On-Premise Vs. the Cloud

However, the on-premise technology model has serious issues as well. One of the biggest is the cost, which often involves large up-front capital expenses. Then there is the ongoing need for maintenance, upgrades, and monitoring. All in all, it means that the IT department may be spending valuable time on noncore activities. And finally, the use of point applications like Excel can lead to a fragmented environment, in which it becomes difficult to centralize data because there are so many files spread across the organization.

Because of all this, companies have been looking at another approach – that is, cloud computing. The interesting thing is that this has been around since the late 1950s, when computer researcher John McCarthy invented time sharing that allowed multiple users to access mainframes. This innovation would eventually result in the emergence of the Internet.

But the pervasive adoption of the PC in the 1980s and 1990s would establish the on-premise model as the preferred approach. After all, the Internet was still fairly archaic and not in wide-spread use (it was mostly for the military, universities, and large businesses). Thus, with PCs, a company could network them together to enable collaboration and sharing of data and other resources.

But as the Internet became more robust, there was a move to so-called cloud computing. One of the first business applications in this industry was developed by Salesforce.com, which made it possible for users to use the software through a browser. In the book Behind the Cloud, written by the company’s cofounder and CEO Marc Benioff, he writes: “I saw an opportunity to deliver business software applications in a new way. My vision was to make software easier to purchase, simpler to use, and more democratic without the complexities of installation, maintenance, and constant upgrades. Rather than selling multimillion dollar CD-ROM software packages that took six to eighteen months for companies to install and required hefty investments in hardware and networking, we would sell Software-as-a-Service through a model known as cloud computing. Companies could pay per-user, per-month fees for the services they used, and those services would be delivered to them immediately via the Internet, in the cloud.”²

The vision was bold and not easy to pull off. But as of today, Salesforce.com is the dominant cloud applications company. In fiscal 2020, the company posted revenues of more than $13 billion and the market cap was $131 billion.

Regardless, the fact remains that the technology continues to gain traction. According to a report from Gartner, the market for public cloud services is forecasted to jump from $214.3 billion in 2019 to $331.2 in 2022.

Here’s what Gartner’s vice president of research had to say: “Cloud services are definitely shaking up the industry. At Gartner, we know of no vendor or service provider today whose business model offerings and revenue growth are not influenced by the increasing adoption of cloud-first strategies in organizations. What we see now is only the beginning, though. Through 2022, Gartner projects the market size and growth of the cloud services industry at nearly three time the growth of overall IT services.”³

Besides the impact of Salesforce.com and other cloud applications companies, another critical development was Amazon.com’s AWS platform. Launched in 2005, this allowed any company to build their own cloud-native applications. The business has actually become much bigger than Salesforce.com’s as revenues are expected to exceed $30 billion in 2019. To get a sense of the strategic importance of AWS, Slack is expected to spend $250 million on it for the next five years and Pinterest plans to shell out a hefty $750 million during a six-year period.⁴

AWS essentially handles the complex administrative and infrastructure requirements like storage, security, compute, database access, content delivery, developer tools, deployment, IoT (Internet of Things), and analytics (there are currently more than 165 services). This means the development of applications can be much quicker. The costs are generally lower and the fees are based on a per-use basis.

With AWS, other mega tech firms were caught off guard and scrambled to develop their own cloud platforms. The two most common ones include Microsoft’s Azure and Google Cloud. In fact, many companies often use two or more of these in order to provide for redundancy (this is known as a multi-cloud strategy).

As for RPA, the cloud has different implications and impacts. One is that a platform needs to deal with complex distributed applications, which can be difficult if a company develops custom programs on a cloud service.

What’s more, most RPA platforms actually started as on-premise software and generally did not transition to the cloud until recently. This does seem odd as cloud computing has been around for a while and appears to be the default approach for software companies. But developing cloud-native systems is not easy for RPA as there needs to be deep hooks across many applications and environments.

In some cases, an on-premise RPA system may be loaded onto a cloud service like AWS. While there are benefits with this, it is not cloud native. This is because you will still need to upgrade and maintain the software.

And then, we will take a look at how to use flowcharts, which are quite useful in an RPA implementation.

Web Technology

The mastermind of the development of the World Wide Web – which involved the use of hyperlinks to navigate web pages – was a British scientist, Tim Berners-Lee. He accomplished this in 1990. Although, it would not be until the mid-1990s, with the launch of the Netscape browser, that the Internet revolution was ignited.

