In a survey of 335 technology leaders of the Institute of Electrical and Electronic Engineers (IEEE), the world's largest global technology expert organization in electronics, communications, and computers, "AI technology," including predictive AI, Generative AI, machine learning, and natural language processing, was selected as the most important technology number one in 2025. In particular, 91 percent of the respondents said, "Public perception of AI will change significantly in 2025." Expectations are approaching that AI, which provides more accurate results and can make judgments and improvements, will change our daily lives. What will AI look like in 2025, which will usher in an era of innovation?
AI Era, Where Are We Now?
According to the IDC 2024 AI Report, which was commissioned by Microsoft by IDC(1), a global market research firm, the adoption rate of AI by organizations around the world has increased dramatically from 55% in 2023 to 75% in 2024. AI is now becoming not just an innovative technology but a business-critical technology. Following the introduction rate, what is the usability?
At the presentation of the 'Digital Marketing Insights (DMI) 2025' session held in November last year, Kim Choong-seop, CTO of 'Purple Io', a fashion e-commerce tech company, explained the development of AI technology by dividing it into five stages below.
|
📌 Step 1: The stage where AI is rarely helpful (for certain tasks)
📌 Step 2: AI is a little helpful to (any task)
📌 Step 3: Increasing the productivity of AI-using (employees)
📌 Step 4: AI does it on its own
📌 Step 5: AI performs most of the tasks from start to finish
|
He said we are now "in a transition phase from stage two to stage three." We are entering an era of practical AI, where we are actively using AI to improve productivity from the stage of auxiliary technology. AI, which was partially supporting the work, has become a key driver!
AI in 2025... Here are 4 AI prospects
① AI Agent
According to Gartner's top 10 strategic technology trends in 2025, AI agents are expected to independently determine 15 percent of daily tasks by 2028. AI agents are software that performs tasks on its own without human intervention. They are AI that automates processes, makes decisions, and interacts with people intelligently. Kevin Weil, the chief product officer of Open AI, also pointed to AI agents as the mainstream technology that will emerge this year.
AI agents are designed to be specific to a specific field or task, and to be able to recognize, learn, make and act on the environment like humans. Full-time work is possible to complement human physical limitations while providing insight into situations that humans cannot see. AI agents are expected to develop to be increasingly intelligent and to make human-like decisions.
② GAR
It is expected that the Retrieval-Augmented Generation (RAG) technology, which received attention in 2024, will evolve into a new dimension called Generated-Augmented Retrieval (GAR). RAG is a technology that combines search and generation, and language models derive results that fit the context of the user's search term. When asked a question, they access reliable external information and search for the latest information to generate a response. It was mainly applied to chatbots. GAR is a technology that complements generation with search, and AI models search for information more precisely. Based on the user's initial input, the AI model generates an expanded query and then searches for more appropriate information. As it improves search quality, more accurate search results can be obtained. GAR can be applied to search engine optimization and advanced search. It is expected that the search function can be further advanced as it can be used complementarily with RAG.
[Find out the difference between RAG and GAR at a glance]
| |
RAG |
GAR |
| key goal |
Improve quality and accuracy of answers |
Improve search accuracy and efficiency |
| key step |
Search for information and generate answers |
Optimizing search steps with generated queries |
| Order of Work |
Search → Create Answers |
Query Generation → Search |
| application case |
Live answers, chatbots, etc |
sophisticated information retrieval |
PLATEER's D2C e-commerce platform solution, 'X2BEE', has also launched an AI chatbot function. One of the X2BEE AI functions, 'AI Chatbot', is an operable AI that handles tasks as well as conversations on various topics. It integrates LLM learned based on large amounts of data with RAG that generates accurate answers to provide the personalized experience that e-commerce needs beyond the limits of existing chatbots. If you are curious about an X2BEE AI chatbot optimized for e-commerce, click the shortcut link. [Experience X2BEE AI Chatbot]
③ Vertical AI Solutions
General-purpose AI refers to an AI model that can be used in various fields and situations. ChatGPT is a representative example. The problem is that these general-purpose AIs can handle knowledge in various fields, but lack knowledge specific to a specific industry or domain. It is also difficult to reflect real-time data because it is distributed after learning. In areas where regulations are advanced, such as law, medical care, and finance, they can provide incorrect information and cause problems. The most important disadvantage is that it is difficult to customize according to corporate needs. There was a disadvantage that it could not be finely adjusted to meet the specific needs of the company.
Accordingly, customized AI solutions that reflect the unique characteristics and requirements of each industry and sector are expected to be in the spotlight in 2025. Vertical AI solutions such as medical AI with high accuracy, manufacturing AI with complex data analysis, and X2BEE AI specialized in e-commerce are expected to be active. In addition, flexible AI solutions that can be selected and combined according to the company's situation are expected. Even if it is not AI using a huge language model, it will be a trend to find and use the most suitable tool for the company's needs.
④ Multi-modal AI
Multimodal AI refers to artificial intelligence that processes and understands multiple types of data such as text, images, voice, video, and sensor data at the same time. In the past, AI was mainly a single modal AI that processed only one type of data. Recently, multimodal AI, which understands and integrates various data types at the same time, is emerging one after another. For example, Open AI's GPT-4 can process image input as well as text. The voice assistant Siri can process both text input and voice command and communicate with people.
New year's multimodal AI is expected to include new modals such as touch and bio-signals. The advancement of multimodal AIs on the market is also expected. Platier will also be interested in the future of interacting with AI systems in a variety of ways, including photos and videos, following text and audio.
(1) Provide DX-related research and consulting services for international Data Corporation, IT, and global markets
Subscribe to our newsletter
