[PYTHON] Summary of know-how and tips for AI new business planning that AI engineers want to know
In this article, we will inform you of know-how and information on entrepreneurship and new business planning for AI engineers and business people related to AI.
There are many contents related to new business planning that are not specialized in AI, but please forgive me. AI-related content begins later in this article.
This article is [1] Quoted and posted famous advice on startups [2] On the other hand, describe my personal opinion (Ogawa) I will write with the composition.
** List of article series for AI engineers ** Part 1. Summary of Python implementation know-how and tips that AI engineers want to be careful about Part 2. Summary of know-how and tips for AI new business planning that AI engineers want to know (this article)
** Table of contents of this article ** Introduction 0. People / organizations you need to know when learning about entrepreneurship and new business planning
- How to find an idea
- How to grow ideas
- How to interview users
- About AI new business planning
- Create MVP
- How to make a pitch
- How to create a team for a new AI business at the end
Introduction
The three circles of the Data Scientist Association, shown below, are well known.
As the ability required of data scientists including AI engineers
・ Data science power (AI algorithm system) ・ Data engineering capabilities (data infrastructure, computer basics)
In addition, ** business power ** is also required.
● Announcement of data scientist missions, skill sets, definitions, and skill levels
In this article, we will summarize information about the part of business power that plans a new business.
We will post stories that are common to both starting a business as an entrepreneur and planning an in-house new business as an intrapreneur.
0. People / organizations you need to know when learning about entrepreneurship and new business planning
· [Paul Graham](https://ja.wikipedia.org/wiki/%E3%83%9D%E3%83%BC%E3%83%AB%E3%83%BB%E3% 82% B0% E3% 83% AC% E3% 82% A2% E3% 83% A0): VC (venture capital) "[Y Combinator](https://www.amazon.co.jp/dp/ Founder of 4822249468 /)
· [Peter Thiel](https://ja.wikipedia.org/wiki/%E3%83%94%E3%83%BC%E3%82%BF%E3%83%BC%E3% 83% BB% E3% 83% 86% E3% 82% A3% E3% 83% BC% E3% 83% AB): Paypal co-founder and now an angel investor. Author of the book "Zero to One"
・ [Sequoia VC (Sequoia)](https://ja.wikipedia.org/wiki/%E3%82%BB%E3%82%B3%E3%82%A4%E3%82%A2%E3%83% BB% E3% 82% AD% E3% 83% A3% E3% 83% 94% E3% 82% BF% E3% 83% AB): The largest VC in the world. So far invested in Apple, Google, Yahoo, Instagram, Airbnb and more. Sequoia's entrepreneurial theory is detailed in this Medium. https://medium.com/@sequoia
1. How to find an idea
1.1 Starting from a unique truth that only we feel strongly
Peter Thiel, the author of the book "Zero to One", is famous for the words he always hears in interviews.
"What is the important truth that few people agree with?" ● The essence of startups learned from Peter Thiel's words
** [Ogawa] ** It is highly possible that someone else (especially a large company with abundant financial resources and human resources) is already working on ideas and ideas that everyone agrees with. There is also a high possibility that it will be imitated and crushed later.
Instead **-Facts based on what you, your team, and what you can experience and think about. **-The truth (fact) that others cannot notice or confidently agree with.
It is recommended that such "truths that only we are aware of" be the starting point of our ideas. That also leads to discrimination.
1.2 Technology-savvy people are leading the way in ideas by themselves
Paul Graham, co-founder of Y Combinator, said:
It can be said that those who are strong in technology are leading in terms of ideas by themselves. This is because technological progress is rapid, and in the past, unrealizable ideas can be solved without anyone noticing. Young people have the same advantage. Summary of advice from Paul Graham to startups
** [Ogawa] ** "People who are strong in technology are leading in terms of ideas by themselves," and the reason is stated. That is, Since AI engineers are constantly researching the cutting edge of AI technology, it is possible to realize what other ordinary people thought "it would be impossible to realize such a thing ..." with the power of advanced technology. You can know the possibilities quickly.
From that point of view, in my case (Ogawa), when reading a paper on advanced AI technology or listening to a presentation, ** Papers that use AI to solve new problems, rather than papers that improve performance in existing tasks (state-of-the-art) ** The emphasis is on research.
However, even in the state-of-the-art system, technical papers that change the deep learning algorithm system itself, such as the BERT paper, are very important as new important technologies.
1.3 Start with issues, not ideas
Paul Graham states ** the importance of avoiding starting ideas ** when considering entrepreneurship and new business planning.
** The best problem to solve is your personal problem ** Successful founders know "problems that others are unaware of." And the best ideas are "I want it," "I can create it," and "Most people don't realize it's worth it." ..., Rather than thinking "Is the idea good enough to start a business?", Just fix what you think is the problem. By doing so, you can finally make what many people want. Summary of advice from Paul Graham to startups
** [Ogawa] ** In order to think of ideas for starting a business or planning a new business, it is important to let go of thinking. Rather than thinking "Is the idea good enough to start a business?", Just fix what you think is the problem. As you can see ** Instead of thinking about ideas, determine the problems / issues to be solved first **.
If the problem to be solved is decided first, innumerable ideas will be born.
It is a pitfall due to the idea first, It's easy to avoid the misunderstandings of your own, such as "You can blindly go to a specific idea !?"
