print newspapers, was even more labor-intensive. Articles often had to be retyped manually or, in a more “advanced” process, scanned from printed pages, converted into text using optical character recognition software, proofread, and then incorporated into the HTML publishing workflow. Original reporting followed similarly manual processes from writing through editing and publication.
HTML page design itself was a complex undertaking. One challenge stemmed from the fragmented Vietnamese font ecosystem. Northern publications typically used the TCVN3 (ABC) standard, while VNI was more common in southern Vietnam. Unicode did not become widely adopted across Vietnamese websites and online newspapers until after 2005.
Browser compatibility presented another obstacle. Internet Explorer and Netscape often rendered websites differently, particularly in their support for JavaScript, which enabled animated headlines, moving text, and image effects.
Internet connectivity in Vietnam during the early years depended largely on dial-up connections with extremely slow speeds and a limited user base. As a result, page-loading speed became the primary concern for online newsrooms, often requiring image quality to be reduced to the bare minimum. It was not until the widespread adoption of ADSL broadband technology in 2004-2005 that online newspapers gained the conditions necessary for rapid growth. During the same period, CMS technology became increasingly common, while Google Analytics emerged as a valuable tool for monitoring audience behavior and optimizing site design.
From hyperlinking to AI
As one of the earliest news organizations to develop an online edition, Tap chi Kinh te Viet Nam, formerly Thoi bao Kinh te Viet Nam, has undergone multiple waves of technological transformation. The emergence and rapid adoption of AI since late 2022 triggered a giant leap forward in digital publishing technology at VnEconomy, arguably representing an even more profound shift than the earlier transition from manually-linked static HTML pages to dynamic, database-driven publishing systems.
Beginning in early 2023, the magazine invested in and partnered with technology company Actable AI to develop its own proprietary small language model. A strategic question quickly emerged: why build a proprietary model when large language models such as ChatGPT and Gemini were already delivering impressive results?
The decision to pursue the more challenging path of developing an in-house model reflected a distinct newsroom philosophy. First, the most important values in journalism are authority and accuracy, requirements that can only be fully guaranteed when a newsroom maintains control over the data used to train its AI systems. Second, a proprietary model can continuously incorporate new knowledge drawn from the publication’s extensive content archives, while becoming increasingly cost-efficient over time. Third, ownership of the model enables deep integration of AI capabilities across every stage of newsroom operations, from identifying story ideas and recommending topics to supporting research and data analysis.
From the outset, the magazine’s AI strategy was never about following trends. The goal was comprehensive integration into the content production and distribution process rather than isolated experimentation or individual use cases.
Today, AI is embedded throughout VnEconomy’s CMS, significantly transforming newsroom workflows. Reporters can confidently instruct AI to assist with article production because the data sources used by the system are fully verified, eliminating the risk of fabricated information. Journalists can instantly access relevant data and information within the CMS while working on stories. Conferences, seminars, and interviews can be automatically transcribed, and AI can even generate preliminary article drafts to support reporters.
Comprehensive AI integration has also helped reduce labor-intensive tasks while significantly improving productivity. VnEconomy’s English-language edition operates with a relatively small team yet delivers a greater volume of content with faster publication times and reliable translation quality tailored to economic and business reporting.
The daily podcast production process likewise requires only minimal staffing, as AI automates multiple stages, from selecting articles for summarization to generating audio content using proprietary voice technology.
As AI becomes more powerful and widely accessible, its use in journalism has become increasingly routine. Asking AI to summarize documents, create illustrations, or generate infographics is now as commonplace as using a camera, a pen, or a notebook once was.
Yet AI’s true value can only be measured and validated when it is tightly integrated into a newsroom’s standardized workflows. The distinction lies in moving beyond individual preference or convenience and embedding AI into formal editorial processes. Only then can news organizations effectively monitor, manage, and quantify its impact and productivity gains.
(*) Mr. Nguyen .the Deputy General Managing Editor of VnEconomy.
Google translate