How many businesses actually use machine learning?

Everyone is talking about machine learning and AI… but how prevalent is it?

A recent survey by MIT Sloan has shown that less than a third (23 per cent) of businesses have adopted any level of machine learning automation, and of those who have, only five percent are using it extensively. It’s a similar situation with its subset, deep learning; an advanced form of machine intelligence technology that is expected to take time to gain wider understanding and acceptance.

“Machine learning is still something people are talking about, rather than using in any great numbers,” highlights Dorian Selz, Squirro CEO. “A lot of hype surrounds machine learning and AI but it’s still very early days so it’s mostly been embraced by early adopters and innovators. But the potential is astonishing and as soon as organisations see the sizeable and tangible benefits it can bring, it will be deployed in far greater numbers.”

However, things already look to be changing, with sectors such as cybersecurity, healthcare, retail and oil and gas pushing ahead with implementations. Analyst firm Forrester has also noted that the percentage of organisations using machine and deep learning had a steep increase from 2016 to 2017.

Wondering about the current state of AI and what happens next? Check out: Experts on hype, past, present, future and ethics of AI

“In 2017, 51 per cent of organisations surveyed implement, implemented or were expanding their use of AI over 40 per cent in 2016. Decision management (26 per cent), machine learning platforms (26 per cent) and deep learning (25 per cent) are the top AI building blocks in use,” says Diego Lo Giudice, VP and principal analyst at Forrester.

Machine learning is particularly finding a home within the customer service and support arena, thanks to the direct effect it has on the customer experience.

“Companies can enhance their ability to predict rather than react to rapidly changing demands and expectations,” notes Gajen Kandiah, president of Cognizant Digital business.

Furthermore, as Darren Roos, president of SAP S/4HANA Cloud, points out, adoption also frees-up time for employees to focus on the issues that matter to the business and its bottom line.

“Not only does it mean the organisation can harness further insights on the customer demands of today, but it can allow employees to be more impactful on the business, giving them roles that can touch customers in a way that a machine can’t,” he highlights.

According to Capgemini’s Digital Transformation Institute, businesses are using AI to increase sales, boost operations, facilitate customer engagement and generate business insights. And it’s working, as the Institute states that three quarters of firms it interviewed for a recent study are already seeing a 10 per cent uplift in sales since implementing the technology. Furthermore, 73 per cent think AI can increase customer satisfaction scores and 65 per cent believe it could reduce future customer churn.

High performance computing and machine learning could be a double-edged sword for datacentres: What will AI mean to the traditional datacentre?

“Tools are much more advanced than just five years ago and the primary beneficiaries of this are businesses and their customers,” says Kevin J Smith, Senior VP at Ivanti. “Businesses are able to offer services around the clock, faster response times, self-learning and self-help applications for customers, and the automation of common tasks that in turn free IT people to focus on the more strategic work,” he highlights.

“They offer better predictions, faster time to outcome and they learn and improve over time,” adds Walid Negm, CEO of Aricent. “The AI dividend is the learning feedback loop as a product or service gets better at what it does, such as raising the confidence of a prediction, more tailored personalisation or more accurate sensing.”

So what are the best uses of machine learning within business? According to Paul Hardy, chief innovation officer at ServiceNow, there are three kinds of decisions as targets for automation ­– anything requiring rating, ranking or forecasting.

“When building a roadmap, focus on those services that are most commonly used, as automating these services will deliver the greatest business benefits,” he advises. “At a high level, where are the most unstructured work patterns that would benefit from automation? Commit to re-engineering services and processes as part of this transformation, and not simply lifting and shifting current processes into a new model.”

Two clear trends are also appearing around management of decisioning applications. The first is outsourcing decisions via subscription APIs to specific vendors; the second involves building a model in-house.

Mark Sykes, COO of Kx favours the second route, either by the company alone, or in partnership with external experts, as he believes there are too many risks associated with outsourcing.

“These include security of data, supplier risk – many of these are startups – and the costs imposed upon the business when APIs are changed or discontinued. Such costs are virtually impossible to budget for accurately, and never happen at a good time.

“In-house solutions give the business much more control over the management and evolution of the finished product, and the knowledge of best practices learned throughout its implementation become useful assets for subsequent projects,” he notes.

However, one of the biggest current barriers to machine learning adoption is the quality of data, as poor data will lead to results without value and could actually increase risk.

Therefore, a business’ first step towards automating business decisions must around data curation and automating data management, as “there’s no use investing in automated systems if the data feeding into them is inconsistent, corrupted or disparate,” says Greg Hanson, VP of EMEA cloud at Informatica.

Looking forward, Ivanti’s senior product director of ITSM, Ian Aitchison, believes that self-service IT support will be one of the first areas impacted by machine learning, with ‘virtual agent’ AI bots helping employees with challenges. Second, he believes, will be complex data decision support, which will recommend proactive IT actions from data mining across large business IT data sets. “Kind of like ‘big data for IT’, but with automation and machine learning-based decisions,” he explains.

Deep learning is set to take things further, thanks to its ability to analyse big data and make recommendations and spot trends and patterns that employees simply can’t due to the enormous volume of modern enterprise databases.

But, as Aitchison highlights, the future of IT is more than just learning and decision-making.

“Connected to this is the conversational communication to interact with seemingly intelligent IT solutions that can decide and action the right responses,” he says. “For example, an end user can chat with a bot to report an IT fault, and the bot can drive automated corrective or enhancement actions to help the user, driven by learned responses to previous similar requests.

“Gartner described conversational AI as the biggest imperative for IT departments since cloud and mobile. These types of platform paradigm shift generally occur once a decade.”