AI Trends and Business Applications in 2025

• The presentation discusses AI trends and applications for businesses in 2025.
• Lộc Đặng, founder of Mindmaid which is a Chatbot AI development platform, presents.
• The main topics covered are AI trends in 2025, AI applications in business in 2025, and a FAQ section.

AI Agentic Automation

• AI Agentic Automation represents a new trend in automation using AI Agents

LLM Technology

• ChatGPT's LLM technology has rapidly advanced AI Agent capabilities.
• Levels of LLM apps include General Questions and Answers, Chatbots, RAG, Agents and LLM OS
• LLM Implementation has 5 stages, LLM stage 1 time is design . LLM stage 5 is data AI,

GPTs on ChatGPT Plus

• GPTs on ChatGPT Plus are considered AI Agents.
• GPTs allow users to create custom versions of ChatGPT that combine instructions, extra knowledge, and skills.

RPA and Agentic Automation

• AI Agents are driving the transformation of traditional RPA into Agentic Automation.
• AI/ML models automate repetitive tasks in scoped workflows, and handle complex general-purpose work with better planning and consistency.
• Rules-based systems handle rigid tasks whilst Agent-based systems handle flexible tasks
• AI Agents can automatically infer and use appropriate tools to solve problems
• Agentic Automation offers advantages over RPA, including user-friendly prompt engineering, processing complex workflows, and creating natural collaboration experiences.
• Microsoft's CEO announced that AI Agents would replace traditional software.
• Microsoft separates agents into personal, organizational, business process and cross-org agents

AI Inferential Ability

• Newer AI models like OpenAI's have significantly improved inferential capabilities.
• AI models are being trained to "Think before you speak" for better decision making
• AI can provide a chain of thought process to deduce an output

STEM and Code

• AI has been significantly improving at STEM fields and code
• Evals of OpenAI's models, including o1-preview and o1 as well are getting better results
• LLMs can now automatically generate plans to solve problems.

Multimodal AI

• AI is becoming multimodal, with the ability to understand images and sounds
• Multimodal capabilities means that Al understands information in its entirety
• Top multimodal models include gpt4-0, gpt4-o-mini, gemini pro 1.5
• It allows AI to reason with different data types simultaneously, and has been expanded to audio
• Gemini is a Family of Highly Capable Multimodal Models
• Multimodal AI can be used for various applications through realtime APIs with video & audio

Tech Industry Impact

• AI-Assisted Coding is rapidly evolving and highly effective.
• Salesforce will stop hiring software engineers in 2025 due to significant Al-driven productivity increases
• LLM Apps & LLM as an OS are gradually replacing traditional software architecture.
• Nvidia is developing Project DIGITS which is needed for these applications

User Experience

• User expectations of technology are shifting, and Chatbot AI and virtual assistants are replacing traditional search and browsing methods.
• There's a move towards Al Agents and Digital Humans.

Effective AI implementation

• To effectively implement AI, businesses should consider the changing landscape of 2025 in their strategy.
• The AI technical talent pool in Vietnam remains limited.
• Businesses should develop an AI application roadmap, build a culture that encourages AI usage, and cultivate internal AI personnel.
• Companies can experiment with some pilot AI projects
• The projects could focus on marketing, customer service, internal training, and data analysis and internal products
• The suggested timeline is 9 months to 1 year for these projects
• Build a culture of AI adoption through education