Loom Systems CTO: Why DevOps needs more Artificial Intelligence

Artificial Intelligence (AI) could help the DevOps profession, massively. 

Think about it, all that control and ability to connect and coordinate things, elements, code streams and workflows (that’s what DevOps is all about, right?) yet do so more automatically.

Slow adoption curve

CTO of Loom Systems Ronny Lehman argues that AI adoption in DevOps is on a relatively slow adoption curve — as is the adoption of machine learning algorithms by DevOps practitioners.

But, predicts Lehman, there are a handful of machine-learning ‘dependents’ that will soon become heavily utilised and relied upon by the DevOps community.

The following points A) to D) are contributed and attributed directly to Lehman himself — Loom Systems is an advanced AI analytics platform designed to predict & prevent problems in the digital business age.

Alerting rule-engine

A) Maintaining the alerting rule-engine, which is an ongoing and repetitive task with a straightforward methodology (act on false-alarms and missed detections). For a human, this means fast burnout. For AI, this is a classical problem that machine learning is a conventional fit for.

Alert prioritisation

B) Traditionally, alert prioritisation is manually maintained but essentially impossible in today’s over-flooded alert environment: How do you prioritise? it makes a lot of sense to prioritise based on data features such as past behaviour, magnitude of the current alert, the number of  other alerts being concurrently triggered, and of course, their source — but smart machines can do this process better and faster, based on endless important parameters that signify a shift in a component or application’s behaviour.

Root cause analysis

C) Obviously I cannot stress the importance of this capability enough. When something breaks, typically lots of alerts will be fired. Correlating, and even knowing the causality between the different alerts can help not only in grouping together related issues, but also in telling apart the Root-cause from the Direct-cause. This is the key to identifying and solving issues fast. Machine learning uses methods such as clustering to help DevOps cut through the noise and determine the sources and correlation of all alerts.


D) This is all about applying recommendation-algorithms to the ticketing system. Machines today can go beyond alerting and even analysis to apply algorithms that offer recommended actions based on real time and comprehensive analyses. An automated query to an application’s knowledge base, for example, is faster and more precise than any human is capable of. In fact, we’ve already grown accustomed to experiencing machine-learning in support sites, where we get recommendations for similar questions when we start typing our own. Machine-learning can do magic by recommending the relevant resolution.

Ronny Lehman of Loom Systems has experience in cyber and algorithms projects and implementations. He also has a track record in behavioural biometrics, authentication and malware detection solutions through machine learning and signal processing.