Imagine a self-driving forklift reacting to an obstacle only after a minute. Or, the next best action for a customer being given to a call center agent only after the opportunity to provide this customer with a special offer is gone. Or, speech recognition and –generation systems that are used in customer engagement and customer service and that do not understand the customer.
Things can be as mundane as having the ability of early identification that a shipment will be late plus suggesting, or directly initiating options to resolve the problem.
Real Time Means IoT
In today’s fast paced world companies need to more and more rely upon devices that include sensors and effectors. These devices – or things – are interconnected. And they connect the physical and digital worlds. This, on one hand, enables the automation of complex processes, and on the other hand produces and needs a massive amount of data to maintain efficient operations.
The entirety of these interconnected devices is also known as the Internet of Things or IoT.
These sensors are used in all sorts of business processes, customer facing or otherwise, starting from marketing, going on through sales, and service, but also covering operational processes.
The challenges that arise from the fire hose of data that gets produced by all these things is that their noise to signal ratio is mostly extremely high and that this fire hose needs to be meaningfully integrated into the available business systems.
Or else it is useless.
In other words, the data that comes in from the sensors needs to be analyzed for patterns and/or abnormalities, depending on business objectives. The results of this analysis then will be used to (semi-) automatically take decisions and to drive effectors. While this analysis and decision process can happen via Big Data warehouses in a batch fashion there is an increasing need of real-time execution.
Cue Machine Learning and Artificial Intelligence
AI and machine learning add exactly the ability to enable the real-time requirements of IoT landscapes. There is far too much data for successful manual intervention. This massive amount of data can be handled only by machines.
But why intelligent, learning machines?
Because it is also impossible to foresee all possible combinations of data. That’s why the machines need to learn about the effects of inputs and the possible actions instead of being programmed with a finite set of rules. Think of the forklift again that needs to learn that the rain drops that trip off its sensors are not an obstacle, while the person in front of it is; or the offer management software improving the next best offer based upon continuously learning what finds interest – and what not.
If this sounds like it is all about doing more with less – aka increasing efficiency – it is a wrong understanding. AI, especially in combination with IoT technologies, provides its real benefits not in scenarios where efficiency is increased, but where there is a contribution to an increased customer experience by improved engagements. By doing more or entirely different things. Ultimately it is about doing things that haven’t been possible without the ability to work on massive amounts of data in real time to provide a business outcome.
This is also something that some vendors have realized after relentlessly pushing the buzzwords AI, machine learning, and IoT. Real business value can only be delivered if there is a way of seamlessly integrating the IT stacks behind these two technologies. And, as Constellation analyst Holger Mueller rightfully says, since today’s technology is in the position of being able to solve problems that businesses haven’t yet realized, there also needs to be a design component to it.
This is, e.g., the idea behind the new SAP Leonardo infrastructure that got announced during the 2017 Sapphire now and that I commented on earlier.
There are mainly three, involving lack of interoperability standards and poor integration between AI and IoT.
IoT Interoperability Standards
Currently there are no real standards that foster interoperability of the vast number of different sensors. There is also not yet a frontrunner that could emerge and set the standards. Add to this the necessary connectivity with business systems. Instead we do see a lot of API-driven approaches, which make each implementation a custom implementation. This makes it a risk to settle on or even decide for a software platform as the connecting tissue. There is a strong need for more ecosystem play.
AI Interoperability Standards
What I just said about IoT platforms is equally true for AI platforms. It only takes a look at engagement platforms like Siri, Alexa, Cortana, or the Google Assistant. And then there are countless other vendors that can do virtual private assistants, be they phone makers like Samsung or companies like Nuance Technologies. And this is only one type of platforms.
Integration of AI and IoT
This necessity compounds the problem. So far, most vendors concentrate on one or the other, thus creating technology silos. IoT or AI, the latter more and more looking at incorporating technology into business software to improve outcomes. A few, like SAP, try to build platforms that break these silos by recombining them in their software stacks.
What it Takes
In one word: Planning.
And a bit of luck.
Yes, luck. Luck, because in order to be able to offer leading engagement ability for best possible experiences early to some extent means trial and error. A chosen platform might not be up to the task or its vendor perishes; the sensors may measure the wrong things; the AI is trained the wrong way.
What this basically means is that executives right now need to carefully observe the market and at what is possible with technology. For leaders and early adopters it is important to start to evaluate and assess platforms and solutions already now, starting from clearly identified needs, followers need to closely observe them. But neither should already do a decisive move unless they have a strong customer facing strategy and there is
- a viable vendor or group of vendors that
- closely fits into the existing strategy and that
- clearly supports the current needs and those that are foreseeable for the next five years.
In the next few years about three to five mainstream platforms will evolve, along with several specialized platforms that concentrate on narrow problems. These platforms will support a small set of standards.
In addition to this it requires the willingness and ability to put whole business models to the test and come up with business models, solutions and services that, within an ecosystem — maybe a mini-ecosystem — deliver solutions that are geared around adding value to the customer. These solutions will revolve around pro-actively solving issues. Proactively, i.e. before they come up. And they will not be about making the sale but about sharing the value added at the customer. This is a shift from inside-out thinking to outside-in thinking.