Staying relevant during the transition to the platform economy: The case for a full-stack AI platform
Assessing the impact of the platform economy is in the eye of the beholder. We at arago are passionate about providing our clients AI and technology capabilities that (currently) only technology giants such as Google, Amazon or Baidu have. But their capabilities are not commercially available. In contrast in discussions with many of our clients, we get a sense of how companies of the established economy are often paranoid about being disrupted by platform companies and start-ups. And this sense of being under constant attack is further fuelled by the many metrics that paint an intriguing picture of how platform companies are challenging the status quo of the global economy. For instance, in 2019, 7 of the 10 largest global companies by market capitalization were platform companies. These metrics should be seen as an indication that the management strategies of the industrial age, with their focus on mass production that thrives on repetition and standardization must be rewritten. The recognition of these seismic changes is widespread. And we see that in our discussions with the many C-Suite executives who come to us for help. But they are struggling to translate these challenges (and often the associated paranoia) into actionable mandates for their operational teams. Unsurprisingly, not every company can reinvent itself into a platform company. Yet, in order to stay relevant, they must devise strategies in order to be prepared for a continuously changing environment where they can mitigate the disintermediation of customer relationships.
Thus, the established economy must develop a playbook for how to react to this permanent state of being attacked in their core businesses. The strategic levers for those playbooks include harnessing ecosystems and creating new business models. But more than anything, the technology stack underpinning those companies must provide the flexibility to react and adapt to changes in the environment. And this is where the next development phase of AI will come to the fore. To turn all this into compelling management strategies, organizations need leaders who have the vision to blend data and technology innovation to create new business models. Yet, success will hinge in equal measure on their ability to drive change through the organization and transform their talent pool. Having said all that, what are the concrete steps for leaders to accelerate the transformation of their organizations? To get the necessary insights as well as some much-needed grounding, I sat down with our founder and CEO, Chris Boos, in order to advance the discussions.
Chris, the term platform has been abused a fair bit. Could you outline the thinking and strategy behind Arago’s approach?
Platform and platform companies are two of the most confusing terms these days. Everybody is talking about building a platform or running a platform. So, we at Arago need to convey our differentiating strategy and capabilities. Our vision is that we want to be a strategic AI supplier for the enterprise and in particular for the established economy that has to find answers in order to compete with the new giants entering into their backyards. We believe that in a typical corporate environment it is very difficult to create a fully-fledged AI stack because a very long ramp-up time is required. Another reason is that it needs a lot of diverse intellectual property and it is very hard to find the talent for that. We have built such a stack and we have collected the necessary data to make this stack available even to problems inside enterprises that only have little datasets. Consequently, we are supplying that stack to the enterprise, and we are equally supplying to them a toolset so that they can use it immediately to automate tasks and processes. What makes it a platform is that we are also offering access to the data and to the tools as an API. Crucially this is not in the form of “build your own AI”, as you can get from many great cloud providers, but it is on top of the automation that you are already implementing with AI so that you can build new business models. And that is exactly what we are running as a platform. So, we want to be the supplier that gives the enterprise the same model, the same type of tools, the same principles that the typical giant platform company is using today. And we offer them a ready-made, easy to deploy, solution and at the same time we give them full access to the APIs to build on and put their own services onto the data applying AI. This allows them to automate tasks as well as processes and at the same time the platform collects and structures data for them.
Why is an AI stack so important for this strategy? Most importantly what is the benefit for clients?
The challenge with AI is that there are so many different models and approaches. For example, the hype topic of the last years was Machine Learning (ML) or even more specifically Deep Learning (DL). With DL you can perfectly build what would be an instinct in a human or animal. And a lot of people believe that is all AI is about. But AI is much more than that. AI is problem-solving. AI is the ability to do inference on top of known or observed causalities. AI is the ability to automate things that humans do not just do instinctively but also through thought and the ability to reason. AI can cover prediction as well as judgement if we go beyond applying a single algorithm family. Therefore, if you want to offer the ability to automate throughout the business and not just as a point solution with a specialized AI solution, it is important that you own an entire AI stack. Instead of having just data collection, recognition and then immediate execution of whatever you have perceived, you want data collection, cognition, interpretation, simulation, decision-making, execution, AND reflection. All these points are part of effective decision-making, process management, and automation. And therefore, you don’t want to end up with pre-defined and industrialized rule sets. Rather, you want to end up with things that react to variety and potentially look like gut feeling just because there is so much variety in a system that you can’t determine the rules for it using Machine Reasoning and rule-based or deterministic models. Therefore, it is important to have that toolset and a full AI stack that you can offer to the enterprise space. Without a full AI stack, an enterprise needs to build a single point solution for each problem they want to automate of apply AI to. Without an AI stack, the next time the enterprise wants to tackle another problem, they have to start over from scratch and spend time collecting new data. So whatever you do, when you run an AI stack you should be automating, performing, and creating; creating money or new business models and at the same time creating more data and more knowledge so that you can automate the next thing with much less effort. And that’s the point. You can introduce point solution after point solution, but it would take you years and years to take you through the entire enterprise to scale and parallelize the AI capability. With the HIRO™ AI stack, the ramp-up time for building and using more than one AI solution dramatically decreases and it means that our customers can apply AI very quickly and expand AI rapidly across multiple processes throughout their business.
