I recently had a request from a partner, telling me that they “needed an AI for a customer and what I could offer”. I was close to responding with “I have a few green ones lying around, how many would you like?”. But joking aside, this is a good example of how hyped AI has become and how far from reality we are at this stage. Everything is “AI powered” in an attempt to promote the product and many organizations seem to feel they need an “AI” to show how “up to date” they are. While I am all for effective marketing, pushing everything as AI is plain ridiculous and as customers we should question everyone pushing AI at this stage and refrain from paying a premium for anything with an AI label on it – the vast majority of which implements no machine learning anyway.
But before we look at if and why you should even care about Machine Learning and AI, it is important to understand the difference. There are a range of articles on the subject, one of the nicer and well-written brief summaries that I have recently come across is the one by Calum McClelland which you can read here.
So why should you care about AI or machine learning anyway and what is the real challenge in standing up a meaningful solution?
The business rationale
Beyond having your teams “tinker” with machine learning technology to get a feel of what it can do and how to build models, the immediate next question you have to answer is “why”. Building meaningful solutions will mean a lot of investment in time and money as you are going through a steep learning curve and in many cases your organization will need to transform as well and think about products and services differently. This is one of the core elements of the so called digital transformation, transforming the organization such that it can apply technology effectively in everything it does. This also means that it is a cultural change in many different ways. The key takeaway here is that it is not just about building an “AI”, it is really about taking different offerings to market and looking at a revised value chain as well as an organizational transformation. Given the effort involved there must be a strong business rationale.
Examples can be as diverse as the organizations out there, but I have seen good examples in digital learning both in K-12 right through to professional and in-company education where machine learning is used effectively to provide customized training and help leaders understand how best to help their staffs professional development. Other examples in industrial environments for example pertain to preventive maintenance and IoT, moving from a very traditional business and service model to a data driven model that can support customer cost savings, more environmentally friendly technology deployments as well as added value and a closer customer relationship as a consequence. The opportunities are endless, but as I am sure the two simple examples illustrate, the difference between the traditional approach in education and industry and the examples listed require a lot more than just a bit of technology.
Most people worry about the technology, but the fact of the matter is that for most uses cases that customers have, the technology is available and not the issue. Knowing which technology to apply and how definitely requires specific expertise and finding real experts is not simple, but just like the technology, they are out there.
We covered this domain in more depth in an earlier blog, but in addition to finding a meaningful offering and building the business case, the data is the most challenging element and will in many cases at least partially dictate what it is you can do, be it because you don’t have the right data in the right format and the right volume for training a good model or because of privacy or other regulation. As we discussed in our blog on the subject, this is the one domain that is often forgotten and least understood, yet it will dictate what you can do and is therefore absolutely key to understand. Machines as we know operate on the basis of “garbage in”, “garbage out” and there is no “intelligence” that can fix that on the fly. However, poor solutions taken to market too early can very much have an adverse effect on your reputation and customer confidence, something that you should always keep in mind.
The added value
So can AI or machine learning or analytics for that matter (which is the answer in many cases) provide real added business value? Well, it depends. If you have worked out your business proposition and business case well, have listened to your customers and tested that the offering is something that customers want, the product quality is good enough and it is not hyped beyond its capability, yes, it can add a lot of value and help you differentiate.
Keep in mind that there is also no shame in starting small, and experimenting. You would be surprised how many customers are willing to work with you and how much good feedback that you can get if you just play it open and as an experiment. In our experience, until you know exactly what you are doing and what customers want (who ever does know that for sure), an open an inclusive approach works very well and can have a very positive effect on your reputation even if the experiment fails.
So in closing, it is not about “how many AI’s”, it is all about what these technologies, be they AI, machine learning, deep learning, data analytics, etc. can do for your business and how they can help improve and extend your offering and make it more meaningful. This does require understanding new technologies, but it starts with an awareness of the fact that you need to “transform” to apply this technology successfully and there needs to be a business case and a customer need. So how many AI’s do you want and would you prefer green or red?