In addition, when setting problems / issues to be solved, it is good that you and your team have experienced and thought about them. This is also strongly linked to "1.1 Starting from a unique truth that only we strongly feel."
Others are more familiar with problems and issues that are not familiar to us, and above all, we are not passionate about solving them. It is also difficult to collect information and interviewees.
However, the word familiar does not have to be taken very narrowly. Especially in the case of new business planning rather than starting a business, it would be good to have a business partner company that usually sells existing products.
However, not all existing business partners are good, ** ・ Friendly and easy to communicate ** ** ・ I want to solve this customer's problems from the bottom of my heart ** It's important to be a customer who seems that way.
1.4 Become a friend of advanced technology and solutions, and work in love with problems and issues
** [Ogawa] ** I tell team members to stop thinking from ideas, solutions, and solutions as much as possible.
And what I always tell the members is ** "Become a friend of advanced technology and solutions, and work in love with problems and issues" **
That is the attitude.
This may seem contrary to Paul Graham's quoted advice: "Because it's an AI engineer, let's take advantage of knowing advanced technology."
But that's not the case.
Love blinds people.
If you get to know advanced AI technology and fall in love with advanced AI technology itself, you tend to try to solve anything with that advanced technology.
As a result, they are excited about cutting-edge technology and fall in love with them, creating and selling some products and solutions. Then, we will search for problems / issues that can be solved by the product later and try to apply them to the solution.
This method is difficult to find customers and difficult to succeed because you will later find someone who will need it even if you pay for it.
"Advanced technology, solutions, ideas" and "problems / issues" are both important, ** I want the target of love to be a problem / issue, not an advanced technology **.
However, ** We are always friends with advanced technology, and in order to solve problems and issues that we notice, it is important to have a team structure that allows us to immediately implement and try advanced technology **.
Otherwise, we will differentiate ourselves from other companies, take the lead in our efforts, And above all ** Rapid construction of prototypes & continuous idea verification and improvement process of showing to customers for feedback ** cannot be realized.
The following is an example for Facebook.
Knowing how to hack also means that when you get an idea, you can do it. It's not an absolutely necessary skill (Jeff Bezos can't), but it's a strength. Student Directory with Face Photo When considering an idea to bring college online, not only can you think "it's an interesting idea", but "it's an interesting idea. Let's make an early version tonight". If so, it will be a great strength. Even better if you are a programmer as well as a target user. The cycle of creating a new version and user-testing it can be run in one head. ● How to get ideas for startups (Original author: Paul Graham)
However, as a precaution for teams with implementation power, I would like to mention that we do not try to make a ** perfect product ** from the beginning to solve the problem.
It doesn't matter if it's rough It is called ** MVP (minimum viable product) **, "Is the problem-solving solution in this direction really valuable?", Prototype the smallest configuration to confirm it.
Since it is misunderstood that it is a product when it is called Product, there is also an idea that it is better to call it Minimum Viable Prototype these days. (Book "Inspired").
MVP will be explained in detail later in this article.
1.5 What you notice, not think about the problem
Paul Graham describes how to find the problem:
** The problem is not to think, but to notice ** The idea of starting a business is not "thinking" but "noticing". At Y Combinator, we value the ideas that come from the experience of the founder. If you notice the inefficiencies that exist now, it can be a problem. Summary of advice from Paul Graham to startups
** [Ogawa] ** Problems to be solved and their solution ideas are not easy to think about.
This does not deny the team's brainstorming and thinking.
However, for the time being ** Forcibly define problems / issues to concentrate their resources in short breaks such as just one day, one week, one month **, That means it's not a good idea.
Then what should I do?
always, · Realize yourself, your team, your company, these visions ・ Feelings of contribution to customers ・ Better the world and create a more beautiful future
By living with these thoughts every day, we can suddenly come across and realize the problems we need to solve.
1.6 Seeing ideas from the future
Also, in Paul Graham's words, he often uses the phrase "seeing ideas from the future."
** Notice ** When you're living in the future in something, the way to notice startup ideas is to look for things that seem to be missing. If you really are at the forefront of a rapidly changing field, there will be clearly something missing. What's not clear is that they are startup ideas. ・ ・ ・ You should pay particular attention to things that annoy you. It is convenient to take the status quo as a matter of course not only to make life (locally) more efficient, but also to make it more patience. Everything you'll get in the next 50 years, but not yet, if you knew it, would make your current life quite limited. It's the same thing that a person sent 50 years ago from the present age would feel on a time machine. When something annoys you, it may be because you live in the future. ● How to get ideas for startups (Original author: Paul Graham)
** [Ogawa] ** I will explain Paul Graham's concept of "living in the future and looking for something lacking in the future." There is a business model called "time machine management" (implementing successful overseas business in advance in Japan), but it is different.
"Seeing ideas from the future" means imagining a future that has not yet visited the world, including overseas, and working on solving the problems necessary to realize it.
It's similar to "backcasting thinking" in design thinking, but it's also a little different.
Backcasting thinking is a way of thinking that identifies the phenomenon that becomes a problem when going through time and time as it is, and works on solving the problem. Backcasting is a stance that starts from the future, which is an extension of the current situation, and returns to the present.
It is important to acquire a culture in order to realize "seeing ideas from the future".
