Business process management
Business processes are (people or system) activities and information that produce some business outcome aligned with your business strategy. Business process management (BPM), according to Gartner, is any method used to discover, model, analyze, measure, improve, or optimize these business processes. You typically use BPM software (BPMS) or a software suite to run your company’s business process management.
Business process management systems began as standalone workflow tools with limited, point-point integration. They have since evolved into whole BPM Suites that can orchestrate processes consisting of both human and software distributed across various systems.
Today, most BPMS come as Intelligence Business Process Management Suites, which combine business process management with analytics and process intelligence. These suites usually come with digital process automation (DPA), which came as a result of several recent high technology achievements.
Digital Process Automation
While business processes management makes managing business rules, data, and human decisions simpler, Business process management systems are complex to implement and maintain, requiring large capital investments, strong business commitment and operating resources.
With recent technological achievements and new cultures - such as social media -, organizations should rethink their existing business models and explore new and improved ones. One way to do this is which digital process automation, which uses technology to automate process execution. This reduces costs while making your business more flexible.
All modern BPS systems come with the following features.
Cloud based BPM (BPMaaS and iPaaS)
According to recent studies, although on-site BPM increased over the last decade, cloud based BPMs continue to grow in popularity and forecasted to reach 23B USD by 2024 (more than triple compared to current figures). Additionally, businesses are turning to the cloud to increase their modular scaling, while reducing the high resource utilization and complexity of their BPM systems and services.
Internet of Things
Every device should communicate with its environment and drive or initiate business processes.
Analysis of input data should provide intelligent outcome with feedback to machine learning environments.
Less code means more adaptable to changes. But what is low code? There are low code platforms that actually use graphic tools to generate code, finally ending to rigid environments. The alternative is responsive environments, where each component of a virtual model constitutes metadata. Thus, changes to model, implies changes to component’s metadata with the underlying code, in turn, gets modified to reflect the changes
Intelligent processes and Artificial Intelligence
This is the cornerstone of digital transformation and digital process automation. The usage can be illustrated with the following example:
In a “traditional” process, an e-commerce customer buys a product by submitting an eform. If payment is authorized, the process creates a sales order, then confirms it with the customer before collecting customer feedback at the end.
The following figure adds a task called “Credit Risk”, which analyzes the customer’s credit risk, regardless if the payment is authorized or not. The analysis may performed using a model that rates customers, based on demographic criteria (country, sex, age etc) and credit attitude. As new orders are posted and paid, model data may change, and in the future the analysis for the same customer may give different result.
Moreover someone can automate the process with a natural language processing bot for voice ordering and a digital sensor for measuring inventories. Thus, providing product availability at site level.
An Artificial Intelligence system is a software module that, for a given input, can perceive the environment it operates to provide the optimum real-time solution by its own. Note that an intelligent system may provide different output for the same input in different timestamps.
There are several types of Artificial Intelligence systems, with the following currently be the most interesting:
- They base decisionson a set of business rules (bots)
- They use recommendation engines with content-based or collaborating filtering
- They may be based on neural network with applied probability theories. These are not distinct since bots or recommendation engines may be based on probabilistic analysis.
Based on the system type, an artificial intelligence system may be more or less deterministically defined or chaotic in nature, depending on the input data structure and the system’s machine learning capabilities.
Consider the case of a rule based system (bot) that takes as input a customer, a product, and any customer approvals, and outputs the price of the product. There is no space for machine learning in this system. On the other hand, consider the case of a voice recognition bot. In this case, this there is plenty of space for machine learning (eg for adopting new languages, pronunciations, words etc) as someone can see in the scottish elevator example. In any case, the final decision depends on the system itself and the environment it operates.
Rule based systems (bots)
Rule-based systems, -often called bots- store and manipulate knowledge to interpret information in an useful way. They consists of
- A rule list
- An engine that receives input and takes action based on comparison with rule list
- A working memory where rules are uploaded.
- An input, which can be either automated or manual (human)
Rule based systems (bots) may rely purely on logic, or they may be able to learn. In the first case, systems take data and predefined logic as input and provide a more static rule-based and not generalized output. Meanwhile, the second case rules are dynamic based on historical data and real time analytics and create generalized outputs based on probabilities.
One very popular class of bots are chatbots that communicate via instant messaging (IM) , internet relay chat (IRC), or a web interface. They allow people to ask questions and receive appropriate answers.
Recommendation Engines (or systems) is a subclass of information filtering system that predicts the user preference for an item. It appears as an intelligent and sophisticated salesman who knows the customer and can make intelligent decisions about what should benefit the customer most.
Although recommendation engines started in the retail industry , they are currently used in several industries. A few use examples include:
- Customer who bought this also bought... (www.amazon.com)
- Up next (Recommended videos)... (www.youtube.com )
- Other movies you may enjoy... (www.netflix.com )
- People you may know... (www.linkedIn.com)
According to studies, 35% of Amazon total sales and 78% of Netflix total watches are based on recommendations.
There are several types recommendation engines depending on their filtering algorithms such as collaborative filtering, content-based filtering, demographic filtering, etc., with the first two (and the combination of them) being the most widely used.
In the Collaborative filtering, the logic is based on the customer past purchases or searches, while recommendations are derived from past purchases or searches of similar customers
Collaborative filtering produces accurate recommendations only if there is a large user base -so called cold start-. Cold start plus the required high computing resources are the main challenges for collaborative filtering
In the content-based filtering, the logic compares various candidate items with items previously rated by the user -or currently examined, resulting with the best matches as recommendations.
The major challenge to content-based filtering is that there is no knowledge transfer between different content types. For example for the same user, recommendations for news articles don't necessarily produce recommendations for movies.
Hybrid recommendation engines that combine features from both collaborative and content-based systems are often more effective. You can implement them by:
- Making separate predictions and then merge the results
- Adding content-based capabilities to collaborative filtering (and vice versa)
Despite their wide current use,, there is a large space for recommendation engine research and evolution , in industries such as financial services, health services, etc..
Probabilistic machine learning systems
Machine learning uses statistics to allow computer systems to “learn” (i.e. progressively improve performance on specific tasks) without being explicitly programmed.
From the commercial point of view, probabilistic machine learning systems are known also as predictive analysis systems, with a wide range of usage in speech and language technologies, marketing personalization, computer vision, financial prediction and automated trading, self-driving cars, etc..
According to Forbes, machine learning industry is continuously developed with high rates:
- Patents grow with 34% rate from 2013 and on
- Spending will grow from 12B$ in 2017 to 57.6B$ in 2021
- Implementations and pilots will doubled in 2018 compared to 2017 and doubled again in 2020
Additionally, other studies show that healthcare, education, marketing, financial services and transportation industry business landscape is going to change due to machine learning.
Technically speaking, probabilistic machine learning systems model data in order to evaluate input:
- A model describes data than can be observe from a system for a given input
- Applying probabilities we can estimate noise and uncertainty associated
- Applying uncertainty techniques, we can infer unknown quantities, adapt models make predictions and learn from data
Thus for a given input the following process apply:
Artificial intelligence and machine learning, is a promising industry. Digitally automated processes make use of intelligent systems and provide the initiative for artificial intelligence further development. These systems will increase the rate of adoption of cloud based solutions, with low-code and adaptability features to become essential.