Rotem Alaluf is the CEO of BeyondMinds, an enterprise AI software provider for delivering AI solutions and accelerating AI transformation.
And indeed, making the technology beneficial in the real world is one of the core challenges we are facing today. Despite the huge potential, the cold reality is that only a tiny fraction of the value we can create using artificial intelligence is achieved today, and the industry experiences huge failure while trying to create value from the technology.
In this series, I’ll take you through some of the reasons for this failure, explain how the demand for experimenting with the technology hurts our ability to create value with it and what can be done in order to mitigate it.
A History Lesson And What Can We Take From It
As AI is the arrowhead of digital transformation, enterprises must quickly learn how to use the technology to create meaningful value. To date, only a small portion of the AI initiatives in enterprises bear fruit, and leaders are searching for the right way to create an AI multiplier within organizations.
As someone who deals with artificial intelligence on a daily basis, when I look at the adoption of transformative technologies such as AI, I try to look for patterns from history (e.g., how enterprises acted initially and how it usually ended).
I won’t take you all the way back to the Industrial Revolution and electricity, although the potential impact of AI could be similar or even greater. Examples from the last few decades include enterprise resource planning (ERP) systems, customer relationship management (CRM), IT service management (ITSM), human resource management systems (HRIS) and more. When you think about it today, it may sound crazy to develop an in-house CRM or ITSM system. Why should you do it instead of getting the help of Salesforce, ServiceNow, Workday and others, making the process faster, easier and more successful?
It may surprise you, but 20 years ago, many companies tried to develop those in-house. The CIO hired IT personnel and third parties and got help from system integrators to build those heavy systems. The fact is that the internal development almost always failed, and today, most companies aren’t using internal CRM or ITSM systems. They came to this conclusion after wasting years and millions of dollars and failing in execution.
I’m sure some of you will say that AI is different, that the commonality between different applications and different organizations is low and that it must be developed in-house in order to create significant value. That’s the same thing organizations said about CRM and ITSM systems years back. And isn’t it said that history repeats itself?
The main difference, in this case, is that the real-world organizational level of AI infrastructure is far more complex than ITSM and CRM systems, and the instruction playbook on how to build it is still a work in process. With the vision of bringing real-world AI value to every organization, I’ll try to assist in writing this playbook, and hopefully, my part won’t be in the appendix, or worse, in the “what not to do” part.
As a community, we must set a higher standard. Just as CRM, ITSM, ERP and other systems have been developed and great companies have been created, real-world AI requires a similar tool. It’s more complex and has higher value than the examples I gave, but this just makes it more appealing and crucial to solve.
AI In Enterprises Today
Although the term “artificial intelligence” was coined more than 60 years ago, only in recent years — with the rise of better and more accessible computing resources, new and more powerful techniques of machine learning, and deep learning and massive data collection in organizations — have we seen an acceleration in adoption by traditional industries.
Based on recent surveys, we understand that the trial in deploying and creating value with AI is far from being successful. A whopping 87% of AI solutions never make it to production, and only about 1 out of 10 organizations create real value from AI. The success achieved is mostly by solving point solutions, which can create value but are not a sufficient sign for the ability of enterprises to massively adopt the technology.
From the enterprise point of view, it’s like having a CRM system only for one big customer out of thousands or having an ITSM system for one group in the organization — it’s helpful, but not a game-changer. Having 10 different CRM systems for 10 varied customers would be a pain. Make it 100 CRM systems for 100 customers, and the overhead of maintaining those systems would be a nightmare. So, why do we assume it will work that way with AI, a technology that is far more complex?
In part two of this series, I’ll take you through why the demand for proof of concepts (POCs) took us off the rails, what the core barriers of AI in production are and what we can do to overcome them.