At the core of this was HTML or hypertext markup language, which was a set of commands and tags to display text, show colors, and present graphics. A key was that the system was fairly easy to learn and use, which helped to accelerate the number of web sites.

Yes, this means that the text is bold.

Yet HTML would ultimately be too simple. So there emerged other systems to provide even richer experiences, such as with CSS (Cascading Style Sheets, which provides for borders, shadows, and animations) and JavaScript (this makes it possible to have sophisticated interactivity, say, with the use of forms or calculations).

No doubt, RPA must deal with such systems to work effectively. This means it will have to take actions like identify the commands and tags so as to automate tasks.

Programming Languages and Low Code

A programming language allows you to instruct a computer to take actions. The commands generally use ordinary words like IF, Do, While, and Then. But there can still be lots of complexity, especially with languages that use advanced concepts like object-oriented programming. Some of the most popular languages today include Python, Java, C++, C#, and Ruby.

To use an RPA system, you have to use some code – but it’s not particularly difficult. It’s actually known as low code. As the name implies, it is about using minimal manual input. For example, an RPA system has tools like drag-and-drop and visualizations to create a bot.

This is not to imply that low code does not need some training. To do low code correctly, you will need to understand certain types of workflows and approaches.

OCR (Optical Character Recognition)

A key feature for an RPA platform is OCR (Optical Character Recognition), a technology that has actually been around for decades. It has two parts: a document scanner (which could even be something like your smartphone) and software that recognizes text. In other words, with OCR, you can scan an image, PDF, or even handwritten documents – and the text will be recognized. This makes it possible to manipulate the text, such as by transferring it onto a form or updating a database.

Because of all this, OCR in the early years was far from accurate. But over time, with the advances in AI algorithms, fuzzy logic, and more powerful hardware, the technology has seen great strides in accuracy rates, which can be close to 100%.

Then how does this technology help with RPA? One way is with recoding a person’s actions while working on an application. The OCR can better capture the workflows by recognizing words and other visuals on the screen. So, even if there is a change of the location of these items, the RPA system can still identify them.

Something else: Automation involves large numbers of documents. Thus, OCR will greatly improve the processing. An example of this would be processing a loan. With OCR, a document will use OCR to extract information about a person’s financial background, the information about the property, and any other financial details. After this, the RPA system will apply the workflows and tasks to process the loan, say, with applying various rules and sending documents to different departments and regulatory agencies.

Finally, even though RPA systems may have their own OCR, this may not necessarily be enough. Some industries and segments, such as healthcare, insurance, government and banking, still rely heavily on handwritten forms – all of which can be time-consuming and costly.

But there are OCR systems that can help out, such as HyperScience. The software leverages sophisticated machine learning (ML) technology to quickly and accurately extract the information (understanding cursive writing, for example). But there are other capabilities like detecting fields on invoices and handling data reconciliation. Consider that HyperScience can automate up to 90% of the processing.⁵

Databases

At the heart of most applications is a database, which stores data that can be searched and updated. This is usually done by putting the information in tables (i.e., rows and columns of information).

The dominant form is the relational database – developed in 1970 by IBM researcher E. F. Codd – that uses structured data. To interact with this, there is a scripting language called SQL (Structured Query Language), which was relatively easy to learn.

It was not until the late 1970s that relational databases were commercialized, led by the pioneering efforts of Oracle. Then came a smattering of start-ups to seize the opportunity. But by the late 1980s, Oracle and SAP dominated the market for the enterprise (Microsoft would essentially be the standard for the mid-market).

While relational databases proved to be quite effective, there were still some nagging issues. Perhaps the biggest was data sprawl. This describes when there is a growing number of tables that get proliferated across the organization. This often makes it extremely difficult to centralize the data, which can make it challenging to get a holistic view.

Another problem was that relational databases were not cheap. And as new technologies came on the scene, such as cloud computing and real-time mobile applications, it became more difficult to process the data.

Given all this, there emerged various alternatives to relational databases. For example, there was the data warehouse that started as an open source project in the late 1990s from Doug Cutting. The technology would undergo various iterations, resulting in the development of Hadoop. Initially, Yahoo! used this to handle the Big Data demands from its massive digital platforms. Then other major companies, like Facebook and Twitter, adopted Hadoop. The key was that a data warehouse could make it possible to get a 360 degree view of data.

The market has definitely seen lots of growth and change. Companies like Google, Amazon.com, and Microsoft have been investing heavily in data warehouse systems. There are also some fast-growing start-ups, like Snowflake, that are pushing the boundaries of innovation.

