Getty Images

How to scale up your AI initiatives

Executive sponsorship, a strong data foundation and having ‘analytics artists’ are among ways to scale up AI initiatives beyond the pilot stage

Most business executives believe they need to harness artificial intelligence (AI) to stay ahead of the pack and grow their business, but they often fail to scale up their AI initiatives across their organisations, according to an Accenture expert.

Citing a global study by Accenture, Lee Joon Seong, managing director for applied intelligence in ASEAN at the consulting firm, noted that while 88% of global executives believed they needed AI for their business to survive, the same proportion also struggled to scale AI initiatives beyond the pilot stage.

“A lot of people understand the potential of AI and have embarked on AI initiatives, but not many have fully realised their full potential,” said Lee.

The ability to scale is seen as a barometer of success in AI adoption, given the time, talent and resources involved in AI projects. Here’s what organisations can do to scale up their AI initiatives:

Earmarking AI as part of your business strategy sounds obvious, but many organisations still struggle to get that right.

Lee noted that while chief data and analytics officers can build an AI platform, an organisation’s business functions may choose to “get on with business as usual”.

To align AI initiatives with the rest of the business, Lee said sponsorship from the leadership team is critical, citing organisations such as Singapore’s DBS Bank, whose CEO has been very vocal about the use of data and AI.

The missing middle

Another ingredient for scaling up AI initiatives is access to talent, though Lee said organisations must be clear on the sorts of talent they need.

“When you put businesspeople who have weak understanding of AI with a bunch of data scientists and engineers, nothing happens,” he said.

According to Lee, the crux of the matter is what he dubbed the “missing middle”, or the lack of people with analytics skills and the ability to build bridges between business and data science teams.

“We want a lot of data scientists, but we also want a lot of analytics artists or translators who can bridge between the science of data and the art of business,” said Lee.

Lay your data foundation

Lee noted that having a strong data foundation is key as “data is the rocket fuel for growth” of AI initiatives.

Noting that technology is evolving rapidly, Lee said organisations that are managing data at scale today would have to do so on the cloud. Many Accenture clients, however, are still operating traditional data platforms which are becoming obsolete, he added.

Also, organisations often do not know how much data they really need – and even though it is generally accepted that the more data one has, the better it is, that is not always the case, said Lee.

“They also have issues with data quality and other dimensions, and everybody is using a slightly different version of the data with no single source of truth,” he added.

Read more about AI in APAC

These challenges, said Lee, cannot be easily solved in the short term, but on the other hand, organisations have a pressing need to differentiate themselves through the use of AI and data.

Organisations can start by getting a set of data right to support a specific AI use case that has been prioritised at a strategic level.

“If supply chain is an area you want to focus on, then fix the supply chain related data and get that into production first while you think about other data,” he added.

Massive experimentation

Working with a set of data to support specific use cases is a hallmark of organisations with a mindset of massive experimentation and rapid prototyping.

Singapore’s Changi Airport, for example, worked with Accenture to create a smart luggage fitted with sensors that collects data on vibrations and other factors related to the condition of the airport’s baggage handling system.

“So, instead of having people inspect the baggage handling system, we created a luggage with sensors to do preventive maintenance,” said Lee. “That solution didn’t exist before and therefore you need to do experimentation and try it out.”

Democratising AI

Paving the way for AI and data science capabilities to be democratised and accessible to citizen data scientists is another way to scale up AI initiatives.

Lee said that while citizen data scientists may not be AI experts, “they can also develop AI solutions by leveraging tools and capabilities out there to achieve impact”.

“As we help our clients scale and get talent, including analytics artists who understand industry and business, we are trying to help democratise data science and AI by grooming citizen data scientists,” he said.

Lee said the availability of cloud-based AI application programming interfaces today also makes it easier for one without coding skills to build AI models.

“You will still need AI specialists with PhDs because they will help you differentiate, but at the same time, you’d also need a lot of citizen data scientists,” he added.

Next Steps

Swiss retailer uses open source Ray tool to scale AI models

Meeting the challenges of scaling AI with MLOps

Read more on Artificial intelligence, automation and robotics

Data Center
Data Management