As is the case when any business shifts its operating methods, the adoption of AIOps requires changes in many areas. It introduces the dynamism of machine learning to the traditionally static and cautious environment of IT operations. In doing so, AIOps represents a significant leap in the development of IT operating models. This blog post will explore what businesses should consider when thinking about adopting AIOps.

1. Culture change

As discussed in “Business and Network Leadership in the Digital Transformation” and “The Changing Role of the Network Professional,” business leaders and staff alike must move away from traditional ways and views of roles in the workplace, if they are to make a success of digitalization. There is no quick, easy fix that enables a traditional business model to simply keep up with digital transformation, let alone make a success of it. Similarly, the effective deployment of AIOps functionality requires a business culture change. It is essential even when AIOps is restricted solely to monitoring operations, which is presently the case for most deployments.

IT has always played a supporting role for the rest of the business. Charged with keeping the lights on, it has been internally-focused, without responsibilities towards external users. It has therefore typically had a conservative outlook because the rest of the business has not needed it to offer any more than just support. However, if a business is to digitally transform, the IT team must transform as well, because the IT team is now a core business asset.

Adapting a digital approach encourages IT teams to learn more about their environment – they will search for more data and ask more questions of it. The learning mindset will encourage professionals to move away from a position of ‘expertise’ to a more dynamic, agile stance. With learning, dynamism and agility comes innovation, which is key to staying ahead of the competition. This stance will recognize the potential of the automated and machine-assisted collection and interpretation of data that are the foundation of AIOps, and which can deliver results that far exceed what a human alone could achieve.

2. Application and service delivery

We have seen that AIOps is a response to the increasing emphasis upon application and service delivery in business today. The application and service environments are joining together as end users demand faster and more reliable user experience. As the collective business mindset changes, and IT professionals see their roles as facilitators of business, AIOps can be introduced as a means of directing the new method of operations that is required.

Shifting their focus from internal network concerns to leveraging the revenue potential of their business data, IT teams will see that it is vital to bring together service delivery and network performance management. With automation, AIOps combines these disciplines so that IT can effectively manage and use big data, and apply the analytics and machine learning for greater insight and better business results.

3. Eliminate silos

All forms of AI and analytics require a consistent, unified dataset. These processes must draw upon a variety of sources from across the network. If they do not, the data collected will not be representative of the network’s condition, and the resulting analytics, suggestions and actions will be incomplete. Further, the effectiveness of AIOps will be reduced if network data is arranged in siloes, or multiple tools are used to collect the data. Not only this, but siloed tools and information limits collaboration, hinders communication, and slows down development and innovation.

The rate of change in network complexity and innovation is accelerating, and network issues need to be addressed faster than ever before they impact business services. It is therefore essential that IT resources and datasets are unified, and that IT teams have the appropriate tools to achieve this. When data and resources are aligned, it is easier to align the people and processes that work with them. In defining AIOps, Gartner recommends that IT operations be reorganized according to the data sources used, rather than the existing technology types or infrastructure layers. This will make it easier for IT teams to collect consistent data from a variety of sources in one place. With unification and consistency comes scope and depth, enhancing the quality of data fed into the machine learning and analytics processes of AIOps.

4. Access to data and algorithms

The machine learning process of AIOps depends on collecting data from a variety of sources. Algorithms can then be applied that learn patterns of behavior and trigger actions using automation tools. Gartner identifies functionality that collects and provides access to log data, text data, wire data, metrics, API data and user sentiment data as necessary for a comprehensive AIOps platform.

Gartner recognizes that many vendors offer a broad range of these capabilities, contributing towards AIOps functionality, without necessarily covering all of them. In these cases, providers should be open to integrating with other vendors, to complement their own offerings and contribute to a more comprehensive service for end users.

The extensive list of functionalities shows the importance of collecting data across a wide range of sources. Modern networks and IT systems are increasingly dynamic and distributed. It is difficult enough to understand the present state of networks and IT systems, without worrying about why they are in their current state, or what might happen to them in the future. A truly accurate and up-to-date view of the network relies on multiple perspectives of data, and a means of aggregating and storing it. Machine learning can help IT teams with understanding these past, present and future conditions. The greater the range and depth of the dataset, the greater the understanding and more accurate and applicable the machine learning will be.

Gartner’s list of AIOps capabilities[1]:

  • Historical data management
  • Streaming data management
  • Log data ingestion
  • Wire data ingestion
  • Metric data ingestion
  • Document text ingestion
  • Automated pattern discovery and prediction
  • Anomaly detection
  • Root cause determination
  • On-premises delivery
  • Software-as-a-service

Whilst digitalization is a culture-led process, there is clearly a technological element that will follow – but not precede – the transformation of a business’ culture. It is only recently that machine learning algorithms have been applied across IT operations, and so the technologies that deliver these algorithms will not have been part of the traditional and original IT setup. Therefore, it is essential that businesses consider investment strategies both for acquiring the appropriate technology, and the skills needed to operate it. Adaptation of and integration with legacy infrastructure is another concern. Investment in leading technologies and algorithms will likely require consideration across the whole business, not just within the IT budget.

5. Gradual learning and progress

There is no ‘end result’ for digitalization or AIOps. They are ongoing processes that give businesses the ability to innovate, or react to industry and market movements. The complexity and difficulty of introducing AIOps processes means businesses should adopt changes incrementally, especially if IT still currently operates with largely traditional processes. Gartner suggests IT teams select tools that enable incremental deployment of machine learning. This is based on four phases: visualization and statistical analysis; automated pattern discovery; pattern-based prediction; and root cause analysis.

Just 16% of executives believe their business has all the skills and capabilities necessary to deliver its digital ambition.

Forrester, ‘Staffing And Hiring For Digital Business’ (2016)

This enables a gradual learning process, from analytics of large volumes of historical data through to predicting future states. Given many businesses face a skills gap, it is essential to develop a base expertise in core processes. From this point, IT professionals will become generalists in a number of different areas, such as algorithm management, applications and security, whilst machine learning takes the heavy lifting in each area.  If AIOps is to augment humans in effective ITOM, the human worker should understand how it does so. In this way, IT teams can effectively oversee the machine learning, rather than doing the work themselves.

Making data a priority throughout the business is the first step towards the operational changes needed to implement AIOps. As IT teams take on more responsibility for external users, they will become more focused on delivering what will contribute to creating more value for those users. To achieve this, agile and automated processes are needed that free IT teams from the burden of tasks that will be more quickly and effectively executed by machine learning algorithms. Because these requirements involve fundamental changes to culture, technology and training, businesses should begin introducing these changes as soon as possible. The rate at which machine learning capabilities will improve continues to accelerate, but the necessity to master the basics will always remain.

[1] Gartner, ‘Market Guide for AIOps Platforms’ (2017)

Share This

Share this post with your friends!