How AI is reshaping global newsrooms: What Bangladesh journalists can do

Global mainstream media organizations like Associated Press, BBC, Wall Street Journal are increasingly integrating artificial intelligence into their daily workflows to increase productivity of journalists, optimize news production, automate repetitive tasks, and execute complex investigations.
However in Bangladesh, there is a significant gap in terms of institutional formalization and less focus on how we can use available free tools in newsrooms to increase our productivity. 

AI Use in Newsrooms at a glance
Rather than replacing human reporting and sub editing, some mainstream media deploy AI models to manage high-volume data analysis, generate automated copy, and refine content delivery.

Major publications like Bloomberg use proprietary large language models (LLMs) to analyze financial documents and perform sentiment analysis. Semafor uses an AI-powered feed called “Signals” to help journalists search through global news sources in multiple languages.

The Financial Times and the Wall Street Journal use AI models to predict trending topics to help journalists find gaps in their current coverage. 
The Associated Press utilizes natural language generation platforms to automatically ingest financial datasets and draft corporate earnings reports and minor-league sports articles. This enables the agency to scale coverage without draining manual reporter resources.

BBC uses internal generative AI models to reformat stories from the Local Democracy Reporting Service (LDRS) into the standard BBC house style, to generate at a glance summaries, and language translation. 

AI tools, in global media, act as smart assistants that handle, with human intervention, time-consuming tasks, allowing reporters to focus on deeper journalism, for example transcription, translation, drafting and primary editing of a news article, and multimedia creation in rush hours. 

The Digital Divide: AI Integration in Global and Bangladeshi Media
Leading international media outlets have enough budget and RnD teams to deploy custom-built and proprietary large language models in their houses. In Bangladesh, we are still in the exploratory phase of adoption.

In global media, AI adoption is guided by well-defined institutional frameworks, extensive tech budgets, and specialized software engineering teams embedded inside the newsroom. 

Conversely, in Bangladesh, AI adoption is primarily driven by individual usages under tight deadlines rather than a well structured institutional strategy. Bangladeshi journalists and writers extensively use general-purpose tools like Grammarly, Google Translate, and ChatGPT for basic proofreading, text rephrasing,creating news from press releases, and translation from Bangla to English and vice-versa.

Industry leaders at a media dialogue hosted at The Daily Star Centre said that while AI tools are highly affordable, Bangladeshi newsrooms suffer from severe ethical and operational gaps. Most local organizations operate without formal, written AI editorial policies or self-regulating mechanisms.

Editors at New Age newspaper have also noted a growing sense of limitation. They predict that news managers train staff to use generative tools like ChatGPT, however, the output frequently lacks the localized emotion, cultural nuance, and deep historical context required for impactful domestic journalism.

Cautions, Bias, and Human Intervention
The rapid adaptation of AI in journalism also brings some risks worth evaluating. 

Misinformation and Deepfakes: It is nowadays very easy to engineer highly credible fake contents, images, and videos. Without rigorous verification and fact checking, newsrooms risk unintentional risk of spreading state-sponsored propaganda, corporate manipulation, and AI-induced hallucinated facts,which can lead to a post-truth era.

Training Data Bias: Standard commercial AI models are trained on massive, and mostly western-centric datasets that carry inherent linguistic, cultural, and political biases. 
These models frequently misinterpret the historical and socio-political nuances of countries in the Global South, the Middle East, to name a few, leading to produce biased information.

These are only two of many critical problems we face in the newsroom while using AI. 
To counter these, human oversight is mandatory at every stage of the journalistic pipeline. We must prioritize accuracy over speed, quality over quantity, keeping in mind that standard tools like Google Translate, Grammarly, or ChatGPT always require strict human verification before publication.

What can we do now? 
AI, Large Language Models, and AI agents- all these are utility tools like the internet, a keyboard, a typewriter machine, or a smart assistant who can bring the best output under restricted human intervention. 

Here is an example how to chain existing AI solutions in our article writing pipeline. 
Without generating a generic story like ‘write an article on AI use in newsrooms’ and sending it to the subbing desk, we can do the following-
  
Firstly, we can use Google search or Google AI mode to collect data on how AI is used in the mainstream media newsrooms across the world and Bangladesh. This is the research and development phase to collect necessary information. 

Then we will select some five to ten original, authentic, and informative pieces.

Now, we feed NotebookLM with the sources, text, or PDFs. NotebookLM is a Google product used for creating a knowledge base from strictly defined sources so that AI makes few mistakes and does not add hallucinated data.     

Then we will filter out our necessary data (obviously cross check any critical information), and assign it the task to prepare a draft on the topic. 

Secondly, we will go to Gemini, and create an ‘article writer gem,’ a gem is your own custom built AI assistant who is specialised in a specific task like report writing or sub editing. 

Based on the information collected from NotebookLM, we will define the article writer gem how it will structure the final draft.