Could you give us a sense of what kind of innovative business models clients could develop on top of HIRO™?
The first thing that all our clients do when they use HIRO™, is that they optimize things that are there already. But they don’t automate single-use cases, say in server management, they don’t automate just “low disk space”. Rather they automate server management as a domain of knowledge, that’s the point to keep in mind. And from server management, you could go to network management. And from network management you can go to security management, and from security management you might want to go to HR management. And this can be deployed to other domains. You could take this to all the business processes until you get to order-to-cash or even audit. The strategic lever is that because you automate something you are getting a much better bottom line from the automation, but then you progress to get a better top line as well. You are getting a much better top line because you are now able to offer completely new services that were either impossible or economically not viable. For instance, if you are selling animal-nutrition to farmers you could participate in the actual result of farmers selling their produce to the supermarkets. If you can guarantee that you have good quality assurance for the farmers, then you have a new business model. Or you can build logistics where you don’t follow the same tour for vehicles but completely individualize the tour depending on traffic, not just the load of the truck but depending on all the influences including maybe even who was home and who wasn’t. These are the types of business models that push disintermediation by putting our enterprise customers into the same boat as their clients. That is part of the challenges that we tend to discuss around the notion of Digital Transformation.
Another key differentiation is that we provide an offering for developers on our platform. We offer folks who manage Digital Transformation services a development environment so that they can put their ideas to work on top of a data pool, on top of an AI stack, and on top of a marketplace where they can put in their solutions so that they can much faster deploy the cool new models to the enterprise world. Three things are important in that regard. First, achieving simple automation and thus improving your bottom line. Second, adding new business models in terms of things that were economically unfeasible before or that disintermediate the top line. And lastly, if you are a service provider, developing your own services on top of the HIRO™ platform.
Why is data so vital for effectively running an AI stack in the enterprise? And how can enterprises own their data and their knowledge anyway? How is this part of the Arago HIRO™ platform?
Data is said to be the oil of the 21st century. I don’t overly like this phrase, but data is certainly important for ML and any type of statistical problem that we must deal with. Normally it means that in order to describe a small environment, you need a lot of data. You collect this over time. And thus, what you are teaching the machine is the average over time. That works very well but only in stable environments or environments that produce lots and lots of data. Most enterprise use cases do not produce enough data. There is simply not enough data, or the enterprises don’t have access to all that data to describe those more complex environments. The only way around this is if you start putting whatever you are trying to automate inside a larger context, let’s say HR processes for expats, but you only have 100 expats. That’s too small a dataset to run anything on. But if you could put this dataset into a bigger context, then it works. And now you may say, that means that everybody must share their data, or give it to a single platform. This is not what we are doing on the HIRO™ cloud services. For us, it is very important that clients own their data. Just as important for us is to have clients own the knowledge they put into the system. So not just the stuff that gets machine-learned but the intellectual property they put in directly, remains their property and under their control. We will never be the competition of our clients and we will never sell their data. Rather we have built the platform to support clients. We have processed so much data from all types of industries and enterprises through our initial data collection in the IT world where we are coming from. Without leaking any data, we have a fairly good semantic understanding of how an enterprise works across different sectors. We semantically and statistically determined how things are interconnected and that way we can put data that comes into a specific structure and thus more context. This is how we can solve problems today even with small datasets. It’s also why we can decrease the ramp-up time so much. In mature processes it takes just a couple of weeks. In totally new domains that we have never been active in before, it is approximately 3 months. That is exactly why we are working on these large datasets. Because we are extracting semantics and we are extracting interdependency out of those datasets. Anyone will only ever have access to that data if our clients choose to share the data with that person. Which incidentally is a feature on the platform that we offer to clients so that they can make more business, but it is not a business model that we are running. Clients can rent out their own knowledge or give access to their own data, but only their own data and nothing else.
Lastly, could you gaze into a crystal ball? Where do see you Arago heading over the next 12 to 18 months?
I do not have the proverbial crystal ball to be able to tell what exactly Arago will be doing over the next 18 months. But what I can tell you is that we have worked very diligently over a long period of research to put the algorithmic base into place. We have then used this algorithmic base in IT Automation to collect a dataset that yielded a structure on which we can now build all types of automation. This allows us to automate business processes much more easily. What you will see in the next few months is that we come out with more ready-made solutions that are deployable within weeks. And we will automate business processes for clients by processing individual client data, individual approach clients have, their individual IP, knowledge, ways of doing things. At the same time, we are actually emulating the problem-solving capabilities humans have today and put them into an AI where people can just use it.
You can see the full series of interviews with Chris here:
