Why and how to build a tech company?

Starting a tech company is a terrible way to make money, and I don't think a lot of people talk about this. Tech startups are riddled with uncertainty and the risk-reward is barely in your favour. Lets say, you have a brilliant startup idea today and some VC were to today give you $100k for 20% of your company. That is a lot of money and now you need to plan how to make the best use of it. You have now essentially started a clock on yourself where you need to build a decent product as well as get to product market fit before you run out of money. You have a very short runway where you can prove to your investors that this can potentially work and I need you to give me more money to extend my runway so that I can take this thing off. While this does sound reasonable, and many companies have been built this way, it is also necessary to take a broader perspective to help see some problems in this. The larger perspective is that at age 25, I want to find out what is the best way for me to build a company that can generate most wealth in the long term, period. I have to be objective in the way I think about this because my goal is not to work in tech but to make the most wealth and to do that I have to compare tech with all other forms of businesses out there. You could say that this is an emotionless way to think about starting a business but I think this is somewhat needed too. To give you a disclaimer, I have been working in AI for 2 years and have been excited about deep learning and generative AI long before chatgpt was released. Yet, I force myself to be blunt and objective so I can identify problems in every industry and not let my past experiences in a particular field limit my future opportunities. Thus, giving me an unbiased overview of all forms of businesses. So lets talk about the tech industry, unlike the norm, I am not going to start with the positives of starting a tech company but would first want to shine some light on some issues I see. ### 1. _No sales till product market fit_ Continuing from the above example, you have a great idea for a tech startup and someone has agreed to give you $100k for 20%. You, undoubtedly give it your best and more to work on the company. Despite that it will take at least 1-2 years where you will pivot your product and mission multiple times. Once you get hit with the reality of customer preferences, you will likely have to change your product all over again and get back in the race to find the small segment in your industry where a customer exists and spend your resources to develop a product around it. So you lose 1-2 years of borrowed time and money where you are not seeing any sales. These 1-2 years are very uncertain and there is no telling if your company will be alive by the end and where it will end up. From the investors POV, they are making enough bets on multiple companies that the one that will succeed will overshadow all the other failures. For you on the other hand, this uncertainty may not the best way to spend your $100k. Compare this with other businesses like: opening a food cart with some of the tastiest food from your country, design print and sales of t shirts for other businesses, manufacture and distribution of low scale products like cutlery, stationary, chemical dyes etc. The benefits with these so called unsexy businesses are that you start getting customers and sales from early on. The revenue coming in can then help you expand your business and give trust with your investors. This may be an oversimplification for a tech startup but by comparing it with other businesses we at least identified an issue and will dicuss later on how can we circumvent this. ### 2. _Strong engineering biases need to be deleted_ Coming from a research background, I had the opportunity to work with some of the sharpest robotics and AI PhDs in the country. What I noticed was that engineers who have spent years learning about a particular technology, fall in love with the technology rather than the problem itself. The excitement of being on the edge of any subject is enticing but can also blind you from other fields and possibilities. To give you an example, an engineer who has spent years learning about machine learning and is surrounded by CS majors starts to see everything as an ML problem. This ends up shortening your horizon, making you forget about the end goal and develop a very unidimensional thinking. You may notice similar experiences with product designers who can become so focused on design principles and aesthetics that they lose touch of the practicalities of cost of production or a customer's purchasing ability. This expert bias can be corrected if you encourage having a diverse team with healthy team dynamics where everyone has a fair appreciation of design, business and technology. ## How would I go about building a company? For all the abovementioned problems I see in opening a tech company, there are certain benefits which you would not get anywhere. Keep in mind that we are talking about companies which can be started with just $100k - **Scaling a tech company is much easier.** You will not require large capital for manufacturing plants or warehouses or logistics which means barrier to entry is low in terms of capital. - **Target audience is very large.** Software products can cater to any part of the world as compared to other traditional businesses like FMCG products / hospitality / services etc which are limited by logistics or location. - **Very less competition.