Asia is the fastest-growing region in the global e-commerce marketplace. As many as eight in 10 consumers in Southeast Asia are now digital, according to research conducted by Bain & Company and Meta, and they are churning a lot of consumer data.

With so much data to manage and crunch, artificial intelligence (AI) and machine learning are common technologies used to automate analysis and decision-making. However, developing AI models and data sets for use across an organisation, maintaining AI systems, and recruiting talent with AI expertise are huge challenges, limiting the integration of AI across organisations and deeper within functions.

According to IDC, insufficient pre-built automation within the tools that facilitate the setup, preemption, and auto-remediation of networks and data centres is a key challenge that APAC enterprises face. Automation in the cloud may be the solution.

The case for automation in the cloud 

Customisable automation capabilities across IT operations and business processes, and full-stack capabilities in areas of automation (e.g. process mining, computer vision, AI/ML) are most valuable in automation solutions.

But training AI models is challenging: If data sets are flawed, teams risk introducing biases which will require more person-hours to retrain AI models, slowing down product development and GTM plans. Losing time in a fast-moving sector can greatly impact bottom lines and competitiveness. At the same time, hiring the right AI expertise in an extremely competitive job market can set companies back financially.

Automation in the cloud means that companies do not need to invest in infrastructure or AI training. It allows for closed-loop automation where an AI agent is able to see, and based on what it observes, decide on a course of action before carrying it out. This includes even the difficult, time-consuming part of understanding and deciding on a course of action that is based on the telemetry collected.

By allowing companies to plug and play, overall costs are reduced in the long term and IT teams can focus on strategic and creative tasks, which increase productivity and optimise customer experiences.

Things to consider

When companies start out on their automation in the cloud, they should begin by identifying operational areas, then processes and tasks in those areas that are repetitive and take up an inordinate amount of manpower. In terms of AI investments, companies should target areas which have the highest returns first. This is determined by analysing if automation will lead to additional operational efficiency, long-term cost reduction, and/or a strategic competitive advantage.

In-house AI development tends to be long term. Companies can begin with an off-the-shelf solution first on a carefully augmented pilot project before considering more ambitious ones that integrate AI/ML operation into more diverse functions.

In the same vein, companies need to decide which tasks they can handle internally and when to tap on external resources for assistance.

For example, for automation that requires AI/ML, companies can leverage the cloud for ML training (with the cloud providers’ CPU and GPU compute), but develop and train the ML models themselves, then run these trained models locally for automation. This mix could offer the highest returns in terms of efficiency and cost, especially if the internal teams have the requisite expertise and experience.

However, for areas that are not core parts of the business, the company should consider continuing with off-the-shelf automation solutions as they would make more economical sense.

Fixing the plane as it flies

For companies to reap full benefits of AI/ML in the cloud, automation systems must be well integrated into other business and operational processes — and with each other — as they are implemented. To achieve that, AI/ML solutions must provide well-documented and complete API coverage. A lack of API and integration between critical automation systems (whether they are AI-enabled or not) are often the Achilles heel of such projects, and this often leads to even more manual processes.

Contrary to popular belief, a complete off-the-shelf product that has AI/ML automation does not always require a high level of expertise to use. One simple example is how cameras today have facial recognition, which can be set up as a trigger to start recording. This will make the surveillance process easier.

Another example is the setup of an office wireless LAN. In the past, it was often necessary to manually survey the radio environment to configure access points carefully and ensure sufficient coverage with smooth roaming. Today, cloud AI/ML-enabled wireless LAN controllers do this automatically while also continuously tuning the AP for optimal performance in a changing radio environment. This removes the need for time-consuming training and processes necessary to achieve the same result.

Automation in the cloud should be an easy decision

Automating in the cloud may seem like an intimidating choice because it is not well understood, but it follows the same journey as any automation project that companies are already doing, only with less time, and more precision and efficacy.

Juniper Networks’ global AI research findings found that 92% of APAC organisations have already utilised some form of AI-powered solutions to automate or aid decision-making. The upside of AI is well understood: 52% of respondents agree that AI will assist in reducing risk and increasing quality at work.

Clearly, more companies are automating, but what will differentiate them is how they are able to adopt automation most strategically and efficiently.