The word culture is difficult to define,
** "Cultivation is knowledge of what human beings and human society are and how they have been formed." ** I define.
When planning a new business, it is common to use the framework of PEST analysis (recently PESTEL).
PEST is an acronym for "Politics, Economy, Society, and Technology."
However, there are few working people who can use this PEST properly.
For most working people, modern politics, modern economy, modern society, and modern science since the 19th century, how did these change and become established in Japan and continue to this day? Also, I haven't been able to understand and understand how the world was born, established, and continue to this day, and make it my own.
Properly learn about the transition and current history of the birth of "modern to modern politics", "modern to modern economy", "modern to modern society", and "modern science to modern science" in Japan and the world. Without it, it is difficult to see the future society in the future.
If you can't see the flow from these pasts, you can't see the future that will flow, and as an entrepreneur (intra-entrepreneur), you can't draw what the future should be and what kind of future will be beautiful and accepted by humankind. is.
The future image imagined by humans who lack the cultural knowledge of how the present is born from the past is just a delusion.
Well-educated, and "I think the future society should be like this, but it's not enough, it's not beautiful as it is." With that in mind, you will be able to see your ideas from the future.
It's not easy to get a perfect education, but ...
For example, when I write a book or article Look up domestic and foreign books, Such books and articles should be in Japan in the future, but they don't exist at present and aren't beautiful ~ I try to write about what I feel like that.
This is the end of "Chapter 1 How to Find Ideas".
2. How to grow ideas
2.1 Repeat dialogue with customers
Paul Graham's advice is three points for creating and nurturing good ideas.
- Make what people want
- Get to know the user
- Let's not scale
** Get involved with users ** Product development is a dialogue with users that really begins after it is released. And if you want to benefit from user interaction, you'll need to change your ideas. The idea at the time of starting a business is not a blueprint but a hypothesis. Do whatever you think is the best at that time.
** Let's not scale ** Do things that do n’t scale – do things that do n’t scale to know the user. That way, you can provide amazingly good service to your customers and get to know them better. It's better to be loved by a few users than to be loved by many.
Summary of advice from Paul Graham to startups
** [Ogawa] ** Once you have decided on the problem / issue to focus on and have an idea for the first solution / solution, For some reason, people tend to be falsely convinced that the idea will succeed.
Perhaps it's because our team, which seems to be the best, came up with a lot of ideas to solve the problem and decided that it was the best.
Also, if it is a new business plan within a company, it may be because it is an idea that the management has endorsed compared to other ideas.
It's a sophisticated idea through the "internal process in the family", so I'm sure it will succeed ...
However, this first solution / solution was only rubbed by "internal processes in the family". I wasn't crushed by a real customer.
As long as we build a solution that solves a customer's problem, we need to interact with the customer (user) and get feedback to make a better solution.
In some cases, the current solution should be completely abandoned.
However, the larger the company, the more And the closer it is to B2B than B2C, the more
It is difficult (difficult to implement) to engage in dialogue with customers and improve initial ideas and first solutions.
However, the act of interacting with customers / users about ideas and omitting the process of refining the solution is
** Invest resources in building unpurchased solutions that customers don't really need ** It is very likely to generate complete management waste and loss.
To avoid that, refine the solution. The important thing in the process is to ** do something that doesn't scale **.
The word "do something that doesn't scale" is a word for startups, so it's hard to understand. I will explain.
Entrepreneurship of a startup is a world where business is scaled up. Therefore, the act that does not lead to scale-up is, frankly, wasted time.
However, the process of interacting with customers / users about ideas and refining the solution is It's an act that doesn't directly lead to a quick scale-up, but it means that you should cherish it.
If you neglect the process of refining your solution, It also puts resources into building problem-solving solutions that customers don't really need.
In addition, it also leads to the risk of creating a solution similar to other companies that focuses on the same issue.
Even in the case of discrimination from other companies, the "truth that few people agree with" that only we know Finding in the dialogue with the customer (user) creates true differentiation.
Interact with customers (users)
** 1. Are the problems / issues that we have decided to solve really the problems that customers want to solve even if they pay? ** ** ** 2. Is the solution to the problem really what the customer wants to pay for with the current idea? ** **
Interact with the customer (user) to confirm these two points. The state in which this can be confirmed is called ** PSF (Problem Solution Fit) **.
As the PSF requirements state that the customer is ... ** PSF can never be achieved by in-house processes alone **.
And the important point is ** This dialogue with the customer must be done by the entrepreneur himself or the person responsible for the idea of new business planning **.
PSF is difficult because the person in charge of ideas cannot feel the real situation of the customer by letting the sales employee bring the product or prototype and listening to the customer's reaction from the employee.
At the very least, you need to accompany a sales employee.
And if a company wants to realize new business planning, as described above,
Build a mechanism for customer dialogue process for PSF, After confirming the PSF, we will move to the process of making a solution in earnest.
It is necessary to build a flow of the step-by-step new business planning process.
If you order to propose a new business without building this process, as mentioned above,
** Invest resources in building unpurchased solutions that customers don't really need ** It creates a complete management waste and loss.
2.2 Identify customer Burning Needs
** Identification of problems and issues ** is very important in both PSF (Problem Solution Fit) and idea creation.
There are many problems and issues to be solved. All human beings want to live a better life, There is no end to the problems and issues such as "I'm glad if you can solve it".
