In the Press

The field of artificial intelligence (AI) has experienced remarkable advancements in recent years, particularly with the evolution of Large Language Models (LLMs), retrieval-augmented generation (RAG), and most recently, the rise of autonomous AI agents. Drawing on experience from our engagements, over the next three weeks I will be rolling out a series of op-eds from an executive lens. I will map out AI concepts into 3 capability layers that should not be ignored to achieve return on investment on their AI investment. This first article focuses on LLMs, which can be used as the baseline capability that enables RAG and autonomous AI. The 2nd op-ed will focus on explaining how RAG increases accuracy and auditability by anchoring outputs to approved sources, and finally the 3rd op-ed will focus on how autonomous AI can automate business processes through workflows with governance and controls.

What is Large Language Models (LLMs)?

LLMs are advanced AI systems that are trained on large amounts of text data and empowered with the ability to use this data to generate content that can interact with their human counterpart in natural language. Basically, we gave AI the ability to understand lots of information from textbooks and the ability to be able to talk to us humans about it. LLMs can be trained on vast datasets from different disciplines, and this is the reason why they are so popular. The ability to give it all the engineering texts that ever exist and letting it write your training materials, summarize information, and even answer your engineers' most difficult questions in real time can be a very powerful tool.

Some of the strengths that LLMs have are:

• Talk to a Human: LLMs can comprehend and respond to complex questions in a conversational manner and are intuitive to human user interactions.
• Anything you can do, I can do better: LLMs can summarize information, create reports, draft documents, and even generate code on demand. What would take a person hours of research, drafting, proof reading, brainstorming, etc., can now be done in minutes with higher accuracy and consistency.
• Multilingual: LLMs can support multiple languages, making them useful for international business. Imagine being able to have meetings with teammates from Japan, the United States, India, and Vietnam and have LLMs translate, take notes, and coordinate meetings for you based on your language of choice.
• Adaptability: LLMs can be customized for specific tasks, industries, or organizational needs. Being able to have LLMs just for engineering, cooking, etc. can even further enhance your AI investment by having a specialist who can talk, think and solve problems specifically for your industry needs.

Of course, with any technology that we implement, we always try to make sure we are transparent about weaknesses as well. From our experience, these are concerns we have experienced with LLMs:  

• Not 100% accurate: Sometimes LLMs generate information that is inaccurate or entirely fabricated (aka hallucination). Providing lots of information for LLMs is a good way to make sure it has the appropriate information to help, but information overload can also create potential inaccurate information being disseminated. To help combat this, we recommend creating an AI LLMs council. No, we don't mean putting a bunch of humans together to review all AI answers. We suggest implementing multiple LLMs AI agents to "fact check" each other and ensure the information being generated is of the highest quality.  

• Built in Bias: LLMs may reflect biases present in their training data, leading to unintended or inappropriate responses. Often this is overlooked and is worth addressing. Your LLMs are only as good as the information they are trained with. So, if the datasets that are being fed come with existing bias, it will be reflected in the LLMs generated answers.  

• Hidden Cost: Training and running LLMs requires significant computing resources and energy. Often, this is the hidden cost that lots of organizations run into at the end of the month. Before investing in LLMs, make sure to do the appropriate scaling cost exercise.  

For organizations, LLMs can be leveraged in several ways:   

• Customer Support: Automated FAQs and chatbots to provide instant answers and support to both internal teams and customers, enabling a true 24/7 self-service business model.  

• Document Generation: Creating reports, contracts, proposals, or marketing content quickly and efficiently. By far, this is the most common usage for LLMs and the easiest way to track return on investment (ROI).  

• Data Analysis: Summarizing large datasets or extracting insights from unstructured text in minutes vs. hours/days.  

• Language Translation: Providing real-time translation services for multinational teams.  

• Coding Assistance: Helping developers by generating code snippets or debugging suggestions. A Large Language Model (LLM) is an AI capability trained on large volumes of text to interpret prompts and generate language-based output summaries, drafts, analyses, plans, and code. Executively, it functions as a rapid language and reasoning engine: provide context and intent, and it produces a usable first draft that teams can refine. This is why LLMs are typically the entry point for enterprise AI — they reduce cycle time in communication and knowledge to work immediately. 

Conclusion

For most organizations we work with, LLMs establish the baseline of their AI investment as LLC can promote operational excellence through reducing cycle time in communication and analysis across the enterprise. However, LLMs alone do not reliably reflect what is true for your policies, documents, and current operating context. For those deeper needs, the next layer is RAG. In next week's op-ed, I will address how RAG can enhance your AI investment by grounding responses in approved sources, improving accuracy, traceability, and defensibility.

About the Author

My name is Huy H. Nguyễn, and I am a managing partner at Bayen Group. We specialize in partnering with organizations to plan and implement the Technology Enterprise Modernization Roadmap. If your organization is starting or in the process of its own Digital Transformation, don't hesitate to reach out to us. We would love to be your guide through the Digital Transformation journey.

Work with Us

Why Bayen Group
Our Projects
Our YouTube Channel