Exploration & Exploitation: Successful business model where people and AI will grow together
Amidst the recent trend of big data, since many activities of people and companies have been digitized, the amount of data available has increased, and the opportunities to utilize it have expanded. So AI technologies such as Deep Learning and Reinforcement Learning are attracting attention in order to leverage big data effectively.
Natural language processing, pattern recognition, image processing, speech recognition, machine translation, robotics, and a variety of other functions have been achieved with unprecedented accuracy. AI is now being used in a really wide range of ways to relate to every aspect of our lives, from medicine to transportation to power supply. With the rapid development and proliferation, AI is often seen as a big threat. It’s a collective feeling like AI will take away people’s jobs and replace human beings, using the technology to automate various work processes that humans do.
In fact, IT before AI has digitized various people’s work, improved efficiency and made people more productive. It has resulted in an increase of the amount of work per person, then allowing a smaller number of people to do the same work. Will AI technology make it further efficient and take it to the level of totally unmanned operation? That’s a frequently asked negative question, but I wouldn’t call it a groundless fear. Certainly business, and society, is changing in that direction. So, if AI technology has permeated the world, or if AI automates with a high degree of precision what people have been doing, then what work should they do?
I have given the following response to this question in one of my own speeches.
As we have studied AI, we have come to understand that there are things that humans can do and things that AI cannot do. The thing only human being can do, is to change the “framework” defining a problem to solve. AI can beat the top Go players, but it’s humans, after all, who can come up with more interesting games than Go. In other words, humans are creative in the sense that they go beyond the framework. AI, on the other hand, has the power to handle the long tail and big data within a given framework. It will become increasingly critical to think of businesses that combine these two forces.
Today’s AI technology is mainly based on machine learning, as mentioned above. Not only in machine learning, but also in the second generation of expert systems-based AI, there is basically some kind of “framework” first.
In that, it reaches the goal by doing the same thing over and over again, or by enhancing the accuracy. Frameworks can be rules or models, depending on the implementation, but in principle, AI cannot do anything beyond that framework. An AI that plays Go cannot play football. With transfer learning, you might be able to convert a model from AI of fighting Go to AI of fighting chess, but you can’t drive a car with that model.
An Overview of Transfer Learning Without Formulas
Transfer Learning — Learning How to Learn
medium.com
There may be some sense of level or abstraction, but only within a given “framework” AI can do its job. But human being is different. If you are tired of playing Go, you can play poker, you can play mahjong, or you might get hungry and want to start cooking without any context. It is a characteristic of human beings to be able to go beyond the “framework”.
Of course, there are areas in which AI excels. AI can do the same thing a ridiculous number of times and can perform complex calculations instantly.
It can search web pages all over the world, and if there are 10 million people, it can make 10 million recommendations to each. AI can process large amounts of transactions, data, at the scale of the Internet.
When you think about these two characteristics, that is, humans set the framework and AI processes it super efficiently within that framework and then scales it up. That kind of collaboration is going to be valuable. It is what businesses should be aiming for in the future AI era. In fact there is a movement that is making this combination of people and AI a reality.
If you look at the startups that are making great strides these days, you’ll notice that some of them are primarily offering machine learning-based AI solutions. They run their business as an industry or domain specific vertical SaaS with a focus on a specific group of customers. For instance, HR Tech, Legal Tech, Real Estate Tech, Medical Tech, etc.
As an example from the US, the project management company Procore is a unicorn that specializes in the construction industry, has achieved high growth rates and has a global reach.
Veeva Systems is a content management, data management and process management company for the pharmaceutical and medical industries. They have been steadily expanding the capabilities of their platform in the life sciences space. It is known as a successful case of vertical SaaS as it engages in PoC with lead clients and deploys effective functionality on its platform to a large number of customers.
In Japan, let’s take a look at FRONTEO, a leading legal tech company.
FRONTEO has developed KIBIT, an artificial intelligence that specializes in text data analysis. The KBIT-powered data analysis platform, Lit i View, reduces the time required for e-discovery. They also provide digital forensic support and have solved many client problems.
Targeting and concentrating on an industry or sector, as in each of these cases — construction, life sciences and legal — will of course limit the size of the market. However, it is a good way to continually improve the market fit of products supported by AI. Not only that, these vertical SaaS startups are enabling new ways for people and AI to collaborate. Plus they’re also achieving high growth rates.
If you just look at the vertical SaaS and feel they are eager to grow in number as a startup, you might think that they are just trying to advertise and sell to clients to get them to use it more, in general. I think the subscription-like model reinforces that image. But in fact, it doesn’t stop there.
Clients don’t just want to use the SaaS services, they want to take on the real challenges that they can’t solve just by using SaaS, or they want to make new challenges that they haven’t taken on before.
As mentioned at the beginning of this article, AI has been developing at a rapid pace, achieving new accuracy and performance one after another, expanding its capabilities, and is sometimes seen as a threat to humanity. In the face of the endless expansion of AI performance, it may seem as if one no longer has to play a business role, but this is not the case.
