Partial automation: The key to easing fears and pushing digital transformation

Many enterprises trying to reach the highest levels of digital transformation are facing a problem: they aren’t adopting autonomous operations and artificial intelligence for IT operations (AIOps). While more than half (59 percent) of organizations consider themselves digital transformation adopters or even leaders, a mere 15 percent are implementing automated processes at the same level.

This gap is problematic for enterprises striving to be digital leaders. After all, part of digital transformation’s promise is using data to increase agility, maximize emerging opportunities, provide personalized experiences and, importantly, guarantee a business’s apps and digital services are continuously available.

But human teams can’t deliver on this promise without automation. Because systems generate massive amounts of data (with data velocity continuing to skyrocket), human operators can only catch so much. And businesses will never be able to maximize their potential without automation tools streamlining and simplifying their workflows.

Overlooking autonomous operations and AIOps is a tremendous missed opportunity -- and it shows in company growth. Thirty-six percent of organizations with the highest level of automated operations expect their revenue to increase by 15 percent or more. Only 10 percent with a low rate of automated processes expect the same.

Despite the promise of automation, a vast majority (90 percent) of organizations still see IT staff toil away on manual, repetitive tasks that could be automated. This toil holds the entire business back. By tasking IT teams to perform repetitive tasks rather than push forward complex and creative initiatives, companies waste resources and slow their own transformation efforts.

If autonomous operations and AIOps can accelerate what just about every corporation is after -- digital transformation -- why aren’t more enterprises embracing these technologies? Why are businesses still exhausting their human technology talent with unnecessary manual interventions? And where can they start on their journey to full automation?

Trust issues don’t need to shut down automation

A combination of factors has led to automation’s lagging adoption, but one of the major reasons is plain nervousness. People are largely uneasy about full automation, still wanting some degree of oversight. This creates tension between the desire for full automation and an unwillingness to let technology operate completely alone.

Even in a cutting-edge, streamlined CI/CD kind of world, some DevOps practitioners desire a "yes, please proceed" button.

This desire isn’t baseless. A slight mistake can escalate to an absolute catastrophe before a human even finds out what’s going on. Consequently, many enterprise IT functions prefer a human-in-the-middle approach to automation, where a human operator interacts with the automated technologies. This partial automation can reap huge benefits and build a user’s confidence, paving the way for full automation.

Before allowing a tool to operate without supervision, teams must first prove that it works. For example, AIOps solutions support the continuous availability of an enterprise’s apps and services by automating incident detection, root cause identification and incident response. Some of these tools feature semi-supervised machine learning (ML) capabilities where users set up their own incident clustering algorithms. While there’s no training involved, the human operator has a role in explaining how it wants the technology to work.

AIOps also offers supervised ML techniques that act as a reward system. When the system attempts to make a decision, operators can tag the data as "good" or "bad." Through these reinforcements, the system learns positive behaviors that it should replicate and negative behaviors it should avoid.

Human operators can also collaborate with AIOps based on the technology’s confidence level. It works like this: a system creates a recommended outcome and then rates how confident it is in that decision. If the confidence level is only around 50 percent, that’s inviting the user to take a closer look at the decision, essentially asking the human to double-check the work. 

Still, sometimes customer trust comes from transparency. And the AIOps market is responding. Select AIOps solutions let users peek behind the curtain to see how the technology correlates underlying alerts into an incident. If teams don’t like the correlation definition, they can change it. But more importantly, they understand how the automated technology works.

The diminishing cost of change and rise of flexibility

Historically, manual processes had to happen very, very frequently to justify an investment in automation. Organizations feared ripping out infrastructures and replacing them with an expensive automated process, only for the technology to be obsolete in six months.

But automation has evolved. Technologies like AIOps now have the flexibility to deal with change, holding up to the ephemeral nature of modern IT infrastructures.

Natural language processing (NLP) is one feature driving flexibility. NLP capabilities allow AIOps tools to understand the nuances and elasticity of language. If a bank’s distributed network needs to understand which cash machine isn’t working, NLP looks at addresses just like a human would. It can identify "North First Street" as the same address as "N. 1st St," instead of listing the address twice.

Understanding such structure makes the technology more flexible. And because the software adapts to changing infrastructures without reconfiguration, it’s set up for long-term operation and sustained ROI, which should help the economics of and motivation for automation adoption.

The future of enterprise automation

In the early days of containers, there was a concept of pets versus cattle. A big computer deployed an expensive application -- a pet. IT teams needed it to stay healthy and live a long time.

With today’s distributed architectures, that team no longer owns the computer and could even remove and replace a computer without impacting system performance. It’s cattle.

In our cattle world, improving upon a problem often means completely destroying it and replacing it with something else. But repeating this process can create a tremendous amount of churn (read: wasted resources).

There needs to be a balance. IT teams don’t want pets but should be nice to the cattle as it’s expensive to constantly replace them. That’s how enterprises serious about digital transformation should use automation.

Modern IT functions should move beyond automation that only points out that the CPU is spiking on a resource with the mindset of "let’s eliminate and replace it." A more sophisticated automated algorithm could suggest migrating the workload somewhere else to see if the resource recovers, or, based on past activity, predicts the error condition will self-recover faster than the remediation action, saving downtime and money.

Organizations don’t just flip on full automation from day one. But using partial automation can engender trust in the technology and help enterprises build a case for further investment. Even these baby steps toward full automation can help advance an enterprise’s mission-critical digital transformation strategy and gain a competitive edge in an ever more competitive world.

Photo Credit: Sashkin/Shutterstock

As Chief Evangelist, Richard Whitehead brings a keen sense of what is required to build transformational solutions. He’s a DevOps Institute Ambassador, and serves on the DevNetwork AI/ML Advisory Board. A former CTO, and Technology VP, Richard brought new technologies to market, and was responsible for strategy, partnerships and product research. Richard served on Splunk’s Technology Advisory Board through their Series A, and more recently co-chaired the ONUG Monitoring & Observability Working Group. Richard holds three patents, and is considered dangerous with JavaScript.