Updated: Jan 31, 2021
Digital automation is a hot buzzword these days. While most industries were negatively impacted throughout 2020, the automation industry has grown. Automation companies have raised funds and had exits that added up to $2.2B. This figure does not even include UIPath’s raise of $225M in July 2020, and Microsoft acquiring Softmotive, and Google acquiring AppSheet for undisclosed amounts. Automating workers’ tasks has been happening for decades, but in 2020 automation has experienced explosive growth. There is a multitude of new tools available today that can automate business processes in days or weeks instead of months.
According to a 2018 Stripe survey, developers lose 31% due to inefficiencies from technical debt and bad code. From my own experience working at a large investment bank for over 10 years, building automation solutions from scratch in a large enterprise environment can take 6-12 months and cost hundreds of thousands of dollars. Automation platforms can bypass the complicated development process, and achieve lower error rates, providing detailed audit logs for transactions, and ultimately allowing employees to focus on complicated value driven tasks while simultaneously achieving higher customer satisfaction by reducing response time.
So how does a company begin its automation journey? With so many tools out there, it’s hard to decide where to begin. In this article, we’ll show you how.
Identifying Processes to Automate
The first step is to identify the processes that you would like to automate. Ideally, these processes need to be repetitive, high volume and have little variation. Often, ETL (Extract Transform Load) processes are a great candidate since they follow definitive steps: taking input from one application or document, cleaning the data or augmenting it with data from another source, and then loading it into another application or system.
Low Code/No Code Automation
The first place to look when selecting an automation tool is at low code/no-code platforms that allow creating integrations between different applications. A good example is Zapier for consumers and small businesses. There are also enterprise friendly alternatives like MuleSoft, Tray.io, and Workato. The benefit of these tools is that you can bypass IT involvement and have non-technical business or operations staff build processes with drag and drop functionality. These tools are good for transferring data between different applications. For example, a salesperson can create a workflow that monitors for new leads coming in from an online source, once a new lead comes in the workflow sends a response email, automatically adding the lead to Salesforce, and then notifying the team through Slack. A more technical example is automatically moving specific types of ServiceNow tickets to JIRA for bug fixes and sending an email notification to the Support and Engineering teams. Low Code/No Code platforms can integrate hundreds of modern applications by utilizing their APIs. However, these platforms can have a limited ability to integrate with older applications or custom apps that your firm developed, where APIs may not be available.
Robotic Process Automation (RPA)
If an application does not have an API, the only way to interact with it is through the front end. A user needs to open the application or browser and interact with it using the keyboard and mouse. Robotic Process Automation (RPA) platforms were designed for this scenario. RPA platforms can run on a user’s machine or dedicated windows terminal, and simulate a human’s activity such as mouse movements and keyboard typing. Some of the largest RPA platforms are UIPath, Automation Anywhere, Blue Prism, and WorkFusion. Many RPA platforms have their own low code applications that allow users to record mouse clicks and keyboard typing which could be good for simple tasks. To build enterprise-level process automation a dedicated IT staff trained on the platform will be needed to professionally handle all the exceptions that might happen with this type of automation.
Unstructured Data Extraction
RPA platforms do an excellent job extracting data from structured documents such as standard PDF forms that have very little or no variation. However, RPA platforms run into trouble when you need to extract data from unstructured data sources like emails or free text documents. According to Forbes, unstructured data make up to 90% of all data in large enterprises. Machine Learning (ML) can classify and extract unstructured data from such sources. Major RPA platforms have their own AI solutions which they call Intelligent Automation, but most of them are complicated and hard to use. New companies like Alkymi, Hyperscience, and Rossum recently came into the scene, and are focused on solving this problem with pre-built models that can be used out of the box. They also can build customized models in days versus weeks for proprietary documents.
Specialized Automation Service
An interesting new development is Automation as a Service companies like Ocrolus and Roots Automation, whose products don’t require building new models, installation, or IT involvement. This is great for companies that are focused on getting results, without getting too deep into the details of how they are achieved. Ocrolus focuses on loan application documents where most data are standardized, and utilize machine learning, logic, and humans in the loop to achieve 99% automation. Roots Automation took a different approach - they see their solution as a digital coworker learning and working alongside employees making their lives easier.
Cloud Process Automation
The three biggest cloud platforms - Amazon AWS, Microsoft Azure and Google Cloud - have their own offerings of automation solutions. These products may not be as powerful as some of the tools mentioned above, but they are worth exploring if you are using one of the cloud platforms. Microsoft acquired Softomotive in May 2020 and made it available as part of its Power Platform, Google acquired AppSheet in January 2020 as the no-code integration platform added it to Google Cloud, and Amazon released its proprietary no-code integration platform Honeycode in June 2020. If you are on one of these public cloud platforms, the above tools can integrate more easily into your current infrastructure. They also have their own ML solutions which could be used to automate complex cognitive tasks.
In conclusion, there are many different automation solutions out there. The first step to get started is to identify the process that you want to automate. Then you can select the best automation solution based on cost, percentage of automation, complexity, and maintainability.
If you are considering setting up business process automation in your organization and need help getting started, contact us for a consultation.