** There will be many other companies manufacturing and selling low scale products but tech evolves so fast you, that very few companies will be providing similar products like you. - **Room to innovate.** Unlike traditional businesses a tech company can differentiate itself by improving its product constantly and continue to capture new segments. - **A self improving culture.** A key facet of a tech company is the constant improvement cycle. Tech companies hire permanent employees who will be dealing with new kinds of problems because the company keeps evolving. Engineers and scientists are geared to iteratively keep improving their products and this culture of analytical self improvement is unlike any other industry and pushes your company to keep improving the products, expanding your target audience and grow rapidly. To open a tech firm, the first thing I would focus on is to find a reliable revenue stream and that is especially true in 2023. You want to choose a growing industry and make your first sale as soon as possible. An industry which I think has great potential to grow right now is generative AI for design and I want to build a company in this space. We want to avoid the pitfalls mentioned above of tech companies and may need to take a rather unconventional approach to solve this. Exploring the industry and building a product in this space may lead us into an uncertain space searching for customers and fighting hard to stay alive. This is where I would like to hack this model by borrowing some advantages of non-tech businesses. However silly as it may sound it would actually make more sense to start a tech company by starting a simple design business. This could be something very basic like design and sale of tote bags / tshirts / paper bags to other businesses. The first goal is very straightforward and focused only on making a $500 sale in a design field of your choice. This will force you to take baby steps and have a very small clientele and ensuring that they are kept happy. Once you generate a revenue stream you can then think of building tools which can make your process faster and better. I chose the design market because it is massive, easy to enter, and gives you a good revenue stream right off the bat. As I grow, I can build generative AI tools for the design process for t shirts, tote bags, purses, shoes, posters, websites, branding etc. We will explore this in detail further. This two-fold technique of building a tech company will be self sustainable and the feedback loop from the design and sales to the AI tool building team will make my products more user focused. I understand this is a rather hard pill to swallow but let me tell you that there is no right way to build a business. It just makes more financial sense to me to first develop a revenue model and understand any design market before spending on R&D to build AI tools for the business. Take Amazon for example, when they started out, they used the internet to find customers who are looking to buy books. It was a simple idea to start off and make money. Jeff did not start off thinking how to use AI to build a logistical and retail empire. He started small and built the tech giant it is by not having any external biases towards technology and by great business execution. You will see such anecdotes everywhere, and they make sense too. To build a business I want to do the same, start in a small segment and make use of technology to keep improving but not make technology the main goal itself. ### My Design Company I want to describe in detail as to what I mean by having a design company, because this is a very broad term and a very broad market. Lets look at the potential design markets. 1. Furniture Design - Home and Living products, 2. Product Design - This can range from vehicles, toys, electronics etc. 3. Jewelry Design 4. Interior Design - Houses, offices, architecture etc. 5. Grahpic Design - Websites, Social media posts, Logos 6. Apparel - Clothing, Purses, Bags, Shoes Design is everywhere around us and this list can become much larger but the problem with the some of the above industries are that it requires a lot of capital to enter these markets. Coming back to the $100k example for me, I would like to focus on a smaller segment which would be rather easy to enter for me and not require large amounts of capital. Instead, it would be better to generate a revenue stream in a smaller easy to achieve market and keep growing continously. Over time, once the company grows into something, we can make use of our team, capital and knowledge to enter other larger design markets. I would like to present a few ways I can get a foot in the door. ### __Logo Design Business:__ Coming back to the $100k example for me, the small segment where I would like to focus on right now is small business graphic designs. This includes: Business logos, stationary, packaging design, website design, invitation cards, brochures etc. Living in NYC, my potential customers of small business owners are all around me. Be it restaurants, grocery shops, thrift stores, barbershops, they are everyhwere and are a good market to tap into. To understand this space better I spoke with a self taught business owner who caters to the same exact market and chalked out a business plan for myself to enter this market. You want to first focus on your design business and generate some clients before moving onto building AI tools for your company. To do this I would: 1. Learn graphic design 2. Become a traditional freelance graphic designer using photoshop, illustrator etc for any business needs around me. This would generate a revenue stream for me as well as help me understand my future customer base better. 