Five essential steps to prepare for AIOps
After decades of being conditioned to follow the structured, sequential IT operations practices promoted by the ITIL framework, IT leaders are wising up to the benefits of AIOps.
In the age of rapidly evolving business demands and changing technological landscapes, they’re drawn to the ability to react and resolve IT issues faster and build models to predict when new ones will occur.
They’re enamored of the idea that their IT Ops and DevOps teams can employ machine learning and artificial intelligence to automate time-consuming or repetitive tasks, like alert management and incident remediation, that otherwise tie up their resources. And they’re embracing the concept of analytics tools and applications that can provide actionable insights and greater transparency across their organization without having to staff an army of data scientists.
But even though they’re mentally prepared for this transformation, many companies aren’t physically or culturally prepared to implement any of it — at least not yet. Deploying AIOps isn’t as simple as flipping a switch, and there’s a good bit of preparation that must be done before unleashing an enterprise-altering solution. So, here are five ways to prepare an organization for AIOps.
1. Define how AIOps will be used
As in most endeavors, understanding the “why” of AIOps is at least as important, if not more so, than the “what” or the “how.” That means it’s critical to define specifically what the team needs to accomplish by adopting an AIOps platform.
Perhaps the IT environment has grown exponentially in recent years (as many have) and the team is overwhelmed by the sheer volume and noise of redundant alerts that prevent them from addressing critical issues faster. Maybe the company is plagued by service interruptions and performance degradation because of underutilized capacity. Perhaps they’re driven to overhaul their customer experiences and worry that downtime could sabotage the best of intentions. Or it could be that there’s simply no visibility into the root cause of an issue and no way to contextualize data to remediate the problems more efficiently.
Whatever the reason, the clarity with which AIOps use cases are defined will determine the tool’s usefulness.
2. Set success benchmarks
Investing in any new tool, let alone one as far-reaching and important as an AIOps solution, is a big lift. Enterprises aren’t in the business of buying software for the sake of buying software, so there must be success benchmarks in place to determine whether the investment was worthwhile.
IT leaders preparing their organizations for AIOps should prioritize success criteria and define how they’ll be measured. Typically, these ROI metrics will include mean time to resolution (MTTR), prediction and prevention of outages, increased employee productivity, and cost savings derived from reductions in man-hours via automation of repetitive manual tasks, or the elimination of multiple point tools.
3. Segment data that matters
“What gets measured gets managed,” the saying goes. But measuring everything isn’t reasonable when there’s just so much that can be measured. To realize full value from an AIOps investment, IT leaders need to focus on the specific data that matters to their organization.
Enterprises with expansive customer bases, like e-commerce, healthcare organizations or streaming content services, will want to ensure platform availability, low-latency data transmission, and service quality by analyzing data that predicts or avoids service outages.
Likewise, an app or software development firm will be more interested in data that highlights application performance or issues, dependencies, and effects on other systems, and information about users’ app experiences that help develop regression tests that can deliver enormous improvements in quality and cycle time.
4. Make an adaptable data collection and analysis plan
Segmenting data naturally depends on efficiently collecting it. And collecting the right data requires a comprehensive and well-thought-out data aggregation plan that enables a company to become an AIOps-powered enterprise.
AIOps tools rely on data from the highest priority endpoints from among the potentially thousands of devices, components, or customer touchpoints common to today’s sprawling IT environment to be valuable and effective. So, it’s imperative that a data aggregation and analysis plan clearly define and prioritize what types of data IT leaders need and assess whether it can be gathered with existing IT tools.
That also means planning for how to handle the various formats and states of data — structured, unstructured, or semi-structured — the systems will collect and devise strategies for streaming it to the data repository. It’s worth noting that because the needs of business and of its customers will change over time, the data produced will also evolve. Any data collection plan must be developed from a long-term perspective that tries to anticipate changes, to not get caught off-guard down the road.
The other element of the plan must address how the collected data is analyzed. The greatest advantage of AIOps is its ability to ingest the ever-expanding universe of analytics, logs, metrics, events, and data that’s churned out of the typical modern enterprise beyond human capabilities.
To that end, clearly defining the types, purpose, and priority of specific datasets ahead of deployment makes it easier to roadmap algorithm tuning to eventually filter out unnecessary data signals and help IT teams understand which data signals and alerts to pay the closest attention to and which are merely noise.
5. Setup the automation
If the purpose of AIOps is to separate the proverbial wheat from the chaff — to surface important issues from among a mountain of signals from and add context to those signals — then an organization must have a consistent stream of signals available to ingest. That is, before a company can enjoy the benefits of AI-powered operations, it must first have a well-established culture of automation.
Establishing automated workflows, runbooks and processes for fundamental activities like application performance monitoring, security breach alerts, and resource provisioning is paramount for AIOps readiness. Expanding automation to as many functional areas of the organization as possible provides dual benefits of alleviating IT teams from the burden of repetitive and time-consuming manual tasks as well as providing the AIOps engine with the steady stream of rich data it will need to effectively optimize IT Ops and IT service management practices.
IT infrastructure continues to become more complex, more involved, and more resource-intensive than at any time in history. As a result, organizations of all stripes are seeking comprehensive solutions for automating many of the most time- and resource-intensive IT Ops activities to free their teams to take a more strategic role in driving digital innovation and transformation.
AIOps is the logical next step in the evolution of IT management and can be a transformative force for putting an organization on a new path toward competitiveness and profitability. But only the teams that take a methodical, measured, and strategic approach to implementing AIOps will reap the full benefits of such an exciting, generational technology.
This article was contributed by Jiayi Hoffman, data science architect at OpsRamp.
18 July 2019