by Kumar Singh, Research Director, Automation & Analytics, Supply Chain Management, SAPinsider
Our fascination with Artificial Intelligence (AI)
For hundreds of years now, humankind has envisioned machines doing our tasks for us and has imagined creating intelligent machines and robots (The Czech writer Karel Čapek introduced the world to the word robot, by way of his play Rossum’s Universal Robots (1920) and the novel iRobot, from my favorite author Issac Assimov (1950), is still considered THE classic by science fiction enthusiasts). Our vision of these machines are that they need to be far superior to humans in many aspects (intelligence, strength, agility etc.). And imagination is a very powerful tool- since it forms the genesis of innovation (simple to complex).
As computing power and technology advanced exponentially in last few years, this advancement allowed us to harness the power of AI and ML algorithms that have been around for a while. And as we explored further and applied many of these successfully, we started envisioning scenarios where algorithms can take over the day to day operations planning aspects. This vision was further fueled, in the world of operations, by rapid automation of warehousing and manufacturing operations.
But the key question that still remains years even after so many years – are we there yet ? While we may not be there in terms of algorithms running our organizations end-to-end seamlessly, we are already in the era of Enterprise AI platforms. Enterprise AI is a powerful capability to build and many organizations are either already close to building this capability or already exploring building this capability. But a initial key question is – What exactly is Enterprise AI and what are Enterprise AI platforms ?
Defining Enterprise AI
You will find many definitions of Enterprise AI. And they all come in different flavors. Some are completely off, some come very close but are loaded with technical jargon while the rest walk the fine line in the middle. These are all designed based on what those “defining” these definitions are trying to sell you. So let me define it in a very simple way.
An organization has been able to build Enterprise AI capability when it has embedded AI and ML capabilities across its operations and planning processes. It is an enterprise wide application of AI capabilities, mass usage of AI and ML enabled solutions and algorithms to facilitate optimal operations and decision making.These AI and ML applications “live” in the same environment as other business applications and frequently exchange information with them, in addition to tapping into the same “single source of truth” data.
An important point to understand is that two, three or even ten successful insertion of ML algorithms do not give you Enterprise AI capabilities. Having scattered, siloed AI & ML algorithms and tools across your enterprise is NOT Enterprise AI. Enterprise AI capability is when you build AI capabilities across functions, and then eventually connect those capabilities so that you develop an enterprise wide network of those AI enabled capabilities. A key point to keep in my mind is to not confuse AutoML tools with Enterprise AI tools. If you are thinking about developing internal Enterprise AI capabilities, AutoML tools can help facilitate and accelerate that process, but by themselves, they are NOT Enterprise AI solutions.
While in its initial stages, Enterprise AI capabilities may be fragmented, eventually, your organization will have one single Enterprise AI platform.
What is an Enterprise AI platform ?
A true Enterprise AI platform will typically have dozens or in some cases, hundreds of AI/ML enabled applications within the platform (which is the very essence of a platform. One of the key aspects of a platform is that it is designed primarily to bring a variety of tools together in one single environment, so that they work together and exchange information seamlessly with each other). A true Enterprise AI platform is the pinnacle of your AI journey and once you have developed that capability, you can harness the full potential of AI and ML in your end to end operations.
While theoretically, the platform can be on-premise or cloud, realistically, it needs to be in the cloud. We will not get into the defining the benefits of a cloud platform in this article but one key additional aspect in case of enterprise AI platforms is that true AI and ML algorithms, and some advanced approaches (like deep learning) are data and computing power intensive.
Having understood what Enterprise AI platforms are and their capabilities, let us try to answer the question around whether you should build or buy Enterprise AI platforms.
SAP Provides great foundational capabilities
While we will cover in a seperate article how SAP Data Intelligence can help facilitate and accelerate building Enterprise AI capabilities (irrespective of whether you plan to build or buy), the illustration below from SAP shows their vision of how Enterprise AI (“Intelligent tools”) fit into the SAP ecosystem.
Picture source: SAP
As you can see, foundational capabilities like Digital platform are critical and SAP provides plenty of solutions. Also, with RISE with SAP, as organizations standardize their solutions and configurations, SAP has a great opportunity to offer its own Enterprise AI “plug and play” solution. We will discuss that in a seperate article.
What does this mean for SAPinsiders – Should you build one or buy one ?
This has always been one of the most important questions that decision makers face, not only in Enterprise AI applications but across functions. So what is the answer when to comes to building Enterprise AI capabilities ? The answer depends on many factors. Let us explore some of those factors below (this is focussed on organizations that intend to use Enterprsise AI capabilities, not planning to develop this as a product they intend to sell):
Size and resource: If you are a giant in your Industry, with mega resources in terms of people & technology, it may not be a bad idea to consider building your own capabilities in this area. Unlike off the shelf Enterprise AI tools who sometimes have to design their algorithms in a way that it is not highly customized (and hence can be leveraged for a wide base of customers across industries), you can build customized Enterprise AI algorithms. And of course you are a large organization, you have some of the key resources (talent, foundational IT capabilities and $$$s to spend on hiring partners). You may have to invest in a private cloud to train your algorithms, specifically if you are leveraging number of deep learning algorithms.
Patience and vision: Building a true end-t0-end enterprise capability is a journey that is not short. Irrespective of the amount of resources you throw to the initiative, it will take years. And this is something you need to be comfortable with and cognizant of when planning to build. You also need to be comfortable with the fact that you will always have to keep making tweaks to your applications on the platform to align with new realities and evolving organizational objectives.
Foundation or Readiness: Where you are currently in terms of your digital capabilities are also important. If here is too much that needs to be done in terms of foundational elements, maybe it is time to re-consider. Building true enterprise AI platform capability needs a very high level of tech innovation culture.
If you were thinking about building these capabilities internally and are disappointed after reading the three aspects mentioned above, there is plenty of good news as well. And the good news is that there are many Enterprise AI solutions available now and there may be one that may perfectly fit your requirement if you know what to look for.
And if you decide to build, a very lucrative aspect of this option is that first, you can identify which unique algorithms are better fit for your enterprise and then build only those capabilities in your platform (like the supply chain algorithm examples shown in the illustration below). But again, building a platform internally is a massive commitment.
Kumar Singh, Research Director, Automation & Analytics, Supply Chain Management, SAPinsider, can be reached at email@example.com