In the meantime, there have been new approaches that have gone against the model for relational databases. They include offerings like MySQL (which is now owned by Oracle) and PostgreSQL. Yet these systems did not get enough traction in the enterprise.

But there is one next-generation database technology that has done so: NoSQL. It also began as an open source project and saw tremendous growth. As of now, MongoDB has 14,200 customers across 100 countries and there have been over 70 million downloads.⁶

Where relational databases are highly structured, a NoSQL system is quite flexible. It’s based on a document model that can handle huge amounts of data at petabyte scale.

Another major secular trend is the transition of databases to the cloud. According to research from Gartner, about 75% will be migrated.⁷

So why the cloud? For the most part, this should make it easier to allow for improved analytics and AI.

And going forward, there is likely to be much innovation with database technology. Yet relational databases will remain the majority of what companies use – which also means that this will also be what RPA interacts with as well.

APIs (Application Programming Interfaces)

An API – which is the acronym for “application programming interface” – is software that connects two applications. For example, let’s say you want to create a weather app. To get access to the data, you can setup an API, which often is fairly straightforward, such as by putting together a few lines of code to make data requests (say, for the city). By doing this, you will increase the speed of the development.

APIs are pervasive in enterprise environments since they are so effective. They also have different structures. Although, the most common is a REST (representational state transfer) API.

Even if there is an off-the-shelf API available, there are still issues. One is metering, which means that you may be limited to a certain number of requests per day or hour. Or there may be higher pricing. Next, APIs can still have bugs and glitches, especially when in complex IT environments.

Because of the difficulties, RPA has proven to be a very attractive alternative. Again, the development is much easier and there is less of a need for integration. But, interestingly enough, an RPA platform can be a vehicle for delivering advanced APIs within the enterprise.

AI (Artificial Intelligence)

A typical RPA system does not have much AI (Artificial Intelligence). The main reason is that there is a literal carrying out of tasks, which does not require any smart system. But as AI gets more powerful and accessible, RPA will increasingly start to use this powerful technology – which should greatly enhance the outcomes.

But before looking deeper at this – we will cover AI in various parts of this book – it’s a good idea to get a backgrounder on the technology. Like RPA, this has also been the subject of much hype. It’s not uncommon to read blogs and news articles on how AI will ultimately conquer disease, help with climate change, and even predict earthquakes! While such things could be achievable, it does seem far-fetched that they will happen any time soon. Keep in mind that AI is still in the nascent stages, even though it has been around since the 1950s!

OK then, what is AI? Well, a good way to think about it is as follows: It’s software that ingests large amounts of data that is processed with sophisticated algorithms that help answer questions, detect patterns, or learn. Interestingly enough, AI is actually made up of a variety of subcategories (Figure 2-1 shows a visual of this):

• Machine Learning : This is where a computer can learn and improve by processing data without having to be explicitly programmed. Machine learning is actually one of the oldest forms of AI and uses traditional statistical methods like k-nearest neighbor (k-NN) and the naive Bayes classifier.

• Deep Learning : While the roots of this go back to the 1960s, the technology was mostly an academic pursuit. It wasn’t until about a decade ago that deep learning became a major force in AI. Some of the important factors for this included the enormous growth in data, the use of GPUs (graphics processing units) that provided for ultrafast parallel processing, and innovation in techniques like backpropagation. Deep learning is about using so-called neural networks – such as recurrent neural networks (RNNs), convolutional neural networks (CNNs), and generative adversarial networks (GANs) – to find patterns that humans often cannot detect.

• NLP (natural language processing): This is AI that helps understand conversations. The most notable examples of this include Siri, Cortana, and Alexa. But there are also many chatbots that focus on specific uses cases (say, with providing medical advice).

When it comes to AI, the excitement is often with the whiz-bang algorithms. But the reality is that this is often a small part of building a model. Keep in mind that it is absolutely essential to have high-quality data. And this usually means spending much time cleaning it up as well as weeding out outliers, which can be a tedious manual process.

Another confusion about AI is, well, the impact from science fiction movies. These portrayals are about strong AI or AGI (artificial general intelligence), in which computers act like humans. Systems can think, engage in conversion, and walk around.

But with today’s technology, it’s only about weak AI. This means that the applications are focused on narrow areas, such as helping predict weather or mining insights about customers.

Then so what about AI and RPA? It’s certainly a major focus right now in the industry and we’ll see many developments in the years ahead.