Issues / issues that should be focused on when starting a business or planning a new business
** "If the customer is really worried about the problem / problem and can solve it by paying money, pay the money immediately" **
It needs to be such a problem / issue.
The point is that "I want to solve the problem immediately even if I pay for it." Such problems / issues and customer needs are called ** Burning Needs **.
It means the need to extinguish the fire immediately after it is burning.
Autify's blog tells us about Burning Needs as follows.
** What are BtoB products? ** Then, what kind of products can be sold? Basically, there are only two types of products that can be sold by BtoB. ・ Products that increase customer profits ・ Products that reduce customer costs The bigger the company, the more you care about ROI, so it will be difficult to sell products that cannot be explained by these two points.
** Good, please let me know if you want to use it when completed ** In response to this reaction, I thought that I had to develop more because I couldn't sell it as it was. But this is a complete mistake. Products that respond in this way are very likely to have not resolved their customer's Burning needs. Because they don't want it right now.
** Win a contract without a product ** ... I remade the presentation slides, made a simple picture-story show demo app, and took a video. I did everything so far overnight. And when I took these to the customer appointment the next day, the reaction was completely different. Even though I didn't have the product yet, I said "I will buy it!". This is how the customer reacts when they find their Burning needs and achieve the right challenges. ● Solving customer Burning needs
** [Ogawa] ** Whether you are starting a business, planning a new business, B2C or B2B, we will repeat dialogue with customers.
"I would be happy if there was such a solution, and I would be happy if you could solve this problem."
That level of solution or challenge is not PSF.
** If you proceed in this state, you will just waste products that you cannot actually purchase **.
"If you can make such a solution, I will buy it as soon as it is completed." "If you aim to solve this problem, we will provide you with as much information as you need, and I definitely want you to realize it."
That's the state of Burning Needs, which is the seed of a product that will pay you (in short, sell) when it is actually commercialized.
Especially after 2020, in the recession caused by COVID-19, corporate and personal wallets cannot afford as much as before.
Even if you don't have much money to spend, you need a solution that you definitely want to buy to solve the problem.
It's hard to develop a solution that is unclear if you can focus on Burning Needs.
** Invest resources in building unpurchased solutions that customers don't really need ** It creates a complete waste of time and money, and a loss.
This is the end of "Chapter 2 How to cultivate ideas".
3. How to interview users
3.1 Three points to note when interviewing users
In Chapter 2, we mentioned the importance of dialogue with customers (users). At that time, there are some points to be aware of in order to talk with the user and conduct the interview effectively.
From ● Basics of User Interview (Startup School 2019 # 02), points to note when interacting with customers (users) To summarize.
** 3 notes **
** [1] Ask customers to talk about their lives, time, and how to use various tools, rather than talking about their ideas **
User interviews are not a place to market your product. A good user interview is about getting information from the people you're talking to, and getting data that can help you improve your product, marketing, and your position. It is not meant to be marketed to use your product. The essence of good user interviews is to learn about your life. You need to be specific about the problem you are trying to solve and the experience you may have with it. ● Basics of User Interview (Startup School 2019 # 02)
** [2] Listen to the facts rather than telling the hypothetical idea we are testing **
** [Ogawa] ** We would like to examine our hypothesis in an interview. However, the question "What if ●●?" Does not make sense unless it is a very clear question. The person being interviewed just flutters back, "Maybe it's OO ~", and it's not true and doesn't test the hypothesis. Instead, think that there should be these facts in order for the hypothesis to be substantiated, and ask them to talk about the existence of the facts, their lives, their time, and how to use various tools. Then, from the story (fact) of the customer's real life, we verify the existence of facts that lead to hypothesis proof.
** [3] Listen, not speak **
** [Ogawa] ** Clarify the purpose of the interview and the facts to be heard for the hypothesis to be tested. And since the purpose of the interview isn't to market their ideas, there's little to talk about from here. I just focus on listening to the customer's reality.
3.2 5 Questions to Ask in User Interviews
From ● Basics of User Interviews (Startup School 2019 # 02), we will organize the five questions asked in user interviews. ..
[1] What is the biggest challenge you are trying to solve? [2] Please tell us the last time you faced the problem. [3] What made it difficult? [4] Please let me know if you have tried to solve the problem. [5] What did you dislike about the solutions you have tried to solve the problem?
[1] asks about the problems / issues they faced. And the reason why it was difficult in [3] is the details of its contents.
In order to remind you of the events that happened in the past, specify the last scene you faced in question [2] so that you can remember it clearly.
In [4], we ask about the method actually used as a means to solve the problem. Then, in [5], ask them to move away from a particular scene, ask other solutions they have actually tried to solve the problem, and ask what they didn't like.
** Get factual information about actions when customers actually face problems in [1], [2], and [3] **.
** [4] and [5] will help us to obtain information on competitors and collect information on customer dissatisfaction with competing solutions as we build solutions in the future **.
4. About AI new business planning
4.1 Example of AI-driven app business
Finally, we're going to talk about AI.
First of all, I will introduce the app business that created a new world view by making full use of AI.
[● When AI becomes a product: The rise of AI-based consumer apps (a16z)](https://review.foundx.jp/entry/when-ai-is-the-product-the-rise-of-ai -based-consumer-apps), more objectively organizes the possibilities of AI.
