In February 2023, the explosive popularity of ChatGPT brought AIGC into the public eye.AIGC’s powerful generative capabilities have led to its widespread application in algorithm programming, language processing, and more.In interviews conducted by the research center on the theme of “Generative Artificial Intelligence,” we found that companies from various industries expressed positive views on AIGC’s potential for cost reduction and efficiency improvement to varying degrees.So, how exactly can AIGC help enterprises reduce costs and improve efficiency?This article will focus on the areas of marketing, R&D, production, and customer service, where AIGC is currently being applied most frequently.
Currently, the marketing field has become one of the most important scenarios empowered by AIGC.According to a report by Yiou TE Research, among numerous commercial scenarios for AIGC, marketing is currently the most mature in terms of service provider deployment and customer demand expectations, with specific applications including marketing content generation, innovative operations (virtual hosts, virtual scenes, etc.), and marketing strategies.
Image Source: TE Research “2023 AIGC + Marketing Value and Application Research Report”
Marketing content production is considered the most feasible direction for AIGC in marketing, applicable to creative advertising, product synthesis, and topic selection.
On one hand, based on AIGC’s powerful text generation capabilities, AIGC can provide creative inspiration according to the creator’s intent and can even quickly generate ideas directly, producing a large amount of copy, titles, and other content. In a survey conducted by the research center, the CIO of a smart retail service provider stated during an interview that the emergence of AIGC has lowered the threshold for creativity, with the direct adoption rate of content generated by AIGC reaching as high as 70%.
On the other hand, AIGC can help realize the industrialization of marketing creativity, simplifying the process of bringing marketing ideas to life. The chairman of an advertising media company mentioned in an interview that the emergence of AIGC has significantly simplified the process of shooting posters and advertising videos. Companies can use large models like Marketing Copilot to upload product images to the platform and then provide corresponding work instructions to generate promotional posters that match the product, reducing the advertising production cycle. If the background or scene is unsatisfactory, they can also switch backgrounds with one click, instantly transforming from a “beach” to a “street,” greatly improving the efficiency of marketing content production.
Innovative operations represented by virtual hosts and virtual scenes primarily target online marketing.
For example, through extensive proprietary data training, AIGC can achieve customization and large-scale production of virtual hosts, shortening the production cycle of hosts. Additionally, virtual hosts can work 24/7 without interruption, reducing labor costs for enterprises; the personas of virtual hosts are controlled by the brand, helping companies avoid risks associated with host persona collapses; in overseas business operations, virtual hosts can effectively solve communication barriers caused by language differences, reducing hiring and training costs for companies. Currently, several companies such as Xiaoice Technology and BlueFocus have implemented virtual hosts on traditional e-commerce platforms like Tmall and JD.com, assisting enterprises in reducing costs and improving efficiency.
In terms of marketing strategies, companies currently face issues such as low advertising quality and difficulty in targeting users. According to a Deloitte research report, the effective conversion rate of social media advertising is currently less than 3%. AIGC can assist business personnel in forming marketing strategies and advertising placement strategies by learning from extensive user preferences and marketing feedback data, providing more precise target audiences for advertising campaigns, enabling personalized marketing, and enhancing the breadth and precision of reaching consumers; at the same time, it can select the best promotional channels and placement periods to improve advertising exposure and communication effectiveness, thereby enhancing marketing efficiency. Currently, several companies, including Chopsticks Technology, have implemented AIGC in advertising placement, achieving real-time optimization based on public and private domain data, striving to create precise placements tailored to individual users.
In the R&D phase, AIGC can simulate human R&D processes based on given R&D requirements, quickly and efficiently creating entirely new content products and simulating and testing their effectiveness, significantly improving R&D efficiency, shortening product development cycles, and reducing wasted R&D resources.
For instance, in drug development, a startup named Profluent has achieved AI-predicted protein synthesis for the first time using a protein engineering deep learning language model similar to ChatGPT called Progen. Guo Chunlong, CEO of Shuimu Future, stated that entering the era of generative AI allows us to create previously non-existent new proteins or nucleic acid biomolecules based on functional and structural needs within an almost infinite space of protein sequences and structures. Industry experts believe that predicting new protein structures can provide numerous new protein target entry points for structure-based drug design, opening up new possibilities for drug development and improving the efficiency of new drug R&D.
At the same time, AIGC can evaluate existing technologies and products based on previous research and propose optimization plans, assisting companies in breaking through R&D bottlenecks, advancing the R&D process, or accelerating commercialization.
Currently, autonomous driving relies mainly on manually collected road data for image information collection and processes it based on designer-defined rules. As road data continues to increase, the complexity of the system also rises. However, higher system complexity limits development, especially when the system encounters uncollected roads or challenging scenarios, where increased rule limitations and high system complexity make it difficult to complete driving actions accurately, posing a commercial bottleneck for autonomous driving.