For the most part, RPA uses structured data. However, this represents about 30% of what’s available in a typical organization. But with AI, an RPA system will likely be much more effective since it will be better able to process unstructured data. For example, many companies have their own approaches to writing invoices. Because of this, an employee would likely have to spend much time interpreting and processing them. But AI would be able to learn from the invoices and come up with its own rules and tasks.

Furthermore, there are other potential benefits of the technology: judgement, the use of reasoning, and the detection of highly complex patterns. With these, the automation will be greatly enhanced, helping with things like detecting fraud.

Cognitive Automation

In the discussion about RPA, you may hear the term “cognitive automation” (in the first chapter, we called this IPA and was referred to as one of the flavors of RPA). It’s often confused with AI – but the two concepts have different meanings.

Consider cognitive automation to be an application of AI, actually. First of all, it is mostly focused on automation of the workplace or processes in business. Next, cognitive automation uses a combination of technologies like speech recognition and NLP. By doing this, the goal is to replicate human actions as best as possible, such as by analyzing patterns of workers and then finding patterns and correlations.

Something else: Unlike other forms of AI, cognitive automation is usually effective with the use of much less data. There may also be not as much reliance on highly technical talent, such as data scientists.

Agile, Scrum, Kanban, and Waterfall

Software development can be quite complex. Besides the technical aspects, there is a need to manage a team whose members may be located in different countries. In the meantime, technologies continue to evolve. What’s often even harder is maintaining a software system as there is usually a need to add capabilities and upgrade the underlying technologies.

Even some of the world’s most talented software companies have faced monumental challenges. Just look at Microsoft. The company’s Vista operating system took five years to develop. The reason for this was that Microsoft had too many silos within its organization and did not have sufficient collaboration. The result was that the development was glacially slow and there were persistent mishaps. The irony was that one team may have created an effective piece of code but it did not work with the complete system.¹⁰

In today’s world, software development has become even more difficult because of the emergence of new platforms like the cloud and the hybrid cloud. This is why it’s important to look at software management approaches.

One is called Agile, which was created back in the 1990s (a big part of this was the publication of the Manifesto for Agile Software Development). The focus of this was to allow for incremental and iterative development, which begins with a detailed plan. This also requires much communication across the teams and should involve people from the business side of the organization.

Nowadays, Agile has gotten easier because of the emergence of sophisticated technologies like Slack and Zoom that help with collaboration. “Over the past few years, my volume of e-mail has declined substantially,” said Tim Tully, who is the chief technology officer at Splunk. “The main reason is that I mostly use Slack with my developer teams.”¹¹

DevOps

DevOps has emerged as a critical part of a company’s digital transformation. The “Dev” part of the word is actually more than just about coding software. It also refers to the complete application process (such as with project management and quality assurance or QA). As for “Ops,” it is another broad term, which encompasses system engineers and administrators as well as database administrators, network engineers, security experts, and operations staff.

For the most part, DevOps has come about because of some major trends in IT. One is the use of agile development approaches (this was discussed earlier in the chapter). Next is the realization that organizations need to combine technical and operational staff when introducing new technologies and innovations. And finally, DevOps has proven effective in working with cloud computing environments.

According to Atlassian, which is a leading developer of DevOps tools: “The bad news is that DevOps isn’t magic, and transformations don’t happen overnight. The good news is that you don’t have to wait for upper management to roll out a large-scale initiative. By understanding the value of DevOps and making small, incremental changes, your team can embark on the DevOps journey right away.”¹²

Flowcharts

Since an essential part of RPA is understanding workflows and systems, the use of flowcharts is common. It’s usually at the core of the software application.

With a flowchart, you can both sketch out the existing workflows of a department. And then from here, you can brainstorm ways of improving them. Then you can use the flowchart to design a bot for the automation.

The flowchart is relatively simple to use and it also provides a quick visual way to understand what you are dealing with. As the old saying goes, a picture is worth a thousand words.

So let’s take a look at some of the basics:

• Terminator: This is a rectangle with rounded corners and is used to start and end the process, as seen in Figure 2-2.

• Process: This is represented by a rectangle. With this, there is only one next step in the process. Figure 2-3 shows an example:

• Decision: This is a square symbol that is at an angle. There will be at least two possible paths. Figure 2-4 is an example:

Conclusion

For the most part, you just need a general understanding of the concepts highlighted in this chapter. It’s also good to remember that no technology is a silver bullet. They all have their plusses and minuses, regardless of what you might hear in the media!

So as for the next chapter, we’ll take a look at the steps in implementing RPA in your organization. As you’ll see, there is quite a bit of work that needs to be done before creating your first bot.

Key Takeaways