Guest introduction
Li Jiarui is the founder & CEO of Sentence Interactive, an alumnus of Y Combinator, "2023 Forbes China's Top 100 Most Influential Chinese Elites", "Forbes" 30 Under 30, and Microsoft AI MVP.
Co-author of Wechaty, the world's largest conversational RPA open source framework; co-author of the "Conversational AI" series of courses and promotion of the competency certification of intelligent dialogue architects with Baidu; author of the first conversational interactive book "Chatbot from 0 to 1" in Chinese, proposing best practices for the Chatbot life cycle and promoting the landing of Chatbot products; and producing the earliest generative AI open course "ChatGPT from 0 to 1" in China, proposing best practices for AI landing applications.
Founded in 2019, Sentence Interactive is committed to aggregating different IM (instant messaging) software such as WeChat, Feishu, and WhatsApp on a single platform through AI technology and automated processes.
Quick questions and quick answers
1. Can you introduce your products and core application scenarios?
By building a next-generation conversational marketing cloud driven by large models, through AI+RPA technology, Sentence Interactive can provide intelligent and efficient dialogue services based on different IMs such as WeCom, Feishu, 5G messaging, WhatsApp, etc. Specifically, Sentence Interactive uses private data to train expert-level digital employees for the enterprise and work efficiently in different IMs. Through general models, enterprise-specific knowledge, RPA, and vertical industry SOPs, digital sales employees in the big health industry and insurance industry, as well as digital police and digital grid personnel in the government affairs field, have been implemented.
The Sentence Interactive team continues to innovate in the fields of RPA, intelligent dialogue and private domain operation services, and its products and services have been applied in many industries, covering consumer goods, finance, pan-Internet, government affairs and other fields, with customers including Procter & Gamble, Haidilao, State Grid, Chinese People's Insurance, Volcano Engine, Baidu, Public Security Bureau, etc.
2. What is the current situation of product development and market expansion, and what are the biggest challenges and gains encountered in 2023?
We have accumulated relatively deep experience in the field of private domain RPA, precipitated many best practices and customers, and the business has fully realized self-operation. On the contrary, the product of large model application will only start to be done in 2023, we will start to do it at the beginning of 2023, and we have iterated more than a dozen versions in the past year, and we have done a major refactoring, but I think it may still be in the very early stage, because the cost of delivery is still relatively heavy, and with the increasing degree of productization, the delivery cost should be close to 0, at least for R&D, it is more reasonable to approach 0, because this is how our RPA products have come over.
On the market side, we basically haven't expanded much, mainly because the company's business students are easy to make some cases when they do some deep digging among old customers, and in 2023, we will mainly focus on doing some deep KA, and we have recently made best practices and plan to start taking the initiative.
The biggest challenge in 2023 is that it can't be delivered, and the whole year feels like a running state. Although the whole capability of the large model is becoming stronger and stronger, there is still a lot of tuning to be done in the vertical field, and the amount of engineering is actually not small, but with the increasing degree of productization, I believe that it will get better and better in the future, and the current challenges will be regarded as future products.
The biggest gain in 2023 is to learn more and more practices from customers, TO B SaaS entrepreneurship has a very interesting thing, the first product may rely on the founder's insight, the further you go, the more customers teach you how to make the product, you will find that a lot of your cognition is very shallow. For example, when I first said that I wanted to be a digital sales employee, my understanding was that I needed to build trust to conversion, but the customer would be very clear about how to complete a closed loop of sales for me, from complete distrust to distrust, how to do it, from distrust to starting to build trust, what to go through, and then from the beginning to building trust to the initial establishment of trust, to the clear establishment of trust, and then to the inflection point of trust transaction, and what key things to go through in the middle of the transaction, and what should be done from complete trust to referral, the split of these contents in addition to letting me know what our product should doI even began to reflect on whether our current business classmates still have a lot to improve.
3. What opportunities does generative AI and large language models bring to entrepreneurs, and how is this wave of entrepreneurship different from the past?
I think it is divided into two parts, more generally, on the one hand, the large language model greatly improves productivity, on the other hand, for startups, it greatly reduces the difficulty of doing things, with a large language model, if the learning ability is a little stronger, one or several people can quickly build an MVP, for example, I did a sharing "how AI helps a person to carry out cross-border e-commerce business", from the selection of suppliers, the preparation of marketing materials, automated operations, etc., can be in AI With the help of saving money and starting a professional business. For example, if you want to write copywriting, design drawings, and official websites, you have to find professional people to do it, and today you can use a large model to do it, although there is still a certain distance between today's work and that of senior practitioners, but it can help entrepreneurs to verify the feasibility of a foundation.
As the professional skills that AI can provide become more comprehensive and powerful, the main thing we want can basically be built with AI to build a prototype, and even more can be completed in the future. I think in the age of AI, we're going to see companies getting smaller and smaller, but products that have a lot of impact. There will be fewer and fewer labor-intensive companies, and the industry will gradually shift from being driven by people to being driven by algorithms and computing power. From this perspective, AI will certainly replace a large number of people, but it will also open up opportunities for smaller companies and individuals that have never been available before.
