Automating business processes with human-like Artificial Intelligence

Practically all business processes can be automated using Artificial Intelligence (AI). However, sometimes it is unreasonable as some processes would be so expensive to automate that the costs would exceed the benefits. The business processes that are easy to automate relate to financial management, data protection, customer service, and quality deviation reports.

In this blog, you will find three concrete cases of the potential of AI, as well as some thoughts on the following questions: What new automation opportunities will human-like AI create? How to find processes in your organization that are reasonable to automate? What should you consider today to make automation easier in the future?

Human-like AI can be useful in processes that require human understanding

RPA solutions, for example, focus on taking information that is known and moving it somewhere, automating the data entry. The problem with the solutions is, that the key information needs to be recognized and picked up by a human first. The human-like AI is able to pick up information and understand what is being said in a document, providing the same information a human worker would. For example, handling incoming sales orders or medical records are processes that can be automated with human-like AI. In a nutshell, human-like AI creates value in those processes where human needs to think, recognize something and act upon it.

How to find processes in your organization that can be automated with human-like AI – three concrete examples

The first step is to open your own mind and think outside the box. If the process is consuming a lot of manual work and we are talking about a big amount of documentation, there might be a possibility for automation.

Case 1 – Automating the processing of sales orders that are not in a specific format

Normally, if the document contains something that needs human understanding, the handling process is considered impossible to be automated. For example, some organizations receive a huge amount of orders from customers by e-mail, and it is a common misconception that you cannot automate this type of order handling because you need a human to read and understand what is being ordered. However, with cognitive AI, it is possible.

Case 2 – Correcting accounting that is based on inaccurate data

The second example relates to financial management. Sometimes people think that they cannot automate their incoming invoice processing because the accounting is not accurate enough and therefore there is not sufficient training data. This is partially true, but one great benefit of clever AI is that we can easily find the things that are not correct in accounting and exclude them from the training data. This means that even with insufficient data or data that is not accurate enough, there are many great things we can do with cognitive AI to make data more robust, accurate and more usable.

Case 3 – Privacy data screening with human-centric AI

Another good example is privacy screening, where you have terabytes of data. If a lawyer goes through the data to see what kind of privacy data is stored and what is the legal basis for storing this information, it will take about 12 years per terabyte for a lawyer to go through it. With human-centric AI, this amount of data can be cleared in a month.

What are the enabling factors for future automation that you need to think of now?

When setting up the ERPs and processes, businesses should make sure to set them up in a way that everything on the ERP can be traced to the source. For example, actions related to a sales process should be able to be traced continuously to the sales order, so that they can be programmatically combined to a current training data set. If this tracing cannot be done electronically, it has to be done manually which can cause unnecessary costs. In other words, making small adjustments in your existing processes cost you now, but it can save you money in the future.