Blockchain, hyped as the technology is at the moment, has not only quite some potential but has matured considerably over the past months.
Still, after having critically looked at some promising use cases of blockchain throughout the last articles of this column, it is time move back to its original theme for a change.
After all, the main theme of this column is Artificial Intelligence Trends and Tools. And bots, AI, and machine learning are exciting topics in their own right, let apart in their combination.
So, here we are. Let’s have a look at some chatbot and digital assistant developments …
Since these are other technology areas that are also maturing fast. And this is exactly the reason why it is interesting to revisit them.
Chatbots Edging Closer to Mainstream
Chatbots, in particular, the voice-enabled variety, are getting closer and closer to the mainstream. We have recently seen a first partnership of Alexa and Cortana, which was analysed by Brent Leary. As Brent wrote, the partnership between Microsoft and Amazon is promising as it helps developing common standards. These are mandatory for scaling. Walled gardens do not help.
Additionally, vendors are making it easier for businesses to generate the required conversation flows and to, therefore, deploy meaningful chatbots across multiple channels, like digital assistants, messenger applications, or websites.
Especially in competitive and commoditized industries like banking and insurance, companies are aggressively deploying AI driven chatbots as the primary customer service channel. At the same time Grand View Research expects the Chatbot market to increase with a CAGR of nearly 25 per cent until 2025.
Looking at it from a ‘future of work’ angle, more and more organizations realize that it is not man vs. machine but man augmented by machine. This is, for example, one of the major results of a survey conducted by Deloitte in May and June 2018 with mid-market companies. This is an understanding that is of crucial importance.
The fourth main reason for chatbots getting more into the forefront is that we now have the technology and computing power to generate insight out of available raw data. The power of combining insight with ‘human’ a way of interacting with computers is hard to overestimate.
Of course, it is not all hunky-dory in chatbot land. We are still facing some serious limitations.
The most important one is the still limited scope of AI and advanced analytics engines. These engines are a mandatory precondition for building meaningful chatbots. When we talk about artificial intelligence we are still talking about weak or narrow AI. These are AIs that are focused on performing a specific, narrow task.
In contrast to this, a strong or general AI is capable of performing any task a human could perform. This type of AI will still take quite some time to be developed. There is no HAL in our near future.
This scope limitation makes solving the interoperability challenge even more important. The different AIs have different scopes, purposes, and strengths. To be able to solve more complex business challenges it is therefore often necessary to use different AIs in conjunction.
But they also have different APIs, which makes it hard to use digital assistants in orchestration, or through each other.
In the business world, this is particularly a challenge for Microsoft and Cortana, as Alexa, Siri, and Google Assistant are strong in the consumer world and directly exposed through mobile operating systems. Plus, vendors like SAP and Salesforce have partnered with Apple.
The data that helps businesses solve issues for their customers and employees exists, within the businesses and outside.
The main challenge is finding the needle in the haystack, or creating insight out of this data and to make it actionable.
By doing this, it is possible to provide value for both, customers and employees by addressing their needs one by one, in a prioritized fashion. But this needs the ability to use different AIs in orchestration.
Building an AI and bot infrastructure requires businesses embarking on a journey. This journey starts by having strategic objectives in mind at any time – and being able to formulate them. These objectives need to be the customers’ or employees’. It is necessary to maintain an outside-in view to be and stay successful.
Based upon the chosen objectives it is possible to collect and organize the data that relevantly supports the necessary decision-making to meet them. Looking at relevant data is important, as it is a foundation for building trust to not just arbitrarily collect data.
The journey needs to be sustainable in a volatile environment; that’s why it is important to apply a think big – act small strategy with frequent review cycles of strategic and operational goals. This is challenging, as it is difficult to identify what data is relevant for future bots.
Still, it enables organizations to deploy meaningful intelligent bots early and frequently.
Once this strategy and the immediate priorities are clear, the next step is to define a data grammar, to lay out data semantics and create the underlying data structures that define how relevant entities are looking like technically.
With this at hand we can go on identifying the existing and relevant data storages and to start harmonizing them for bot use. Else there will be data silos with redundant and more likely even inconsistent data, which makes it very difficult to generate insight. The AI that feeds the bots needs good data.
In parallel, it is possible to prepare and analyze the existing data with respect to the objectives to be achieved, be they in customer service, sales, or marketing.
Creating a Bot Infrastructure
This harmonization process requires a crucial decision being made: The decision for a platform that immediately helps delivering the chosen bots and that is viable in future. This viability also includes the ability to integrate with different digital assistants. This decision is crucial as it is hard to change.
Most of the technical platforms do have the advanced analytics and the machine learning capabilities that are necessary to build and deliver meaningful bots. For those, which do not have it to a sufficient degree, their ecosystem comes into play to deliver both, the AI as well as the bot infrastructure. Looking into the SAP world as an example, there is not only the SAP own conversational AI (formerly recast.ai) but also great emerging players like Cognigy that got recently recognized by Gartner as a cool vendor. Then there are multiple independent vendors like kore.ai or Intercom, or Helpshift, or agent.ai, if the focus lies on customer service.
Not to mention the services provided by Google and Amazon.
Creating a bot infrastructure that takes the limited abilities of underlying AIs this way into account will lead to having a number of bots serving different but often related purposes.
Therefore, the final step is planning for and building up bot swarms and ultimately their automated coordination and collaboration to ensure handoffs bot to bot and bot to human.
Need help? Just ask me.