In 2020 and beyond, no organisation can ignore the enormous shift in the way that information is collected, stored and analysed.
The AI war is full-on. Like all tech giants, Oracle years ago decided to join the fight for the AI chatbot and voice bot users. And that makes total sense. As a benchmark platform for secure data storage they are able to fill a niche that the Facebook-dominated chatbot sphere can not: A comprehensive architecture that interacts, stores data, learns and helps the organisation make better decisions on auto-pilot in one all-encompassing secure environment.
In 2020 and beyond, no organisation can ignore the enormous shift in the way that information is collected, stored and analysed. Industries like healthcare, high-tech and government are much in need of smart learning systems as the possibilities of our systems continue to grow. With the advent of AI our systems can now converse with users in highly autonomous processes. Yet for these industries Facebook and similar platforms are not an option due to privacy concerns and Facebook’s open-platform approach to security.
Customer data remains the most important pillar for decision making in the enterprise.
Customer data, if well-managed and accurate, remains probably the most important pillar for decision making in the enterprise, and there is nothing new about organisations leveraging data models for prediction and analysis. But what is new today is that Conversational AI is able to work these data pools on two ends: improving the AI through user input and expanding the data pool on one end, while simultaneously creating a deep learning-enabled intelligence that makes increasingly accurate predictions with little to no human interference. The Oracle Digital Assistant is very well positioned in these two fields.
So if Oracle is such a great candidate for these modern AI systems, why haven’t we heard about their bot builder before? Well, it’s an enterprise system. And enterprise software is not typically advertised on national TV or TikTok Ads. In fact the team of the Oracle Digital Assistant (ODA) has been working on their bot platform for years. Its most eye-catching characteristic is that it requires us to code our own bots using their Oracle Bot Markup Language (BotML). Here it is: BotML is both the strength and the weakness of the ODA bot platform.
It’s not surprising that ODA is not popular amongst the pro bot builders yet. Why write code if there are the likes of Manychat and Chatfuel where zero coding is required? It’s not a feasible solution for small and medium size businesses with limited budgets. Coding a bot on BotML takes 3-4x as much time as building one on the WYSIWYG bot building platforms. It requires study, designing the bot flow, writing the code and debugging. Lots of debugging.
Is its high resource impact on the organisation a fatal flaw? I don’t think so. Enterprises have different needs than small and medium size businesses. Big organisations make big targets and that’s why data security is a number one priority, and Oracle’s reputation on data security is unparalleled. Second, if your organisation has running contracts with Oracle and your engineers know their way around the platform, your lead time is greatly reduced. Your new bot will integrate with your existing systems like a charm.
Let’s now look at the tech side of ODA. One of the main advantages of writing bots in OBotML is that it’s extremely versatile. All your flows and elements can be customised in detail and there are few limitations on how these elements work together. A single ‘state’ — which may be translated as a ‘node’ or ‘block’ — can contain parameters for the receiving platform, content, how it relates to the next state, actions and variables. Setting variable values can be done dynamically by using custom Apache FreeMarker code, making it even more flexible.
The Natural Language Processing (NLP) component is not coded. Whoohoo, we got a break here! These cognitive learning protocols are too complex for even Oracle developers to create and have been pre-added. There are two learning models to choose from and we can set a custom value for certainty threshold to better steer intent matching.
It is my strong belief that organisations who start developing these integrated deep learning systems early will have a major edge over their competitors in the years to come.
The data that our bot collects can be analysed through ODA’s rather primitive analysis tool which is really only useful for improving our bot skill. The real power lies in building up our data pools in co-ordination with the Oracle database systems and running our own Machine Learning and Deep Learning queries on this. It is my strong belief that organisations who start developing these integrated deep learning systems early will have a major edge over their competitors in the years to come.
To sum up: Even though Smart Conversational Bots on Oracle are still maturing, the Oracle Digital Assistant is already one of the most powerful bot platforms for sales and customer service automation while populating data pools for data analysis and customer behaviour prediction. With the addition of ODA the Oracle Cloud Platform is no longer a simple data warehouse, but a full-scale AI platform that has the tools to automate everything from customer interaction to predictive analysis and back.
Please note that I’m in no way affiliated with Oracle or the Oracle Digital Assistant. The views presented here are based purely on my own experience working with ODA in the past year. I hope it helps.