In this article, we first take up "TikTok" and describe the difference in the recommendation algorithm from video sites such as YouTube as follows.
** It's like a TV, but you don't need a remote control-thanks to AI ** TikTok is completely dependent on AI, which makes a big difference. Instead of having the user tap the video thumbnail or click on the channel, the app's AI algorithm determines which video to show to the user. TikTok's full-screen design allows you to clearly see both positive and negative signs of the user for every video. (Positive = Like, Follow, Watch to the end / Negative = Swipe to skip, give a low rating). Even the speed at which the user swipes the video to fly is a relevant sign.
On the other hand, Instagram uses AI as a tool, not a real product. While AI has helped determine which videos are recommended for a person's Instagram exploration feed, the likes and dislikes that thumbnails give to the platform are more ambiguous. If a person doesn't click on the thumbnail, can we really say that he didn't like the video?
How is this different from platforms and products like Facebook news feeds, Netflix, Spotify and YouTube? As we all know, which of the services we have just mentioned is something that users should pay attention to (which one). We use algorithms to recommend news, shows, music, videos). In my opinion, the apps mentioned in this post take a different, more AI-centric approach. For example, TikTok never presents a recommendation list to a user (as Netflix or YouTube does), and never asks the user to express their thoughts. A platform called TikTok makes a complete guess at what users are likely to see. [● When AI becomes a product: The rise of AI-based consumer apps (a16z)](https://review.foundx.jp/entry/when-ai-is-the-product-the-rise-of-ai -based-consumer-apps)
** [Ogawa] ** (I have never used TikTok), TikTok is different from YouTube and Netflix so far, basically users search videos and which one from the recommendation list I don't choose that video.
If you are not interested in the videos that are displayed one after another, swipe to skip them. That mechanism leads to the acquisition of information that "I am not interested in this video" that cannot be acquired on YouTube etc., so it leads to more accurate grasping and drawing of the video that the user likes.
This means that users can spend more time in the app and expand their business.
The above article (● When AI becomes a product: The rise of AI-based consumer apps (a16z) -of-ai-based-consumer-apps)) also uses matching apps and English conversation apps as examples to emphasize the performance of ** AI-based personalization **.
It also touches on the possibility that the AI-based personalization could turn a previously unsuccessful business model into a success.
4.2 Expectations for AI business in Y Combinator
In ● Requests from Y Combinator to startups, as of April 2019, VC Y Combinator (Paul that appears many times in this article) ・ Graham et al. Founded) presents a list of problems and issues in the coming era, and states that it wants to accelerate investment in ventures in these fields.
From AI to biotechnology to transportation and housing.
Let's check about AI.
AI AI has a great impact on society. It's so big that we even created a department for AI within YC. AI has a completely different appearance before and after that, and it feels like it will be a turning point in the history of technology. We are interested in applicants for research in all limited areas (drug discovery, programming assistance, legal counseling, fraud detection, etc.), and especially in areas where AI and robotics intersect (manufacturing, autonomous driving, etc.). I have. ● Requests from Y Combinator to startup
** [Ogawa] ** In the Japanese translation of the article, it says "" All limited areas "(drug discovery, programming support, legal advice, fraud detection, etc.)", but in the original text 「any narrow domain 」 is.
In short, he says he is interested in "utilizing AI by focusing on business domains and target industries."
Even when planning a business that utilizes AI, it is important to focus on the narrow domain. A narrow domain reduces competition and facilitates customer interviews and strategy planning.
This ** narrow domain ** will be one of the keys when starting a new business and planning a new business with AI system.
Another point that is emphasized is the area where "AI and robotics" intersect. It will be difficult to surpass GAFA in a cyber-closed world.
Therefore, while being based on AI, ** a business model in which physical interaction with the real world exists by robots, etc. ** is one of the new business approaches (CPS: Cyber Physical System).
4.3 Impact of machine learning on business
This is a little old article (June 2018), but ● Thinking about machine learning (Benedict Evans, a16z), machine learning is a business Introducing an article by an employee of Andreessen Horowitz (a famous VC specializing in tech) about the impact on.
However, I believe that the implications of machine learning have not yet been established. The meaning of machine learning here is as follows. What does machine learning mean for tech companies and companies in a wider range of industries in the future? How to structurally think about new things that machine learning enables, or the meaning of machine learning for ordinary people. And what are some of the most important problems that machine learning can solve?
In fact, there are plenty of questions about the current development of machine learning that are completely useless. For example: ・ Data is new oil ・ Google and China (or Facebook, Amazon, BAT, etc.) have all the data ・ AI will rob all jobs in the future ・ And of course, topics about AI itself ● Thinking about machine learning (Benedict Evans, a16z)
** [Ogawa] ** I generally agree with this article. Especially ** "The implications of machine learning have not yet been established. What the future implications of machine learning will be for tech companies and companies in a wider range of industries has not been established." **.
Today's deep learning corresponds to the dawn of the Internet, Regarding IT and the Internet, around 1995, when the communication standard was still 2G, I could not imagine the current IT society in 2020.
Also, while the movement of people was restricted by COVID-19, it is hard to imagine in 1995 education by distance lectures and business continuity by telekan.