However, the emergence of AIGC has given industry practitioners new hope. He Xiaopeng, founder of Xiaopeng Motors, publicly stated on Weibo, “The next opportunity lies in AIGC and AUTO.” The reason, as mentioned by a leading investment institution in an interview, is that AIGC’s approach is directly end-to-end learning features; as long as the intermediate pathways and training signals are rich, it is possible to achieve a breakthrough. In April of this year, Haomo Zhixing released the world’s first autonomous driving cognitive model, DriveGPT, built on the GPT architecture. By incorporating human feedback reinforcement learning (RLHF) technology into large model algorithm training, it further improves the effectiveness and accuracy of algorithmic driving decisions. It is reported that DriveGPT has increased the pass rate of Haomo Zhixing’s autonomous driving system in recognized challenging scenarios, such as turning and roundabouts, by over 30%, and this technological breakthrough further accelerates the implementation of autonomous driving technology.
In the production phase, AIGC’s generative and reasoning capabilities can optimize the execution and management processes of product manufacturing. In April 2023, Siemens reached a cooperation with Microsoft to utilize AIGC to improve its industrial control workflows for continuous efficiency enhancement.Currently, both parties are integrating Siemens’ Product Lifecycle Management (PLM) software with Microsoft’s Azure OpenAI service language models to support workers who cannot use PLM tools, simplifying manufacturing processes and enabling production operators to report production issues using mobile devices.
In quality inspection, AIGC can optimize traditional industrial quality inspection algorithms, using visual large models to capture product differences and make real-time adjustments faster, improving the efficiency and accuracy of product quality inspections. For example, TCL has collaborated with Baidu Wenxin to build a large model for the electronic manufacturing industry, aiming to address the pain points of numerous production lines, complex quality inspection processes, and high precision requirements in the electronic manufacturing sector. Specifically, for semiconductor display panel defect detection tasks, due to the high precision of the panels, the yield rate is critical, and even a few microns of defects can deem the product defective. AIGC has improved the mean Average Precision (mAP) metric by 10%, while reducing training samples by 30-40% and development cycles by 30%, safeguarding product quality and achieving cost reduction and efficiency improvement in the quality inspection phase.
In production planning, AIGC can analyze historical orders, inventory, and other data to formulate production plans, monitor and analyze production data in real-time, further optimizing production plans and resource allocation, enhancing production efficiency; in safety monitoring, AIGC can detect safety indicators in the production process, issue warnings for potential issues, and even automatically resolve problems, reducing the risk of accidents during production.
Currently, the application of large models in customer service can be divided into two categories: one is to provide back-end services such as knowledge base optimization and opportunity mining, and the other is direct customer-facing inquiry services.
In back-end customer service, compared to the past reliance on manually written customer dialogue processes, AIGC can efficiently optimize and supplement knowledge bases based on conversational insights, providing merchants with precise dialogue services that enhance the business fit for front-end customer inquiries. It can even mine potential opportunities based on front-end dialogue data.
In inquiry customer service, AI intelligent customer service can provide consultation services across all time periods and scenarios, preventing customer loss; compared to traditional machine customer service, which can only answer routine questions, AI intelligent customer service can more accurately capture consumer dialogue intentions based on the perception capabilities of large models, matching response dialogues based on priority within the given knowledge base, thus improving the accuracy and efficiency of customer service responses.
For example, Baidu’s intelligent customer service provider, Jimu Fish, under the support of AIGC, autonomously learns from a vast amount of high-quality dialogues from human customer service, generating customized dialogue flows based on industry and business specifics, helping customers reduce the dialogue writing process from several hours or nearly two days to 5-30 minutes, saving nearly a hundred times the time. In practical applications, it achieves a 99% second-level response accuracy rate, improving the effective conversion rate by 36% for an education enterprise, with effective lead costs decreasing by 23%.
In addition, AIGC is also widely applied in IT development, daily office tasks, and other scenarios. AIGC’s powerful programming capabilities can not only assist in algorithm development, significantly improving development efficiency, but also automate the production of code documentation for easier maintenance in the future. According to GitHub predictions, within the next five years, up to 80% of code will come from AI systems. In daily office scenarios, companies such as Kingsoft and iFLYTEK have already integrated AIGC into their existing products to assist in summarizing meeting minutes, extracting text information, and other daily tasks, greatly enhancing daily work efficiency.
In addition to the cost reduction and efficiency improvement mentioned in this article, some experts also indicate that in the long run, technologies like AIGC can increase corporate revenues, although the timeline for achieving this remains uncertain. According to Goldman Sachs analysts, when artificial intelligence is widely applied, the median earnings of stocks in the Russell 1000 index could be 19% higher than the baseline. This means that the widespread application of technologies like AIGC could potentially increase corporate revenue by 19%, a promising development prospect worth looking forward to.
Under the wave of AIGC, with the continuous expansion of large model applications in marketing, R&D, production, and customer service, the role of AIGC in helping enterprises reduce costs and improve efficiency is gaining increasing recognition.As large model technology continues to improve, AIGC is expected to further bring revenue growth to enterprises.Companies should consider actively embracing AIGC technology and applying it across various business segments to enhance competitiveness and achieve sustainable development.
Source: Insight Academy WeChat Official Account
