After several artificial intelligence projects with startups, I realized that one aspect of artificial intelligence disruption is still relatively uncensored:The correct business model for artificial intelligence companies.

Is the product sold as a service or as a new feature category for users? The mention of income type is an important part of any AI project.

Due to the many technical elements associated with AI, I have noticed that existing traditional business models are not always applicable.

business model:The plan for successful business operations, identifying sources of income, prospective customer base, product and financing details.

One might think that, like their previous cloud/SaaS companies, artificial intelligence startups have a common business model. However, I noticed that it is difficult to apply the SaaS model to AI startups.

In fact, depending on the nature of the AI ​​solution, you will always need data, a lot of raw computing power and algorithms. Customer digestion is inherently more complex than cloud startups, and it requires different things, so technology must also be sold in different ways.

Before entering the artificial intelligence business model, I think it is very important to show the artificial intelligence landscape based on my experience.

AI Landscape can be divided into two parts:

  1. infrastructure:These companies run on the back end and provide computing services to other companies. The business model they follow is usually based on API calls. A good example is IBM Watson, which provides sentiment analysis through its Bluemix platform.NLPAnd entity identification. IBM charges users for API calls.
  2. application:These can be found in the B2B and B2C fields. However, inB2B fieldYou can see the important activity, that is, the company provides a SaaS-based subscription service. These companies typically develop applications for specific use cases defined by customers. Because I am very familiar with it, so I can talk more about this. I realized that after many proofs of concept, some of them were eventually acquired by important companies. Basically, if the things that these AI companies do are considered strategic by the customer, then some big customers will want to get a startup instead of renting the technology.

Let us now focus on AI startups:

There are basically three types of AI startups:

Going back to business model, I observed that two of them started to work well, others existed, but I haven't introduced them yet.

1.

In this model, the new AI solution will increase the effectiveness of the current workflow.

Due to intensive deployment, the sales cycle is very long. Therefore, every transaction must be large in order to maintain entrepreneurial vitality. In addition to important development costs, a large amount of operating costs are required.Typically, the company will charge you for the development of a custom solution and then pay you monthly operating costs and operational support/training.

2.

Business models are more or less similar to SaaS models. It involves AI solutions that can interact on other systems, such as CRM / ERP systems. AI accesses data flowing through these systems to drive business improvement.In this business model, the company will charge you on a monthly basis. According to my experience, it is easier to make it suitable for NLP projects (chat bots, etc.)

Typically, such solutions are deployed quickly, so the sales cycle is fast and has a reliable ROI.However, this business model is also very fragile.If the AI ​​solution does not prove itself "essential," it will be vulnerable to budget cuts.

It depends on the data

Your next business model will be highly dependent on the data you can use.

Obviously, your ability to leverage data will affect your business model. Given the ability to replicate data (many projects start using fake data in the PoC phase), they are not scarce in nature, and the value of the data is generally low and tends to be low overall in most industries.

As a startup, you don't need all the data in the world, just the data you need to solve the specific problem you are looking for (as long as you define it precisely).

It also depends on the nature of your project, let us imagine that you want to build an AI-driven drone. You need to integrate hardware costs and other product-related costs into your business model.

Building a AI for the first time is still more difficult and time consuming than a regular SaaS startup. In fact, data collection and AI training take a lot of time. Deploying TensorFlow still requires rare expertise. For all of these reasons, the SaaS model can be complex for AI startups.

Income share

I have worked with several companies that cannot afford to build AI solutions from scratch. Therefore, they decided to work with AI development startups that focus on custom solutions. Through the revenue sharing business model, both companies managed to find interest. They all agree to build a PoC, and if it works, they share the benefits. The only tricky question is who will provide the data needed to build a PoC. In my experience, neither company is willing to spend too much time collecting data.

It is suitable for software companies that are trying to improve existing solutions without spending too much development cost.I can imagine a future where the AI ​​development team can get multiple revenue sharing contracts to get other companies to do all the business/marketing work.

The only downside is that another partner wants to add some non-competitive terms. From a contractual point of view, both sides are a bit complicated. In fact, both have something to lose. The first one became heavily dependent on AI, and AI could not sell this same solution to another competitor.

I noticed that most artificial intelligence development startups have some paid pilots running, and there are some early indications that once these pilots are put into production, customers will be prepared to pay a lot of money.

Therefore, they are always interested in ensuring new partnerships that can generate new income. In particular, when they have developed similar algorithms to answer your business questions.

SaaS and AI

It may be tempting for artificial intelligence startups to choose the SaaS model, but for various reasons, this choice may be risky...

Price

Pilot

The nature of the AI ​​solution creates a situation in which training data that is usually specially annotated is mandatory, as well as many different data sources. Therefore, you cannot limit AI testing to a small number of users. In fact, it reduces the ability of the solution to adapt to customer needs.The more people use AI, the faster they learn..

For this reason, I saw several startups using project managers to help companies understand and train AI processes. Obviously, this extra resource requires an expected cost in the business model.

Evaluation

Within a few days of deploying a solution, the value proposition of a traditional SaaS solution is usually obvious.

However, before the AI ​​system passes enough training data and is exposed to multiple use cases, it may not perform better than traditional software. Depending on your business problem, the added value of this AI solution will be obvious once the solution is fully operational and after its learning curve has improved.

Due to this critical factor, AI startups have difficulty using the freemium model. AI takes more time than traditional solutions.

My latest AI project does earn revenue, but it may only account for 9% of the amount we spend on it. These numbers have improved as we reduce spending while our subscription revenues continue to decline.But in my experience, AI products are not profitable when they are first launched.

AI as a service

I noticed that most companies use at least one "service". In fact, it enables them to focus on their core business and spend less on an important service. Obviously, the way companies have built technology stacks has changed in recent years, mainly due to major changes in digital platforms and microservices. "As a service": Any software that can be called over the network because it uses cloud computing.

The success of such a solution can be explained by the ease with which such a solution can be purchased. In fact, in most cases, you can buy from a third-party vendor, make some changes, and start using it right away.

For companies that cannot be budgeted or unwilling to build an AI solution, AI-as-a-service is the perfect answer without wasting opportunities. As with other "as a service" options, the same approach applies to artificial intelligence.

Data is the driving force of machine learning.

In the coming year, companies will quickly begin to incorporate machine learning as a service (AIaaS) into their technology stack for a variety of reasons.

From what I saw (Other business models do exist!), if we look at it from a larger perspective, global companies can produce and get so much data, but not necessarily the data necessary to answer specific business questions.

They can easily build and train their own machine learning models. This allows them to offer it to external companies like MLaaS, just as they have more data center space, they can provide IaaS (infrastructure).

In addition, you have an AI development team that builds tailored solutions for your customers. These companies use the tools created by these global companies to build these AI solutions for small companies.A new ecosystem and business model is emerging.

Generally, smaller companies cannot obtain enough data to create powerful AI models; however, they do have valuable and precise data (and excellent business knowledge) to start building excellent data sets that can be used in AI projects.

in conclusion…

Obviously, a new AI business model may emerge. I believe that it takes a while for an artificial intelligence startup to find the "right" formula for artificial intelligence success. It's important to choose a business model that will enable your business to grow effectively.

This article is reproduced from the medium,Original address