2017 was a year which had huge developments in the world of artificial intelligence and big data. The most important thing that happened was that data science, a once very niche field, moved to the center of the public conversation. As a data science evangelist, I feel this is fantastic but as I've discovered during the year, the fact that everyone is talking about AI is not always a good thing. Let me explain a bit why.

Imagine you are the CEO of a marketing startup who needs to raise funding for your new product. Now, raising funding for any startup can be tricky. You need to have a great team, a really great product and a solid go-to-market plan as well as the so-called "secret sauce" which makes you better than your competition. If you are raising money, then the best thing for you to do is to look at what are the current market trends that can make you shine. In 2017 it was VR/AR and artificial intelligence.

On the flip side, if you are an investor then you are always looking for the technologies that are just emerging and will provide you with the most chance for a quick and large return on your investment. This is where the public discussion around AI has started effecting the way investors invest in companies. Whether we like it or not, if a company has artificial intelligence in their marketing pitch, then it is more likely to raise capital then a company which is an "old" SAAS or cloud-based provider.


...


What this has caused, is a massive wave of "AI companies" entering the market. Some who have legitimate reasons for implementing machine learning technologies in their products but also others that try to stick on AI as a marketing badge to make themselves look more attractive. Great, but is this a bad thing? Well, yes and no…

As a person who has seen his fair share of successful and failed AI projects, I can say that implementing AI in your products is not something you should take lightly. I even go as far as to say that if you can solve your problem with any other more conventional means, then you should never use something as complex as Machine Learning. If on the other hand you are looking to get the competitive edge in the market, have the data, and are willing to put in the huge amount of work and effort required, then yes go ahead with AI. Sadly many companies underestimate the complexity of machine learning which is why there are lot of AI solutions out there which are more artificial and less intelligent.

But then how should investors or large companies who are looking for third-party partners validate the vast field of AI companies out there. There is no real silver bullet solution to this but there are a couple of things you should do on a basic level of due diligence.

  • 1.Find AI and Big Data competences who can help you out
  • This one really is a no-brainer. In order to validate if a company is the next big thing in artificial intelligence, you need super smart people. In this case, you need data scientists. In the case of a large company, my suggestion is that if you find them, hire them. This is easier said than done, unfortunately. The salaries of data scientists have gone up in recent years and it might be difficult for a company to justify hiring a data scientist when they don't really know if they want to get into the field. The second, and more cost-efficient opportunity is to hire a consultant or a consulting company. 2017 also saw the rise of AI consulting companies. Companies who have both the business development experience as well as the technical skill sets to work with anything from integrating AI into your existing product all the way to tuning hyper parameters of your deep neural networks. These companies can help you jumpstart your AI business and definitely should not be overlooked.

  • 2.Look into the technology stack and architecture.
  • The second thing you should do, once you have your smart people, is to go deep. Look into whatever company and product you are assessing and find out how it works. This involves understanding if the company is using open source API-s such as IBM Watson and Microsoft Cognitive Services or are they using frameworks such as: Tensorflow, PyTorch, MXNet, Caffe, H20.ai, etc. They might even have proprietary machine learning solutions which might be super-efficient but difficult to keep up-to-date and integrate into your systems. The best way to understand any third party's technology is to sign a NDA and have a face to face meeting with their CTO. This meeting should give you an understanding of their validity.

  • 3.Validate quickly and don't be afraid to cut your losses early.
  • Just like with any other new product or company it is necessary to validate quickly if the solution works as it should and that it matches with your current customer base. This means that you should be able to validate your use-case within one month. Start with a PoC and don't be afraid to show raw things to your customers. AI is still very much science and make sure your customers know that.

There are many more things a company should be aware of when validating AI partners or getting into the AI game but if you follow these three very common-sense steps then you have already reduced the risk of failure in your endeavor.