3. Use existing AI tools to improve my design processes so I can cater to a larger number of businesses. 4. Begin building AI tools using diffusion, LLMs, generative AI etc. that would help other designers like myself. 5. Enter another design industry and repeat steps 1-4. ### __Tote Bag Business (Focused for US market):__ Another way to start building an AI design company would be to manufacture and sell tote bags for small shops and businesses. Why tote bags? * As compared to other apparel businesses, tote bags do not have any sizing issues so are easy to manufacture, and changing the design is fairly simple. You provide your graphic/image file to a direct to film print (DTF print), then use the film print to transfer the design onto the tote bag. * Their major design work happens in the graphic design on top of the tote bag, with no other changes in material or color, making it the perfect play for generative AI to make the design process faster and with creating better looking designs. This can help you provide your customers with changing designs every other month. With a little product development we can use the existing logos of the business and create an image / design file that uses the logo of our customer and creates a graphic design suitable for a tote bag around it. * You do not need to sell to retail customers so it is B2B rather than B2C which is an easy market to crack and has recurring customers. Our target audience will again be small business / shop owners in NYC for the moment. * Manufacture and printing costs of tote bags in India are very cheap and I am in touch with some people who can do the same and export the bags for lower than $2. How would I go about starting a revenue stream in this segment: 1. Learn graphic design and go door-to-door to sell tote bags to shops in NYC. 2. Make your own graphic desigs and use existing AI tools like diffusion to make design samples for businesses who are looking to buy tote bags. 3. Generate a good revenue stream and customer base in this segment. 4. Build better AI tools to improve the process. 5. Enter another industry and repeat steps 1-4 ### Future business in Interior Design (Focused for Indian market) I believe there is a lot of space for building AI tools that can help interior designer / home decor space. To understand this space better I spoke with interior designers, home builders, wardrobe manufacturers, furniture stores etc. to understand all their processes and figure out ways in which we can build tools around this space. #### Customer Journey of a home buyer looking to furnish their house in India: After you buy a house you will hire an interior designer, and sit with the designer to decide the basic house layout and plan a total budget on houw much you would like to spend on the house. The home buyer and designer will prepare a mood board by taking images from pinterest and google to decide the theme or color choices for the house. You will scroll through hundreds of images and then select a handful of them as reference images for the mood board. This will give you a very rough idea of what you would want the house to look like. After this process the main grunt work begins. The home buyer will need to hire a qualified contractor who will have the manpower to build / install all the furnishings. Then you will have to start selecting all your furnishings, this includes flooring, tiles, paint colors, couch, type of wardrobe, type of wardrobe shutters, dining table, bathroom tiles, bathroom fittings etc. For each item, you have to find a good store from your own network or go to one of the stores recommended by the designer. The home buyer will have to go to multiple dedicated stores and browse through hundreds of options for each item. That takes a lot of time and effort because of the sheer number of items to buy, options available and the number of stores for those options. The buyer will not be able to allocate a budget for every item in the house, they just have a rough estimate of the entire spend they would like to do, which makes it almost impossible to stay under budget and in every I case I studied, the buyer ends up spending 15-20% above what they had initially thought of. During the selection of all the products, the buyer needs to keep travelling back and forth from the site and to various markets to shop for every item individually making it a very time consuming and tedious process. Even after all this effort there is no fixed idea of what the house will look like and it will keep changing until the house is actually made. You will not get an entire visulization of what the house will actually look like and what you see is not what you get. Only once the selection of all products is done, will the designer sit and prepare a rough render of the house. These renders take hours to develop and any change in the render of the house takes too long to incorporate so these renders barely give a clear picture to the home buyer. The render will be of the correct dimensions, will include the right color for the floors, paint, size of the dining table etc but it still is far from the real look of the room. The render will never include any art piece / couch or any item which does not come readymade into the interior designer software, until you