Artificial intelligence is the most disruptive of the digital technologies, and we have seen that a business must undergo a cultural change and embrace new methods of operations if it is to successfully introduce AIOps. But what does this mean for the day-to-day tasks of ITOps – the people who are tasked with implementing it? Digitalization is already changing the role of the network professional, and yet AIOps makes further demands of IT staff whilst providing even greater opportunities. This article outlines three of the main demands and impacts of AIOps on the teams who deliver it.
1. Human thinking vs machine thinking
Artificial intelligence is already pervasive across the globe, enhancing and powering various parts of our lives across many industries. It is particularly suited to situations requiring direct, human-scale interactions. Examples include self-driving cars, asking Siri for directions or when Facebook recognizes familiar faces in photos.
But ITOps are not conducted on a human-scale. Indeed, IT networks are growing ever larger, and becoming more complex and widely distributed. Artificial intelligence is not yet sufficiently advanced to look after a network on its own, simply because machines are not yet able to think as creatively as humans. However, machine learning – which is the leading driver of AI development – is far more efficient than humans at executing the fundamental tasks of network management. Machine learning has already been used to automate and enhance trading, an environment where real-time responses and extensive user customization are needed for dynamically-changing conditions. Going forwards, it will become crucial for ITOps teams to harness such efficiency.
The global volume of business data doubles every 1.2 years.
Waterford Technologies, ‘Big Data Statistics & Facts for 2017’ (2017)
With the explosion in available data, IT must now handle datasets on scales beyond feasible human calculations. The underlying algorithms of machine learning can handle and extract knowledge from dynamic and growing datasets at an exponentially greater speed and scale than can a human workforce. Machine learning comfortably copes with data, event and security storms, where human judgement and traditional software engineering struggle.
Simply put, the machine learning aspect of AIOps delivers the efficiency and scale of data interpretation that supports the limitless creativity of human thinking. This shift in responsibility is the basis of a shift in the role of the IT professional from specialist to generalist. Data volumes now mean it is impossible for any one person to attain knowledge in a specific area that is remotely comparable to that acquired through machine learning, let alone across the range of fields required for AIOps.
The IT professional’s move from specialism to generalism is not a downgrade in significance, however. Rather, it emphasizes the creative potential of human thinking that machine learning – and therefore AI – is nowhere near close to imitating. With machine learning responsible for the raw labour of extracting knowledge from data, ITOps will have the freedom to become creative. As a result, IT teams will no longer need to spend time chasing down faults or security alerts across the network, for instance. This freedom will empower teams to implement ideas that would be impossible to execute without the scale and efficiency of algorithms. The potential of the network increases as more insights are exposed, and this encourages new ways to think about problems and plans for the future.
In this way, AIOps contributes to the new digital trend of augmenting the human worker. In the new data-centric paradigm, automation and analytics make decision-making easier for IT teams. As these teams move closer to the business’ frontline, AIOps becomes essential in helping them focus on building scalable networks to meet business growth. IT operations will no longer be a merely reactive firefighting service. Instead, it will become exponentially more informed about the past, present and future state of the network, and therefore more valuable to the business’ success than ever before.
2. AIOps skillsets
AIOps can bring many benefits. But it is not simply a case of turning on machine learning, sitting back and waiting for results. As discussed above, the limitations of machine learning mean that human workers remain integral to ITOps.
Machine learning algorithms can be unsupervised, where patterns are found in a set of observations. However, because the way in which algorithms handle the network will directly impact upon overall business performance, simple pattern recognition is not enough. Therefore, AIOps requires supervised machine learning. IT teams must be sufficiently trained so they can inform the algorithms with examples of both preferred and undesirable network states. Further, the IT teams need the machine learning algorithms to inform them how and why the network is performing like it is. This is only possible if the algorithms themselves are running with a context of what that performance should be.
Over 40% of organizations practicing advanced analytics are challenged by a lack of adequate skills.