While there are certainly many progresses, AI solutions cannot evolve without any direction either. Simply improving performance and expanding with a number of enhancements would be an unnecessary evolution just creating garbage if it had nothing to do with what the customer truly wanted.
It requires the great capable leadership to identify the potential problems or challenges that a company really wants to solve, to identify them as topics worth addressing, and then to shape them into features or a framework to be implemented in an AI solution.
That’s why successful startups have a large corps of salespeople, consultants, or partners in these roles to help them explore the direction of AI development, not just by selling, but by fitting the market and defining and executing PoC’s on issues that have not yet been addressed.
As you can see, there is new collaboration between human beings and AI. People will challenge the problem in an exploratory way, invent new solution through PoC, demonstrate its effectiveness, and extend it. And then AI will implement the indicated functions, and leverage it with big data on a large scale across the internet.
This is a new form of business that can be called the E&E model, combining human exploration and AI-based exploitation.
The term Exploration & Exploitation is a word that originally appeared in reinforcement learning, which is an area of machine learning, specifically the Multi-Arm Bandit Algorithm.
In addition, as mentioned at the beginning of this article, AI has been developing at a rapid pace, achieving new accuracy and performance one after another, expanding its capabilities, and is sometimes regarded as a threat to humanity.
Multi-Arm Bandit Algorithm is a method of continuously maximizing profits by using two different types of actions, “exploitation” of previously experienced means to see how much profit can be gained with limited resources, and “exploration” of unknown means that may yield additional profit.
In general, a trade-off is established between the use of experience and the search for further benefits. So there is always the challenge of how to do it and how far to go to minimize lost income. However, in the E&E model, AI and people can play each role cooperatively to avoid trade-offs and achieve continuous growth.
The E&E model can be seen as a move to disrupt the software market traditionally for specific industries. Labor costs are high because the sales team and consultants responsible for exploration must be high-skilled individuals who define high-value issues. But that’s not a problem in and of itself. Rather, you need to hire the best people and actively put them to work in “exploration”. How do you find hidden demand from clients that are happy with the way business is done now? Delve into the client’s needs, include new workflows and tasks that have never been done before as potential solutions, conduct a PoC with the client, and once the effectiveness of the solution has been confirmed, hand it over to AI platform. AI will then leverage solutions on the platform to provide advanced services to thousands of clients.
This model can also lead to improvements in SaaS products and AI solutions from the customer’s perspective, and a platform that evolves in that way can be a powerful barrier to entry for competitors. And the members who conduct the explorations will always explore new topics in a viable way, based on non-repetitive, but specialized knowledge and discussion. By building on that, they gain more market-impacting problem-solving skills. This means that AI and people will grow together.
The E&E model can be established as a vertical business in all sectors. This trend will extend beyond the SaaS-driven software market to the consulting industry, which is based on highly skilled people. In the consulting industry, the model used to be seeing a linear increase in sales as head count increased. Hence, in order to make more sales, you need to hire more people. In contrast, in the software market, sales trends and the rise in the number of people are essentially independent of each other. Since labor costs are higher than the cost of electricity for a server, the consulting business naturally has a lower gross margin than a software company. The E&E model also makes it possible to break this revenue-staff ratio in the consulting industry.
For SaaS startups and consulting firms, the E&E model has two significant implications.
Firstly, it’s the advantage of leveraging data. The more you do business in the E&E model, the more high quality data you gather from your lead clients to help strengthen your AI (machine learning) model. It’s also a differentiator against your competitors. The business may not be easy to take off at first. However, as it grows, the E&E model also replaces complex operations that require client manpower with software. It will multiple the efficiency of business processes, ramp up your clients’ productivity, and even achieve automation.
Secondly, the E&E model has the potential to not only capture market share within an existing market, but to expand the size of the entire market. Except for the IT and Internet industries, in each industry or domain, for example, real estate, construction, health care, education, etc., software is still said to account for only 5% to 10% of the total market in terms of the direct production of the main goods and services. So there’s still a lot of room for vertical SaaS growth as AI services develop. The E&E model makes AI to progressively automate production and improve business efficiency each time. People will then be able to devote their resources to tackling new issues and themes, and enlarge the market.
The E&E model is a model in which people and AI both strive to grow skillfully through the interaction of exploration and exploitation. In a sense, it is an activity that aims to achieve greater market share and competitive revenues by leveraging the evolution of technology itself. This movement, which enables humans and AI to grow together, will be a pivotal trend in the years to come. Here, humanity will evolve as pioneers of a new era that will make a challenge and open up the world, guiding the future of many technologies, including AI.
Side note
If your SaaS and AI services are built on a platform that can respond elastically to the demands of the business, it will help boost the smooth operation of the E&E model. In other words, it’s also essential whether you utilize a cloud-native architecture.
