News

AI set to replace traditional IT administrators

Demand for higher levels of availability will lead to increased use of automation. Over time, many IT roles will be taken over by intelligent algorithms

This article can also be found in the Premium Editorial Download: Computer Weekly: The march of the robot workers
Cliff Saran
By
Published: 25 Oct 2016 15:10

In the age of the customer, human IT operators will be inferior and a new rulebook for IT operations will be needed.

IT will need to change from supporting back-office processes to interacting directly with consumers and users, who will have far higher expectations of IT systems than traditional business users do.

Speaking at the recent VMworld conference in Barcelona, Joe Baguley, CTO EMEA at VMware, said: “IT needs to shift from being business-aligned to concentrate on understanding much more about what customers and end-users in the business do.”

Baguley argued that most IT systems were designed to support business processes rather than employees, such as an HR system, which is optimised to help HR people but makes little sense to staff.

“The future is no longer about the business or the datacentre, but about customers,” he said.

Speaking in August about how the IT function was changing as cars became smarter, Volvo Cars’ director of business development and strategy for consumer IT, Jonas Rönnkvist, said this was making car companies into IT operators.

“Our core systems are supporting customers 24/7,” he pointed out.

Supporting the customer puts huge demands on back-end systems that were previously designed for internal or business-to-business (B2B) operations.

Wolfgang Krips, executive vice-president for global operations at Amadeus, which runs global ticketing distribution systems for airlines and travel agents, said businesses could no longer differentiate between clients and consumers.

Big shift

“The big shift in the IT industry is that the difference between B2B and B2C [business to consumer] is going away,” he said. “If you don’t understand what is going on client-side, you are not doing the right thing.”

For example, customer loyalty was non-existent in the online travel booking business, he said. As a result, the service needed to be 100% available, representing a fundamentally different understanding of service-level agreements.

Similarly, said Krips, airlines ran a business model designed to fix costs. The business tried to maximise revenue from passengers by up-selling, which meant adding more and more products, and the associated IT to support them, he said.

But Krips warned: “Changes on back-end system have to go in fast, so you have to increase stability and agility.”

Taking people out of IT

Such demands on the business were leading to radical changes in the way Amadeus saw IT operations, he said. “We are automating to take people out of the system. In the future, my people will become automation engineers.”

This is a big transformation. IT has to think about how it delivers services, and applications must adhere to newer paradigms. “You have to find ways to operate in a hybrid world,” said Krips. “Today, if we have a problem, the vendor fixes it. In a cloud-based world, your applications have to deal with the problem.”

There was also the opportunity to make use of artificial intelligence (AI) and best practices derived from fraud detection techniques used in financial services to flag up potential problem areas based on deep analysis of telemetry data from running IT systems, he said.

“Operating at scale requires a totally different arsenal of firepower,” said Krips. “There is no technology, no matter how advanced it is, that I can ignore because the problem space is so complex.”

Deep learning

He said Amadeus was looking at using IBM’s deep learning Watson system to preview incidents in operations. Watson has had some success in healthcare, helping clinicians to tackle certain types of cancer.

But analysis of server logs is an analogous problem, said Krips. “There is no human who can look at millions and millions of events a day. You need machines that can make decisions.”

Krips sees IT operations staff managing only by exception, when the machine cannot figure out how to solve the problem.

According to Gartner’s smart machine research, self-learning automation technology will be a “top five” investment priority for more than 30% of CIOs by 2020.

Read more about machine intelligence

  • Machine intelligence is improving quickly – as demonstrated by the success of Google’s DeepMind AlphaGo – and IT leaders need to plan for its effect on enterprise IT.
  • We look at how companies analyse server and security logs to tackle cyber crime and internal fraud, and optimise the user experience.
  • How AI in HR brings out the best in people

In the IT department, it is not a case of whether or not employees will be replaced by cognitive systems, but when. VMware’s Baguley said: “Graphs of alerts are not useful for humans. You need AI that can take actions.”

Baguley said he expected IT management tools to evolve over time to support machine learning, enabling cognitive systems to take over from human IT operators.

Arguably, companies like Google offer a glimpse of what is possible in terms of machine-learning IT admin algorithms. Mark Shuttleworth, founder of Canonical, said Google could understand what was happening both on a mobile device and in its server racks, enabling it to use AI to manage server workloads more efficiently.

“We are starting to understand that what we call machine intelligence is just a particular type of pattern matching and we are starting to encapsulate software so we can correlate data,” said Shuttleworth.

Greater scale

In the past, it was impossible for IT administrators to compare server logs outside their own organisation, he said. “Since we are now operating software at greater scale, we are using a set of the same patterns [for optimising the software]. With machine learning, I can learn from a million other deployments and provide intelligence.”

Whereas an IT administrator may have looked at logs and taken the average to assess a system’s health, learning from the huge amount of historical data means AI-based IT admins can identify so-called “black swan events”, the minuscule signals of abnormal activity that precede major IT failures.

Analyst Forrester said recently that human office and administrative tasks were fertile ground for robotic process automation. As society’s reliance on complex IT systems increases, people will expect 100% uptime, and the experts believe this is unattainable unless traditional IT administration is automated.

Read more on IT operations management and IT support