Gartner, ‘How to do Machine Learning Without Hiring Data Scientists’ (2017)
In an AIOps environment, therefore, IT professionals become the conductors of the network’s automated delivery. They require the appropriate skills to undertake these tasks, which includes the unification of applications and services. They will understand the algorithms and why they produce the results they do. With understanding, the algorithms can be audited and, if necessary, adjusted for the greatest business benefit. By handing over much of their original labor-intensive responsibilities to machines, IT professionals are now tasked with much broader responsibilities to keep the machines and infrastructure of the business running.
Between 2014 and 2017, demand for engineers developing AI and machine learning algorithms has increased by 485% in the UK, and there are 2.3 roles available per qualified candidate.
Computer Weekly, ‘UK at risk of AI skills crisis’ (2017)
But businesses could face a steep learning curve in establishing effective AIOps. The task of accessing the necessary data and integrating it is considerable, and that is before any machine learning is applied. It is also very difficult to achieve without the skills of a data scientist to help orchestrate a supervised machine learning strategy.
However, one of the purposes of AIOps in particular, and digitalization in general, is to deliver to businesses the benefits of a data scientist without necessarily hiring one. To overcome this issue, Gartner has suggested that businesses can train IT professionals as citizen data scientists, work with academic institutions, or employ third-party professionals or applications.
To begin with, increases in capability will be incremental. Mastery of large datasets is the first step, followed by the gradual addition of new skills. This is as much led by culture as it is by technology, so it is crucial to get everyone on board with this gradual workflow. Only then can teams working with the machine learning algorithms fully understand their capabilities and what they can do for the network and business.
3. How IT operations will operate
Alongside skillsets, businesses must think about how they deploy their operational resources when adopting AIOps. Specifically, ITOps will be reorganized according to the data sources used, rather than by existing technology or infrastructure layers. In the age of digitalization, data is king, and without data there can be no analytics.
As is common to the digital transformation, AIOps requires the removal of information silos. Despite the complex algorithms that underpin machine learning, the quality of output from AIOps still depends on the quality of input. If only a few data sources are used, or plans are executed only within a certain business silo, then the results and analytics will only be as good as those silos.
Indeed, Gartner has specified the datasets that constitute true AIOps, as we have previously covered. This further underlines how important it is to unite separate silos from across a business network. To achieve this, IT teams need to familiarize themselves not just with machine learning, but also the roles of both network engineers and software developers.
Such changes to the way IT operates are necessary because, simply put, traditional IT management techniques are inadequate for managing extensive machine data. The speed and scale of analytical data insight is too great for rigid work streams and processes. They lack the capacity and agility to handle incoming insights and apply them for business benefit. By contrast, more recent methodologies, such as agile and DevOps, are more suited for the speed of collaboration needed to create the conditions for machine learning. Agile and DevOps can help minimize labor and politics across silos when correlating data from across the business. If followed correctly, they can ensure the quality and consistency of data produced.
A 10% increase in data accessibility will result in more than $65m additional net income for a typical Fortune 5000 company.
Forrester, ‘ROI of Data Quality’ (2017)
Businesses will also be further encouraged to move operations beyond the premises. Extending network infrastructure to the cloud, as well as adapting to decentralized data collection and analysis through connected devices and trends such as edge computing, will become ever more important. Operations across distributed environments will provide the data that truly reflects the state of the network, both on premises and in the cloud. Ensuring the speed and quality of these varied interconnections will provide conditions that most closely follow the real-time interactions required for better learning.
With silos being eliminated and businesses increasingly unified in purpose, the impact of AIOps will spread across departments and within applications. As it does so, IT should retain governance of AIOps deployment strategy, and ownership of data collection and analytics across departments. With consistent ownership comes consistent processes. It will also enable and encourage uniform interaction between IT teams and the business development professionals, who rely upon the insights provided by AIOps to improve the business’ offerings.
AIOps is about harnessing AI to enhance the human worker, remove friction from workflows, and increase efficiency across the business. Implementing this disruptive and technology-heavy paradigm across the entire business is labor-intensive and cannot be undertaken lightly. IT teams will take on more generalist, wide-ranging and creative responsibilities – uniting IT functions – whilst supervising the work undertaken by the machine learning. They will collaborate with business drivers across departments to ensure successful delivery. In this way, both the human and the machine can work to their respective strengths for the benefit of the business.