By the time of the Great Hanshin-Awaji Earthquake in 1995, education was delayed only in the Kansai area, so the voice of educational inequality for university entrance exams was not as loud as this COVID-19, and the idea was to teach by distance lecture this time. I feel like it wasn't long.
In the early days of the Internet, the value of IT, especially the Internet, was still unclear. I think it is similar in that the value that machine learning and deep learning will bring in the next 10 to 20 years is uncertain. (I think Professor Matsuo often tells the story of this nuance.)
I often take the electrical revolution of Industry 4.0 and Society 5.0 as an example. ** The washing machine did not become a moving washboard just because the electric motor technology was born by developing from the steam engine technology ** Talk to.
The washing machine has created a solution that can achieve JTBD (Jobs to be done) of "cleaning clothes" by changing the conventional washing method according to the new technology of motor.
But to be honest, for machine learning and deep learning, the solution for this "washing machine" is unknown.
Various personalization technologies, image processing, etc. are similar to moving the washboard electrically.
However, Youtube can be expected to continue business to some extent even if machine learning recommendations do not work, TikTok is a business model that does not work without the presence of machine learning and deep learning in advance, and may be close to the washing machine of the deep learning era (users will leave if the recommendation does not work at all).
So, like TikTok, to build a new washing machine business in the deep learning era ** Limit in advance to business models that cannot be reliably established without machine learning and deep learning **
The idea may be interesting.
The constraint of imposing restrictions risks narrowing your thinking, but it also has the potential to lead to new ideas.
- However, as mentioned at the beginning of the article, a business starting from an idea has a low probability of acquiring customers, so after coming up with an idea within the limits, you should return to the problem / issue to be solved again. There is
5. Create MVP
5.1 MVP creation is a process of new business proposal
I will explain the meaning of MVP while taking up ● MVP is a process, not a product.
If you have read this article so far, I think that you can understand that the process of searching for problems and finding solutions up to PSF (Problem Solution Fit) in a new business is a process that repeats until it succeeds.
However, for large companies, especially executives, IT has always been a waterfall model (SIer has a clear picture of the finished product, and SIer has clearly documented the plans and products to achieve it. I'm used to the type (implemented by our partner companies) and have succeeded in doing so.
Therefore, new businesses that utilize AI tend to have the image shown in the figure below.
● MVP is a process, not a product
In other words, the image is that the finished product is clear, and progress is made in a straight line every six months or quarters.
However, in reality, the image of modern business, especially the completion of IT systems that utilize AI, is as follows.
● MVP is a process, not a product
I'm going back and forth to the Proof of concept, but think of it as a trial and error process to achieve a PSF.
In the world of trial and error, the person who finds the mistake first is the winner. Some people call this philosophy "Fail Fast." On TripAdvisor, we used to call it "speed win." Eric Ries called it lean, and Kent Beck and other programmers called it agile. Whatever you call it, the core is to get product feedback from real users as soon as possible to find out which assumptions are wrong. Whether you're making a product, writing code, or thinking of marketing ideas, you should always ask yourself two questions: ・ What is the highest risk assumption? • What is the smallest experiment that can be done to test that assumption? ● MVP is a process, not a product
Thus, ** MVPs are not finished products, outcomes, or outcomes, but products and outputs of intermediate processes **.
It is important not to be fascinated by MVP.
It's no wonder that you're attached to it because you made it, but creating an MVP is a process of improvement, and its prototype should be constantly changing.
MVP is an intermediate product of the process of breaking down hypotheses one by one toward business realization. It becomes consciousness.
** With AI, you can do amazing things, you can do things that were not possible before **, and I feel that the image has taken root in society.
It's not a mistake in itself, but is it easy to plan a new business using that AI? That is not the case.
It is necessary to understand and systematize the fact that it is an agile type by trial and error, not a waterfall type like the conventional IT system.
Institutionalization is especially targeted at new business planning systems in companies, but it is a review of various systems and stage gate methods so that agile type new business planning and formulation becomes possible.
Creating a new business utilizing cutting-edge technologies such as AI without changing the structure of those in-house systems is a failure to create a new business due to reduced motivation of field employees.
As a result, promising employees will change jobs to a company that has such a system in place.
** MVP creation is a process of new business proposal **, unless a management who can understand the meaning of this word creates a mechanism for that, new business creation in a company, and true collaboration and value creation with a venture will be It's difficult (I feel).
5.2 Create MVPs from suggested slides
In the words of Steve Jobs ** "In many cases, people don't know what they want until they show it to them." ** There is a famous word. [● Steve Jobs Quotations (Japanese, English)](https://estorypost.com/%E5%90%8D%E8%A8%80%E3%83%BB%E6%A0%BC%E8% A8% 80 /% E3% 82% B9% E3% 83% 86% E3% 82% A3% E3% 83% BC% E3% 83% 96% E3% 83% BB% E3% 82% B8% E3% 83 % A7% E3% 83% 96% E3% 82% BA% E5% 90% 8D% E8% A8% 80 /)
It is faster to create a demo prototype that is the core of the product or function and show what works, rather than making a suggestion slide with PowerPoint asking "How about making something like this?"
People respond better to seeing what actually moves than to PowerPoint.
In order to realize this team activity of "trying to make a prototype that moves immediately" ** The development system must not be outsourced or outsourced. In-house production is a prerequisite **.
