Automation or AI

AI is often misunderstood and frequently mistaken for automation.

When discussing AI, one usually thinks about having the AI perform tasks that would otherwise be done manually. However, AI is not automation, although it can be combined with tools to enable such capabilities.

Automation tools are designed to perform tasks automatically, eliminating the need for human intervention. Nevertheless, a human would need to set up such a process. For example, an individual who receives a daily email from a particular person with files attached may decide to set up an automated process instead of downloading the files every day and storing them in a designated location.

An automation process is defined by different triggers and actions. To determine such a process, one must first understand and scope out the problem that the automation will solve for the business.

In other words, automation is not done automatically or generated; instead, it is constructed, built, and developed.

The image below illustrates a 6-step cycle that begins with a trigger and concludes with an update.

Automated Process

Step 1: Trigger

A trigger indicates what the tool should "look for, such as an email from apeaceofamind@apeaceofamind.apam. Perhaps that would be enough, as all emails from this sender would follow the same format, text and attachments. However, if the emails contain different contents, then additional criteria would include the title or subject of the email, as well as words contained in the message that can be relied upon. This is an important detail. If the sender's email/chat name, title, subject, or content changes or fluctuates, the trigger will need to be adapted accordingly. Some tools require the exact wording, while others can look for similar words. It all depends on the tool.

Step 2: Action

Once the tool receives information that the process has been triggered, an action occurs. Whether it is to schedule a meeting, to store attachments, forward an email, or to update a database, it all depends on the end goal, which should reflect the BUSINESS USE CASE. It is essential for the individual setting up the process to understand what it will replace. Rigorous discussions and conversations are necessary because, often, a task may seem repetitive and mundane, but it may involve several minor components.

Step 3: Reaction

A reaction is either incorporated with Step 2 or requires a different step altogether. For example, let's assume the trigger is an email with attachments that need to be stored in a folder. The action would be storage, but the reaction could be to send an email confirming that the task has been done. It could be an update of another file, since the folder has now been updated.

The reaction can also be the end goal, which in the above diagram is marked as Step 6, but in truth, it all depends on the USE CASE. The above diagram is an illustrative example and not representative of all processes. Steps 4-5 can be part of the process, may not be needed or may expand to many more steps. Understanding the process to be replaced will determine the number of steps required.

Step 6: Update

Whether to set up a meeting, send an automated chat message, email or refresh a dashboard, I always recommend having an update step to confirm that the process has been completed. Many times, individuals will opt for an email, "Process ABC completed successfully". However, this may lead to many, many emails, and while the idea makes sense, if the user does not check them but lets the email go to a folder never to be looked at until "spring cleaning", it defies the purpose. Instead, the business stakeholder shall determine what the update should look like. I assure you that different individuals will have different preferences.

How does AI come into it?

Many tools can enable automation, but not many people possess the necessary expertise. AI can enable you to set up such a process. It can provide examples of how to set up an automated process, and an individual can interact with AI to refresh the guided steps by adding minor requirements that are often overlooked. AI can also support the trigger stage, where it can compute whether fluctuations in the text are similar to the setup or not. It can also be embedded in the process, albeit with more regulations requiring a human-in-the-loop. Nonetheless, some instances allow for the AI to respond to particular messages in a specific manner and tone, but each trigger is allowed some flexibility. Of course, the parameters and, more importantly, the guardrails will need to be determined to reflect the company's brand, but it is an option.

AI is not automation; however, it enables anyone to learn how to use an automated tool and set it up. I can prompt AI to generate an image, and I can also set up a process to trigger an AI tool to generate an image, but the concept is different. What we want to achieve with AI is different to automation. However, AI can be integrated with automation tools and be part of an automated process.

I expect some will argue that even the AI needs to be triggered in the form of a prompt. Correct with one crucial difference - I know what to expect from an automated tool, but I rarely know what the exact output of the AI will be. Depending on how and why I interact with it, I may know more or less what it should be, but when it is not my area of expertise, I have no concept of judgment whether it is on point or not.

In conclusion, AI enables automation in ways that support anyone in using any tool and automating processes. Rather than seeing this as a replacement of a role, I see this as a displacement of that role. A person who typically sets up these automated processes can focus on growth and explore various ways to support the business. Whilst their day-to-day will change, their expertise will shift. Instead of delivering a project, they will deliver insight, knowledge, and educational skills to enable businesses to leverage their tools as much as possible.