manually make changes to every product. It will include some readymade options of the right color for a dining table or basin or even a door handle but these options are nothing as compared to the thousands of options you will actually see in the market. Consider the purchase of something as basic as the shutter for a wardrobe. The customer has to go to a shop dedicated for wardrobes, where the shop owner will show you 2-3 complete wardrobes on display, a handful of samples of the different kind of materials you can buy, and the owner will show you a brochure with small samples of the material in 50+ color options for the outer finish of the wardrobe. The amount of options you will get are endless, in just the material there is PU, laminated, acro glass, fluted glass, back-painted and more; once the material is selected you have to also select the color and if thats not enough then you also have to select any particular design or engraving you would like on top. It would not be wrong to say that there are an infinite number of shutter options. It is difficult for home buyers or designers to choose from these options because they do not have a clear visulization of what the samples they choose will look like on a wardrobe and how the wardrobe will look like amongst the rest of the room. Once you choose all your options, the designer will then put your selcted options on a render to get an approval from you. You cannot escape making these renders but because of its dissimilarity from the actual house and the large cost in making them look real, the renders end up being unworkable. The home buyer thus has to end up pouring 1000s of dollars based only on imagination and the designer cannot serve its customers well. The customer journey is very similar when someone is designing a new restaurant / office or any other real estate space, so a similar experience goes on in that process as well. #### The state of the furniture and real estate market in India Apart from the small presence of online furniture shopping or megastores like IKEA, furniture or home decor items in India are majorly done in an offline store. There are many furniture / home decor manufacturers both small and large that make amazingly well designed products at very reasonable prices. The large manufacturers like Kohler for bathroom fittings or Aristo for wardrobes sell through dealerships or traders of their products. The smaller manufacturers sell their products to larger brands who have a brand name and stores in the city. Online stores like Furlenco or Pepperfry exist too and have created a great brand name for themselves, and they too have opened multiple brick and mortar stores in urban areas. They do not manufacture themselves but focus mostly on connecting the customers with the sellers through their website and managing the large and complex scale of their inventory and logistics. <https://yourstory.com/2019/10/behind-scenes-pepperfry-ecommerce-startup-online-furniture> There is a large real estate boom across the board in India. We are seeing exponential growth in residential, office space and restaurants in metros as well as Tier 2/3 cities (smaller cities). Tapping this large market by improving the design and shopping experience for buyers could add large amounts of value. #### An improved way of designing real estate (which includes houses, restarants and offices): It seems almost clear that we need a better way to furnish real estate. I want to build software for interior designers through which you can design real estate using actual products from the market so that the property buyers can see exactly what they are buying, all online. It will not just be a software tool for interior designers but also an ecommerce website for all furniture / home decor markets to list their products. It will be a 3d rendering software, where you can choose every item that actually exists in markets around you. If you want to choose the flooring for your house, you see a high quality render of all the options available in the market along with its prices, vendors, and delivery options. You then choose one of the options and your render will now have the flooring you have selected. When you want to choose the wardrobes, you will see a lot more options then you see in the physical shop, along with recommended colors, materials and designs that you would want. You can then order them from the nearest vendor who will come and install it for you. Same thing will be done for flooring, or bathroom fittings too. For other loose furniture items, you can again browse through multiple options that currently exist in the store and the home buyer can select what they want and visualize all of it in an high quality accurate 3d render. #### Why now? Achieving these features has not been a software limitation but an AI limitation. With newer generative AI tools, it is possible to make renders that quickly and easily. Many features are possible using gen AI like: We can now use the image of a loose furniture item and then place it inside a 3d render. We can convert a rough sketch of a furniture item and output an image of that item in whatever material we would like, then ask a vendor to make our product like that. We can use inspiration photos taken from Pinterest and use neural style transfer to apply the design themes on our current furniture. The possibilities have really only been limited by AI. Generative AI is only going to get better and more multimodal in the coming years making the time ripe for