There are people on the team who can write the code properly, and they also need people who can make prototypes quickly. (However, it is not necessary to guarantee product-level security, and it is sufficient to create a prototype that can verify the idea.)
For example, in the case of Airbnb, some of the founders who studied design at art schools, Brian Chesky and Joe Gebbia, may have the image of bringing design thinking to the business and brilliantly founding it. As a matter of fact, Airbnb now is on top of a very muddy stickiness and an activity that never gives up.
And it is big that Nathan Brecharck was there. He has been an IT professional since he was a student and has been entrusted with system development to earn money (huge).
Airbnb wouldn't have been possible without Brecharck, who could quickly turn Brian's business hypothesis into an IT system. (In the current situation, it may be difficult to expect Airbnb to grow for the foreseeable future due to the impact of COVID-19).
5.3 MVP can be a toy
The prototype that becomes the MVP is good at the toy level. Rather, it is better not to make it strangely.
● Why is it important for startups to start with "toys"? (Aaron Harris).
** Expectations ** Let's say you give people a tool and say that this solves a serious problem perfectly. The user will then get angry at any small flaw in the tool. On the other hand, let's say you give someone a toy and say, "Look! I made it. It's interesting. You can do this." You expect a positive reaction. That's because it's much easier to exceed low expectations than to exceed high expectations. You can say that you have increased your chances of getting materially happy users. ● Why is it important for startups to start with "toys"? (Aaron Harris)
In this way, it is better to show the demo with the attitude of "It is not a product yet, but is there any sign that it will help you or our customers solve the Burning Needs?" ,Is good.
Also,
Companies are about making money and doing business with customers in mind. This is both very serious and risky and scary. Toys, on the other hand, are for playing and trying new things. This is not serious at all. ・ ・ ・ First of all, you don't want to try ideas that don't clearly meet the goal of ultimately building a big company. This means that the eyes of those who make serious things will go to sales immediately. They avoid risk and, as a result, innovation. Companies that are based on new technologies must take advantage of extraordinary ideas. These are ideas that large companies cannot pass because they do not meet certain criteria. ● Why is it important for startups to start with "toys"? (Aaron Harris)
The larger the company, the harder it is to allow employees to spend time making toys.
It's hard to allow time to play with the unseen ideas that "immediately commercialize, aim for black, and scale."
To avoid that, like Google's 20% rule, would it really allow time to try something ...
That said, 20% is one day a week. An ordinary company wouldn't allow you to like one of the five days. .. ..
However, even when planning a new business in a large company, there are actually many solutions around this point.
And for AI engineers who want to start a business, securing time to make toys that large companies cannot do may actually be a shortcut to success. (Facebook also started with a toy closed at Harvard University as a tool to promote interaction between students by making a simple paper version of the university's official student introduction into an online version of SNS)
This is the end of "Chapter 5 Creating MVP".
Above all, I would like to convey that it is only a hypothesis that the problems / issues that we have decided to focus on in entrepreneurship and new business planning, and their solutions, will also work.
It is indispensable to verify whether this hypothesis is true, to rebuild the hypothesis, and to build a high-speed MVP and interact with customers.
(The flow of my AI system construction is as follows)
afterwards,
6. How to make a pitch
6.1 How to make a pitch
Speaking of pitch, it is a presentation that a venture company gives to VCs to convey their attractiveness and get them to invest.
However, it will be necessary to give a presentation to upper management not only in starting a business but also in planning a new business of the intra-planner type.
I mentioned the importance of MVP earlier, but that doesn't mean that no presentation is needed.
Based on what we learned from MVP and the results, we need to pitch to VCs and the upper management of our company.
For the pitch slide configuration, ● Airbnb's initial pitch material is according to YC's 3-minute pitch template But it is detailed.
Here is the actual 2008 Airbnb pitch.
The basic composition of the pitch is
[0] Title (Company name & service provided in one word) [1] Challenges [2] Solution [3] Market size (estimated by other companies in the same industry and Fermi) [4] Traction (current degree of inquiries) [5] Unique insights [6] Business model (showing how easy it is to make money) [7] Team (shows what abilities and achievements people have) [8] Closing (specify the action you want to request afterwards to those who listened to this presentation)
It will be.
In addition, if you are planning a new business within your company, you will need to explain ** synergies with your company's strengths **.
** Completely new business in the excursion ** that is not related to your company's strengths can be imitated by other companies, so it is difficult to succeed after that.
On the other hand, if it is a new business that leverages your strengths, it is difficult for other companies to imitate, and you can use your existing resources, so the probability of success is higher than a complete excursion.
7. How to create a team for a new AI business
7.1 Leader, responsible person
The situation where new business managers or entrepreneurs cannot program or AI is an era that is no longer talked about. (Even if it's not an AI-centric business)
The state in which a new business manager or entrepreneur cannot program or AI is an image that the CEO cannot read or understand the three financial tables.
However, this is also a material that you can rest assured.
The CEO does not have a separate tax accountant or accountant qualification. I am working on finance and financial analysis with the support of a tax accountant, etc., and I am not doing it alone.
Similarly, for programming and AI, the person in charge and the leader do not have to be able to handle it alone, as long as the work can be done with the support of IT and AI experts.
For team leaders and managers, such as new business managers and entrepreneurs in the future It is important to recognize that programming & AI knowledge is required as much as accounting knowledge.
Readers are also advised to read the following articles.
