diwo: a cognitive system for continuous business optimization
diwo® is a platform developed by Loven Systems for supporting analytics-based decision making in businesses, with novel cognitive capabilities to enable continuous business optimization. The design of diwo is based on patent-pending methods built around the concept of an opportunity, and a cognitive decision-making methodology, called SEAL™ (an acronym for Sense, Explore, Act, and Learn).
Informally speaking, an opportunity is a time-sensitive business situation which—if not addressed in time—will have an adverse impact on the financial performance of the business, either due to missing a potential gain or by potentially incurring a loss.
diwo reveals business opportunities in real-time by continuously analyzing incoming data streams from various sources. It explains the potential impact of revealed opportunities, further recommending an optimized strategy to address the opportunity within the current business context.
Furthermore, diwo also provides flexibility for decision makers to explore alternative strategies by tweaking certain decision levers. Once a strategy is approved by decision makers, diwo provides a list of tasks that can be carried out to implement the strategy.
diwo enables businesses to transition from courageous business decisions to smart business decisions by enforcing a systems-thinking mindset, through augmenting human cognition and striving to continuously optimize the financial performance of the business. In diwo, a business is reimagined as a cognitive system whose sole purpose is to optimally respond to opportunities of certain well-defined types, as quickly as they arise.
A business operates in a space of opportunities. While not every opportunity may be practical for a business at any given time, there is a vast space of viable opportunities which slip by—either because they were never revealed, or because they were discovered too late to react in time.
Continuous business optimization works in two ways: first through enlarging the space of addressed opportunities by revealing them on time, and then by providing guidance to address them in the most profitable ways.
A cognitive system is any system that exhibits a certain level of intelligent behavior, as evident from its cognitive responses during interactions with its environment. There are various levels of cognitive ability, ranging from the reflexive level (the lowest level) to the self-aware level (the highest level).
Inelligent systems differ in their cognitive ability, and not all tasks require the highest forms of intelligence. The best approach is to apply the right level of cognitive ability for the task at hand. diwo is a cognitive system which exhibits human-like intelligence, for real-time business decision making.
diwo’s cognitive ability is evident from its three personas which support three different–but complementary–forms of interactions with decision makers.
The diwo ASK persona supports goal-directed guided conversations based on patent-pending technology from Loven Systems. Decision makers can converse with diwo in natural language using either voice or text, and diwo responds in a multi-modal manner by using voice, text, and interactive visualizations.
In contrast to simply answering a question posed by the user, diwo applies deliberative intelligence to understand the problem that the user is trying to solve. With that understanding, diwo produces better answers and gently nudges the user towards solving the problem faster. diwo also learns from user interactions and has the ability to adapt to individual problem-solving styles.
The diwo DECIDE persona supports guided continuous business optimization by revealing opportunities through automated intelligence, and by guiding development of optimal strategies. diwo builds user trust by providing explanations and evidence for its recommendations, while also allowing users the freedom to explore other options. Besides using domain knowledge and continuous learning, diwo DECIDE works behind the scenes, using advanced analytical methods from the repertoire of descriptive, predictive, and prescriptive algorithms as appropriate for the specific opportunities addressed. diwo DECIDE is configured with the appropriate domain ontologies for all the opportunity types it handles. These ontologies can be edited or augmented as new opportunity types are introduced to diwo.
The diwo WATCH persona supports continuous observations and a fast decision-making mode for routine and less complex operational decisions. These observations—along with on-the-fly streaming analytics—are provided in “always on” and “always available” visualizations through browser and mobile application interfaces.
As a software system, diwo is a reactive distributed system comprised of many intelligent sub-systems, which collaborate with each other to create a cognitive synergy—yielding intelligence that is far beyond the reach of a single sub-system.
diwo software design follows the reactive manifesto and all its components are responsive, scalable, resilient, and event-driven.
diwo uses a micro-services architecture realized by asynchronous communicating actors that comprise the diwo Knowledge Management layer. diwo incorporates 10+ actor systems which run concurrently to provide its core functionality.
Front-end and back-end layers are cleanly separated, with a messaging layer mediating between the two. This layer also serves as a gateway for external data and service used by diwo. Internally, diwo actors communicate via asynchronous messaging.
Most analytics, including batch and streaming, are performed using a distributed computing engine. Pre-developed models can be imported into diwo using PMML standard and deployed for scoring in diwo analytics pipelines.
diwo runs on a dedicated cluster of commodity hardware with redundant storage and computing nodes to provide fault-tolerance and resiliency.
With an on-premise application philosophy, diwo is fine-tuned to run on a standard configuration in a dedicated environment, but can also operate in a cloud environment if desired.
• General purpose high performance machines (e.g. M5 on AWS) are required at minimum. This family provides a balance of the computing, memory, and network resources required by the diwo application.
• GPU computing is recommended for machine learning and deep learning purposes. Optimum speed and efficiency of neural net models require GPUs.