On the other hand, mainly from what I have done, because I have been doing enterprise services for many years, I think that in the future, the workflow of enterprises will be reshaped by AI. From my own perspective of making SaaS software, all current SaaS software is very anti-human, and users are actually completing a task according to the idea of the product designer of SaaS software. For example, in our system, it takes dozens of clicks to create a marketing task, which also brings a lot of use costs and training costs, on the one hand, you can say that your product is thicker and has barriers, but on the other hand, I think the so-called "product thick" is to make excuses for a very poor human-computer interaction.
Human-computer interaction should be simple, in fact, many times it should be done in one sentence, and then it should be AI-driven, and all these interactive interfaces will be eliminated, for example, a marketer only needs to say "help me give a coupon to pay customers who place orders on Douyin to promote Douyin activities", and then the system will automatically synchronize data from the e-commerce platform to the CRM platform, automatically analyze customer groups, identify different types of customers and tag, automatically create copywriting and product images to be promoted, automatically create reach tasks, and so on. That's the next generation of SaaS, and that's where AI reshapes workflows. I think there are a lot of opportunities in this area.
When it comes to the wave of entrepreneurship, I think the wave is divided into big wave and small wave, and the arrival of this wave of AI will be a big wave like the mobile Internet. We say that the current AI model has become a foundation model, that is, an infrastructure-level model, and it will become an infrastructure in the future. I've heard a particularly interesting example before: more than 100 years ago, the plug of the electric toaster was a lamp head, because people at that node more than 100 years ago were imagining what the future would be. At that time, the engineers envisioned that all homes would have electricity in the future. What do you use electricity for? Light a light bulb.
At that time, they didn't know that there would be refrigerators, washing machines, and bread machines that would also need electricity, so they didn't design these sockets at home. At first, most people thought that they were going to generate electricity, just like now, to make big models, to do cloud computing, but it turned out that in the end, these would converge into the country's critical infrastructure, and only a few companies were qualified to do it. But there will be a lot of people who have the ability to use electricity to give human food, clothing, housing, or firewood, rice, oil, salt, sauce, vinegar and tea, all to it, which was called electronic and electrified at that time, and there are still many such opportunities on the large model. In fact, many people also use the same way to compare the mobile Internet, so it is often said that the release of ChatGPT is an iPhone moment, so this wave must be much bigger than the various small waves that have been hot in previous years.
4. What are the challenges facing AI applications in the current domestic market environment?
From the perspective of ToB, in fact, whether it is in China or overseas, the real challenge of landing AI applications is actually similar, that is, to find the value point that truly matches the needs of enterprise customers with their own AI capabilities. The capabilities of AI sometimes seem cool, but in fact it is not a panacea, and it takes a lot of engineering and co-creation with the customer to really solve the problem.
From the perspective of enterprise customers, whether it is domestic or overseas, the needs of each enterprise are scattered and diverse. We look at OpenAI's ARR of $1.6 billion, and a large part of it comes from enterprise customers, and that's because OpenAI has the best large models in the world. And more large-model companies don't actually generate much revenue on the B side, let alone those in the middle and application layers. Enterprise customers are reluctant to pay, the core is because they have not solved their problems and have not really achieved cost reduction and efficiency increase for them.
The good thing is that more and more enterprise customers are realizing this, no longer fantasizing that AI can solve all their problems at once, and are willing to work with SaaS application companies to co-create and explore a truly feasible path to commercialization.
5. What do you think are the key factors that AI startups should consider in the formulation and execution of product development and marketing strategies?
On the product research and development side, I think there are two very important points: the first point is what is the core competitiveness of the enterprise, which enterprises should think clearly, do not do everything, because of limited resources; On the first day of product development, we should stand with the target customer, and continue to co-create with the customer throughout the product development process, and find the PMF as soon as possible.
In terms of marketing, different product strategies are different in different industries, and I will only share some of the lessons learned from the cases I have seen. If you are a large enterprise customer, then it is very important to build your own sales team as soon as possible. Release products as early as possible, verify them with customers as long as they can solve customer problems, and commercialize them as soon as possible. Good growth comes from good customer relationships and word of mouth.
If it is a ToC product or a PLG-type SaaS product, if you have verified the product requirements and found the user value of your product, you need to pursue creative and rapid growth. Sometimes the product itself is the best growth booster, think about where your product can bring you cheap or even free traffic, and do accurate reach and conversion for the target group in the right channel and at the right time. If the company has a large amount of go-to-market expenses and the industry is in the early stages, and the market has a sufficient ceiling, then it needs to do a large number of paid go-to-market and factor user LTV into the cost structure.
If your products do not have a significant generation gap with competing products, and there is no large amount of marketing costs, then you should grow while commercializing in a down-to-earth manner. Products in the age of AI are naturally magical in theory, and if you can't spread the word, sometimes it's better to continue polishing the product.
