Industry Research in the AI Era: Where Do Unique Insights Come From?

In many enterprises' AI practices, efficiency is the most noticeable change. Data collection, initial draft writing, chart creation, and page generation have all become faster. AI is also being integrated into various workflows, transforming the production methods of knowledge work.

At the scene of the "2026 4th National College Student Future Vision Industry Research Competition," university student teams demonstrated another possibility after AI enters business research: they are bringing AI into the real world. With the help of AI, they raise questions, connect scenarios, organize information and verify judgments, and also capture subtle signals in the market through their own life perceptions.

The imagination brought by these future business management talents lies in their ability to bring technology back to real life and discover new market opportunities from their daily experiences.

The championship team from Nankai University started with perfume and fragrance products, noticing that fragrance consumption is expanding into more scenarios such as home spaces, personal care, emotional healing, and lifestyle expression. The team from Xi'an Jiaotong University studied new-style health-preserving tea drinks, focusing on the young consumers' psychology of "unable to give up milk tea while starting to pay attention to health."

Behind different research topics, the same capability is being tested: using AI to improve efficiency, raising good questions in the real market, verifying complex information, and forming reliable judgments.

The practices of these university student teams also bring a question back to enterprises: When AI can complete information collection, framework construction and content presentation faster, where is the real moat of industry research?

When AI Enters the Research Process

In this competition, student teams have already begun to embed AI into specific links. Some teams used conversational AI to sort out materials and build frameworks, and used AI Agents to process data workflows and build interactive pages. Some teams adopted a more node-based approach to advance research, allowing large models and Agents to undertake part of the automated work, and then members verified the output results. Once deviations are found, the team can return to specific nodes for adjustments.

"AI is driving industry research to shift from traditional manual workflows to more structured AI-native research workflows. The change is not just that information collection, viewpoint extraction and content presentation are faster, but that the sources, data, facts, evidence, judgments and conclusions in the research process can be explicitly recorded, traced, reviewed and iterated." said Wang Chenhui, Managing Partner and President of Sullivan China. In his view, in traditional industry research, a lot of energy is consumed in repetitive work such as information collection, heterogeneous data integration and data migration.

The quality of industry research often depends on a series of small judgments. How the problem is defined, how the industry boundaries are delineated, which materials are credible, whether user interviews can be incorporated into the analysis framework, and whether the competitive landscape is accurately understood, all affect the final conclusion.

For enterprises, the application of AI cannot stay at the level of tool usage. What organizations really need to reconstruct is how research tasks are broken down into executable research units, how sources and evidence are recorded, what verification status the data and facts are in, who makes decisions on key disputes, and how the responsibility for final conclusions is confirmed.

A person who can call models, write prompts, and generate documents does not necessarily possess research capabilities. The moat of industry research is built on problem definition, task decomposition, verification nodes, and business judgment. Reliable research capability does not mean directly accepting the answers generated by AI, but being able to break down the model output into verifiable facts, explainable viewpoints, and accountable judgments, and after clarifying the boundaries of evidence, counter-evidence conditions and responsible parties, form conclusions that can be used for business decision-making.

The Easier Public Information Is to Obtain, the Scarcer First-Hand Insights Become

AI has greatly reduced the cost of obtaining public information. Industry reports, news, financial data, social media comments, and corporate official website information can all be collected, organized and preliminarily analyzed faster. As information increases, new challenges emerge: Where do unique insights come from?

In Wang Chenhui's view, AI can significantly reduce the cost of searching, organizing and preliminarily analyzing public sources, but public information is not equivalent to sufficient evidence. The differentiated value of people will be more reflected in the acquisition of non-public sources such as offline scenarios, interviews, and channel feedback, as well as the structured recording, cross-verification and judgment refinement of these qualitative data. He mentioned that those deeply hidden, unstructured industrial insights and first-hand data that are still offline remain areas that AI finds difficult to reach. This is also an important reason why the competition continuously requires university student teams to conduct real investigations. For industry research, first-hand insights are becoming a more important barrier in the AI era.