● Startup Playbook An introductory guide to anyone interested in startups, written by Y Combinator's Sam Altman. Whether you're starting a business or planning a new business, this is something you definitely want to read.
Also from Sam Altman, ● Advice for Startups (Sam Altman) However, various advice will be helpful in looking back on our team.
7.2 How to collect team members 1
There are two important points in gathering team members.
The first is to be mission-centric.
It is important to set a clear and easy-to-understand vision for your venture or new business, and to build a mission on which members can rely.
In the words of Peter Thiel,
"Why would a higher-paying, higher-ranking person choose your company as the 20th engineer, whether at Google or another company?" Various benefits can also be offered by other companies. The answer is "There is no correct answer because it depends on each company, but the" mission "and" team "are the keys." ● "I want the 20th employee to join the company for this reason." Peter Thiel
There is something like that.
The 20th employee is not a referral, but a nuance of a new member who comes in as a red stranger.
As of 2020, the salary level of AI human resources is still high.
However, the person who decides where to change jobs based on the amount of salary will leave the company in terms of money.
Rather, it is better to have members who can communicate and sympathize with their mission and mission. (Of course, at a minimum, the salary level should be the same as the industry average, and the treatment should be such that the new members can maintain their lives.)
Sam Altman said:
** Be mission-oriented ** Most of the top companies are mission-oriented. If you don't think you have an important mission, it's difficult to combine the power that big companies need. It's also very difficult without a great founding spirit. Mission-oriented ideas also have the advantage that you can devote yourself to them. It takes about 10 years to create a great startup. If you love and believe what you build, you are likely to give up on the way. As far as I know, there is no way to overcome the suffering of a startup other than believing that "this mission is really important". Many founders, especially student founders, believe that their startup will get off to a good start in just a few years and then work on what they really like. But most do not. Good startups usually take 10 years to get going. ● Startup Ideas, Products, Teams, Execution Ability Part 1 (Startup School 2014 # 01, Sam Altman, Dustin Moskovitz)
7.3 How to collect team members 2
Another important thing in gathering team members is to be able to answer the question "why now?".
This is important not only for gathering members but also for getting investment from VCs,
"No one has focused on the problems and issues that we are focusing on in the past, and we are the first of humankind!" That is impossible.
So why could people in the past not be able to solve the problem / problem, but now and by themselves? , I need to be able to answer this question.
Leaders and teams who can't answer that are just dreamers, not entrepreneurs who can live together.
at the end
It became an article that wrote a lot of things that seemed to be great.
When asked "How are you?", I'm still a chick.
I personally feel that it is a problem / issue, and there are two things that I would like to build a solution to solve the problem.
The first is to establish a method for building a deep learning-centric business model, which is symbolized by the "washing machine" of the deep learning era.
The second is Even if a representative of the Japanese manufacturing industry and a company that can be said to be the representative of Japan's deep learning form a capital tie-up and work hard for several years (It may be very rude to everyone involved) I feel that the stage level of the company is changing, and that no rugged business results have been produced.
How can the manufacturing industry, which is a Japanese specialty, utilize deep learning and AI to produce rugged results that will change the stage level of a company? I would like to solve this mystery and make it a solution.
These two are my personal feelings, and I enjoy working as a team because of the overlap between those feelings and what I should do at the company I am currently working for.
This is a very long sentence, but thank you very much for reading this article.
[Critical responsibility] Information Services International-Dentsu (ISID) AI Transformation Center Development Group Yutaro Ogawa [Disclaimer] The content of this article itself is the opinion / transmission of the author, not the official opinion of the company to which the author belongs. [Remarks] The AI department development team that I lead is looking for members to praise. If you are interested, please click here
References
● Announcement of data scientist missions, skill sets, definitions, and skill levels Summary of advice from Paul Graham to startups ● Paul Graham "Six Principles of Creating New Things" ● The essence of startups learned from Peter Thiel's words ● How to get ideas for startups (Original author: Paul Graham) ● Solving customer Burning needs ● Basics of User Interview (Startup School 2019 # 02) ● Requests from Y Combinator to startup [● Steve Jobs Quotations (Japanese, English)](https://estorypost.com/%E5%90%8D%E8%A8%80%E3%83%BB%E6%A0%BC%E8% A8% 80 /% E3% 82% B9% E3% 83% 86% E3% 82% A3% E3% 83% BC% E3% 83% 96% E3% 83% BB% E3% 82% B8% E3% 83 % A7% E3% 83% 96% E3% 82% BA% E5% 90% 8D% E8% A8% 80 /) ● Startup Playbook ● Advice for Startups (Sam Altman) ● Why is it important for startups to start with "toys"? (Aaron Harris) ● Startup Ideas, Products, Teams, Execution Ability Part 1 (Startup School 2014 # 01, Sam Altman, Dustin Moskovitz) ● Airbnb's initial pitch material is according to YC's 3-minute pitch template ● Airbnb Pitch Deck From 2008 ● "I want the 20th employee to join the company for this reason." Peter Thiel [● When AI becomes a product: The rise of AI-based consumer apps (a16z)](https://review.foundx.jp/entry/when-ai-is-the-product-the-rise-of-ai -based-consumer-apps) ● Thinking about machine learning (Benedict Evans, a16z) ● MVP is a process, not a product
Recommended Posts