Full Automation Is a Near-impossible Goal: Here’s What to Strive for Instead


Automation adoption and implementation are slower than the enthusiasm surrounding it for several reasons. One is: human capabilities are still essential. So should modern businesses put their sights on full automation? And where can tech leaders start automating for the most effective impact? Richard Whitehead, evangelist in chief and CTO at Moogsoft, shares thoughts and key insights.

Is artificial intelligence really coming for your jobs? Not any time soon, according to the latest research. In fact, while 59% of organizations are considered digital transformation adopters or leaders, only 15% are considered adopters or leaders of automated processesOpens a new window . 

When to Implement Automation — And When You’ve Gone Too Far

Automated processes help human teams with repeated manual tasks and take time away from other activities. Engineers shouldn’t waste time on these repetitive tasks, instead focusing on important, value-creating innovation. Making time for innovation can, in itself, justify automation costs. When making automation investments, tech leaders also must consider: How repetitive is the task you’re looking to automate (think, the “rule of three)? Does the automated process save a significant amount of time? Does the task fail to meet one of the five DevOps ideals without automation? How will the newfound free time be used?

Automation saves time and is also necessary for data-heavy activities, like reaching availability goals. After all, modern systems generate an overwhelming amount of data — too overwhelming for manual processes and too overwhelming for DevOps practitioners and SRE to find and fix incidents. On the other hand, automated solutions, like artificial intelligence for IT Operations (AIOps), can automatically and constantly scan data to detect incidents early in the life cycle. Not only does this solution remove toil, but it also automates the entire incident life cycle, seamlessly moving tasks from machine to human.

While AIOps is an example of humans and machines working together, automation sometimes just doesn’t make financial sense. Architectures are — or should be — constantly changing. If teams get to a point where their environment is so stable that full auto-remediation is implemented and there’s no risk, then they probably aren’t innovating anymore. If they are still innovating, with the speed at which technology keeps evolving, changing automated processes to fit these new innovations will just bring teams back to the original problem: manual work.

This is not to say we shouldn’t try to automate repetitive tasks as much as possible to help DevOps and SRE teams. But leaders must recognize when automation is not productive: with the “black swans” that are unpredictable and near impossible to have a consistent automation strategy. The solution for these issues lies in tools that highlight the problem and uncover additional details that lead to quick remediation.

Learn More: The Importance of People-First Automation—and How to Facilitate It

Why Businesses Hesitate to Automate

There is a lot of hype around automation, and at the pinnacle of that conversation is AIOps. By automatically getting to the root cause of an issue early, AIOps can ensure more uptime for customers and save your team from unnecessary toil. This solution is one of the highest levels of automation. So why isn’t everyone on board?

Automation hype makes people think they must aim for full auto-remediation as the end goal. And that scares people. After all, modern systems are distributed, complex and fragile, so many IT teams — even DevOps practitioners – still want to be the final stop instead of letting automation take full control. We refer to that as “human-in-the-middle” AI.

These hesitations are often rooted in an old-fashioned view of automation. To combat hesitations, leaders require more insight into modern automation.

Hesitant Teams Can Start Small

Many organizations try to take on too much at once when building their automation plan. But, in reality, even small steps toward automation can significantly impact productivity. Businesses new to automated processes should partially automate while keeping humans in charge.

For instance, teams can make big gains by simply automating the incident management triage process. Because collating the information required to make an informed decision requires significant time and effort, teams can automate that process, allowing people to make faster remediation decisions.

Trust is one of the biggest barriers to this kind of automation, so teams should tailor their automation tools to build up that trust. For example, if a team is using an AIOps tool and wants to create specific outcomes for the solution, teams can use machine learning (ML) capabilities to create unique algorithms and track and adjust them as needed. This way, once teams discover how to get their desired outcomes, they are more apt to trust AIOps to do the rest.

Humans build trust with each other through transparency — and the same applies to automation. People often don’t trust what they don’t understand. When searching for an automation tool, tech leaders can ask if the back-end model of the system is available to users. As hesitant IT teams understand the behind-the-scenes technology, they are more likely to trust the technology. 

Another strategy for partial adoption is to become the platform’s teacher. Just as a child would respond to a parent’s feedback, machine learning (ML) can respond to a user’s positive and negative reinforcement. If the user marks an automated decision as “good,” the system will continue that behavior. Similarly, if a user marks a decision as “bad,” the system will not repeat the behavior. Soon, with this reinforcement, systems will learn the desired outcomes for each event.

There are also ways for teams to get feedback from the technology itself. When an automated tool makes a decision on how to handle an incident, for example, the system can ask the human if it made the right call. But not on every decision: the tool can present its confidence level based on the variables leading to the decision. So, if the system presents a high confidence level, humans might not give the system feedback. If the system presents a low confidence level, humans probably want to step in and provide “good” or “bad” feedback.

We will never achieve full automation — and frankly, we shouldn’t aim for it. Instead, tech leaders should focus on the repetitive, manual tasks causing teams to toil. With toil off their plates, teams can focus on value creation, including the innovation and experimentation that will accelerate digital transformation and enable competition in our rapidly evolving digital economy.

Where are you on your journey to full automation? Share with us on FacebookOpens a new window , TwitterOpens a new window , and LinkedInOpens a new window .


Image Source: Shutterstock