Real investigation sounds like a traditional method, but it has gained new meaning in the AI era. AI is good at processing existing information, while the real market provides information that has not been fully recorded. A moment of stay in a store, a moment of hesitation during a trial, and a judgment blurted out by an interviewee can all become entry points for understanding the problem.

Taking the fragrance and personal care track as an example, researchers need to enter specific scenarios: which scent attracts users, why they are willing to pay for a certain package, story or brand tone, and how personal care, home fragrance and emotional healing are interconnected. These questions are difficult to answer entirely from public materials.

For enterprises, more data, faster tools and more powerful models can improve the efficiency of trend identification, but they may not necessarily explain why users make a certain choice.

In the AI era, going to the scene and hearing real voices has instead become a key capability for organizations to understand the market. Trend curves on the screen can indicate changes, but inquiries, trials, hesitations and repeat purchases in stores can help enterprises understand the reasons behind the changes.

Therefore, enterprises need to establish their own "first-hand perception system." It can come from store visits, user interviews, channel feedback and community observations, as well as from the continuous information flow between sales consultants, customer service, product managers and research teams. The key lies in whether the organization can capture changes that have not been fully recognized by the market, and transform them into discussable, verifiable and actionable judgments.

How Future Business Management Talents Form Reliable Judgments

AI is lowering the threshold of technological practice. Li Zhuojun, Head of AI Talent Cultivation in Universities at Alibaba Cloud, believes that AI can improve efficiency and reduce costs, but it cannot replace human thinking, decision-making and judgment. "What AI does is to amplify human capabilities, rather than simplifying the human thinking process."

This difference has already emerged in the research process of student teams. The championship team from Nankai University mentioned that important data and factual information need to be further verified by returning to industry reports, corporate public materials, financial information, platform official websites, consumer surveys and expert interviews. As they said, "AI can improve efficiency, but the final judgment on accuracy and usability must be made by humans."

Wang Chenhui proposed that outstanding future industry research talents need to become compound talents who can master the AI + industry research workflow: they not only understand how models and data tools participate in research production, but also can take human responsibility for market issues, business scenarios, evidence quality and final judgments. In his view, "What AI cannot replace is precisely people's deep understanding of the market, understanding of business, and specific business execution capabilities."

A student from the Special Class for Young Talents at Xi'an Jiaotong University also expressed a similar understanding when talking about his growth path: "First, I need to rely on the university's strong science and engineering background to become a research talent, and understanding technology makes me more proficient in the industry. Then, I will cultivate my leadership thinking, business thinking, etc., to become a compound talent."

For future business management talents, the boundaries between technology, research and business are becoming more flexible.

This puts forward more specific requirements for enterprise talent cultivation. AI training cannot only stay at the level of tool courses. Being able to ask questions, generate texts, and make PPTs will soon become basic skills. What is more worthy of training is the ability to identify logical loopholes in a seemingly complete answer, judge whether the data source is reliable, make choices among multiple viewpoints, and transform the model output into accountable business suggestions.

Therefore, to cultivate talents in the AI era, enterprises need to put the training scenarios back into real problems. Give a vague business problem, let the team design a research path, use AI to improve efficiency, conduct market research, verify information sources, accept inquiries, and finally form actionable suggestions. Such training is closer to real business scenarios, and it is easier to precipitate individual capabilities into organizational capabilities.

Tools will continue to become more powerful, and the efficiency of research, analysis and expression will continue to improve. For enterprises, a more important topic is how to train people to raise questions, go to the scene, verify facts, and take responsibility for judgments.

What determines the value of an answer is always whether people truly understand the problem itself.

This article is from the WeChat official account "Harvard Business Review" (ID: hbrchinese), author: Xue Chun, published with authorization from 36Kr.