such a venture. #### So what? Having a 3d rendering interior design software which also acts as an ecommerce platform can totally change the game. The end user that is your home owner will be able to visualize their entire house right from the start and can get a much more detailed quote of what the charges are going to be. Software can be used across the world, and with this enhanced form of furniture listing can boost furniture export everywhere by connecting the buyers to much more reasonable and well designed products in other countries. From a business point of view, our venture earns from a software subscription to interior designers as well as from the advertising costs for listing on our ecommerce platform. #### Where do we go from here? You want to build an MVP first and you can choose many different ways to do this but for now lets start off small in bathrooms. Since this is just a prototype, we want to first build a 3d rendering tool which can be done using Unreal Engine 5. This will allow the user to change the dimensions of the bathroom. Then we want to demonstrate the strength of gen AI. First we look into bathroom tiles. It can be difficult to get the cadd of a tile so you want to prove that using only a high quality image of a tile, you can use the image and add it to any floor / wall of the bathroom without having the manually do it through CADD. We will also show the listing price / vendor and other details of the tile that the user has selected in our MVP. Next we look into bathroom fittings. In this case, you do have access to cadd files from the manufacturers, so you want to list out all the options for the sink, WC, shower fittings along with its details and give the user the option to place the items wherever they like. Adding maybe one or two more such things with generative AI can help demonstrate the power of this software. Then of course we test it with interior designers and take their input on it to make some changes. Another MVP option could be to focus mainly on wardrobes. The number of ways manufacturers can customize your wardrobe is infinite, thus making it the perfect play for gen AI. An off the shelf text to image diffusion model like Dall-E 2 itself would be a great place to start and of great help to the end customer looking to design her house. You can also add a neural style transfer option in the MVP which can take in inspiration images and change the image accordingly. Here we do not even need to get into converting the image to a 3d render since for wardrobes, even the image of the design that gives both the customer and the manufacturer what they need. You want to make a first sale asap and if I were to build an interior design software with a marketplace with all of the features listed above I am sure to fail because it will take too long to make it which will bleed me to death before I even onboard any users. Instead of making an all purpose software I want to focus on a smaller segment first and build a minimum sellable product there. I would like to focus on wardrobes first. The number of ways manufacturers can customize your wardrobe is infinite, thus making it the perfect play for gen AI. You want to create a simple design tool where customers can input text or inspiration images from pinterest and our model will create an image of the wardrobe. An off the shelf text to image diffusion model like Dall-E 2 itself would be a great place to start and of great help to the end customer looking to design her wardrobe. For the inspiration images we can look into neural style transfer option in the product which can take in inspiration images and change the image accordingly. Here we do not even need to get into converting the image to a 3d render since for wardrobes, even the image of the design gives both the customer and the manufacturer what they need to go ahead. You have the ai generated wardrobe designs but you dont sell it as a software to the wardrobe manufacturer company on their ipad. You sell it as an entire system almost making it a tourist attraction for customers to their store to increase footfall. There will be a projector screen the legnth of an actual wardrobe and a dais in front of it with a fixed iPad. There customers can type in things, connect to pinterest and add their inspiration photos, make and suggest changes all in a nice user friendly way which they will be able to see live on a big screen. Even using ai art is kind of a first time thing for many users so this will increase footfall to the store and you tell the store owner to not make this system portable so the wardrobe manufacturers customers will have to come visit the store for all these design work. If it becomes a tourist attraction every other wardrobe manufacturer would want to start buying it creating a good demand for our product. For pricing of course you charge a good amount up front for the entire system but you also charge per image where you want your major profits to come in from. That will pretty much give you continuous income from one software itself. ## Conclusion I have listed down 3 well thought and researched business ideas in the field of design in an increasing order of preference for me. In all 3 scenarios I have spoken with customers, manufacturers and designers working in the field to find any small space for me to operate and build a business around. However, we need to remember that ideas are the cheapest commodities on this planet and they evolve and improve only when you start working on them.