6. In the face of the risk of "dimensionality reduction" that giants may bring, how should AI startups maintain their competitiveness?
No single business can eat the entire market. Giants have the advantages of giants, such as having more resources, whether it is computing power, talent or data are much higher than startups, so giants have stepped down, trained their own large models, and released their own applications.
Startups don't seem to be as good as big companies, but startups also have the advantages of startups. Because entrepreneurship is flexible enough, the ability to iterate and innovate is stronger, and because resources are limited, it is relatively more focused. We can create local advantages first, and then seize part of the market opportunities.
On the other hand, the relationship between large companies and start-ups is not an either/or relationship, but a relationship of mutual cooperation and complementarity. When the AI wave is just coming, it's better to focus on the market and users than on the competition. If you can solve the customer problem by yourself, you can solve it by yourself, and if you can't solve it by yourself, you can find a partner to solve it together, and this partner can actually be both a start-up company and a large factory. Let's work together to serve customers well and make the market bigger, which is a fair opportunity for all companies at the beginning of this great era.
Finally, on a more ambitious scale, I think that in the new traffic distribution, new forms of interaction, and new internationalization, young AI companies can make a different giant, although it seems a bit too big now, but step by step, there is not necessarily no opportunity. Today, whether it is search, e-commerce or social networking, in fact, a lot of changes are taking place, and the successful form of giants may not be the successful form of tomorrow, but the needs of people have always existed. And the older generation always has to be old, and the younger generation is also where we have to go, after all, we have better physical strength and energy, and as long as the young AI company continues to maintain the speed of innovation, the opportunity is still very large.
7. Some entrepreneurs believe that the current capabilities of large language models are not enough, which limits the development of AI applications/products. What do you think about this?
Most of the capabilities brought by this round of AI are attributed to the capabilities of the underlying large models, so it is true that many problems cannot be solved at present, which can essentially be attributed to the lack of capabilities of large models.
But this should not be the concern of entrepreneurs, or the primary concern, but should be the concern of those large model companies and manufacturers. The core of a startup company is to pay attention to customer value, the market, and the problems they want to solve. If the ability of the large model is limited, then within the scope of the large model, some engineering and unique innovations between the means and product interaction layers are superimposed to solve the problems that can be solved for customers. This is what a startup should do, and only then can a startup survive and develop.
Don't wait until the big model can solve everything before entering the game, there is no chance of a startup at that time. On the contrary, it is possible to explore and create value at a time when everyone may not have the best solution, but it is possible to seize greater opportunities.
8. How do you view the new paradigm of "human-machine collaboration" brought about by large models and GenAI, and how can enterprises adapt to and innovate the changing trend of automation process business upgrading?
In 2020, I published my book "Chatbot from 0 to 1", and this year I also published the second edition of the book, How to Build a Chatbot from 0 to 1 in the Era of Large Models, and in the last subsection of the last chapter, I say this about "human-machine relationship":
For enterprises, it is necessary to concretize seemingly abstract and cool AI capabilities to help enterprises better reduce costs and increase efficiency, improve decision-making and competitiveness. Because AI has far more advantages than humans in some aspects, such as AI can automate repetitive work tasks, for example, AI can generate a large amount of average content, for example, AI can pull information from a given knowledge base to interact with people 24/7, and so on. Enterprises should recognize the advantages of AI that are different from people, and "let the machine belong to the machine and let the person belong to the people", so as to better help the enterprise achieve better operation and create greater value.
It should be noted that enterprises should not blindly innovate for the sake of innovation, but should start from specific problems to understand the boundaries of AI capabilities to better serve enterprises. More important than using AI to innovate an organization is to innovate the organization itself, pursue organizational goals and organizational capabilities, and improve the overall management level of the enterprise.
Registration channel
- END -
Author: Li Jiarui
编辑：Qiuping
- Recommended -
Industry Insight | Global venture capital is cold, can the AIGC track stand out? | AI in Finance: While technology leads the future, it faces the dual challenges of talent and data Creative Change and Market Disruption: AIGC's Comprehensive Impact on the Advertising Industry|Creativity and Opportunities in the Future of Digital Human and Generative AI| How AIGC Revolutionizes Content Marketing|Industry Data + Scenarios: Under the AI Trend, the First Companies Are Favorable|AI+ Finance: Leading the Transaction-Centric to User-Centric Growth Transformation|Exploring the Future: AI-Agents Leading the Artificial Intelligence Revolution|From Idea to Revenue: AIGC Shows a New Commercialization Perspective|The AI Era of Cross-border E-commerce: A Comprehensive Transformation from Factory Innovation to Customer Service
Company Research | Baidu|Cambrian|Kunlun Wanwei|Blue Cursor|iFLYTEK
Industry Reports | November 2023 Global AIGC Industry Monthly Report| 2023 H1 Global AIGC Industry Semi-Annual Report
Industry Atlas | 2023 China AIGC Industry Atlas V3.0
For more historical articles and report collections, please click on 👉 the Most Complete Report Collection