As AI expands to touch every aspect of your business, how can your company embrace machine learning? What are the pitfalls to watch out for? And how will active machine learning influence the future of your enterprise?
When AI is discussed in a business context, what is really being outlined is machine learning (ML). ML is the practical application of AI. With a long history that stretches back to the 1950s, this brand of AI is rapidly maturing. As a core component of AI, machine learning forms the basis of many systems your business can use today.
The practical application of machine learning takes many forms: The Facebook Messenger app, is perhaps, the most conspicuous application of machine learning. Used by millions every day, the application can contain many instances of bots that offer specific services. These conversational bots can seem highly realistic, as the machine learning algorithms in use, have been honed for several years by their creators.
Machine learning is also now extensively used in the retail sector. Supporting customers when they have a query with a bot as the first point of contact is becoming commonplace. For example, two years’ ago, Very.co.uk launched its bot within its iOS app to help their customers track orders and payments. The discount supermarket Lidl created Margot on Facebook Messenger that can understand natural language.
According to Juniper Research, retail chatbot interactions will reach 22 billion by 2023, up from an estimated 2.6 billion in 2019. Research author Nick Maynard explained: “By embracing automated customer service with chatbots, retailers can act in a more flexible and efficient way. The wider retail market means that chatbots are no longer a luxury; they are essential.”
Adam Hadley, CEO and Founder, Quantspark told Silicon: “As with any strategy, CTOs and CIOs need to have a return-on-investment front of mind. ML is itself, not a strategy; of course, it’s a tool to realise a strategy. That’s a distinction that can be often overlooked. So, CTOs need to be asking ‘what is the strategic problem I’m trying to solve here?’ From there, they can explore the extent to which ML can help. A separate point worth raising is human resource – as with most things, the quality of your team is essential, and this is undoubtedly the case with ML initiatives. We’ve seen businesses really struggle to hire the right mix of strategic and technical mindsets that are needed to develop an impactful ML solution.”
With Julian Nolan, CEO of Anglo-Swiss AI company Iprova, also commenting: “Within the C-suite, there’s a tendency to either underestimate or overestimate the capabilities of AI, sometimes both at the same time. Many see its use as a paradigm shift or aid to ‘boiling the ocean.’“
Nolan concluded: “AI is also a widely misused term, a catch-all for many techniques ranging from simple regression analysis through to machine learning and deep learning using neural networks, natural language processing and a whole host of other techniques. Even if management moves into AI with eyes wide open, there are multiple pitfalls, primarily in the integration and acceptance of these technologies in the organisation; whether that’s integrating with existing business processes and applications, or resistance and lack of readiness from the wider workforce.”
There is little doubt that machine learning will impact many aspects of your business over the short term. Says Deloitte: “From improving productivity on large construction projects to helping customers choose the right furniture, machine learning and deep learning technology are being applied to almost every aspect of business and industry. The combination of big data and neural networks is helping to unlock the value of data businesses already have by revealing patterns that they can use to create and improve offerings or productivity, or to gain an advantage over the competition.”
Machines are fantastic at finding patterns in large datasets. Some of the most extensive datasets your business creates are based on your customer’s behaviour. Here, machine learning and predictive analytics come together. The simple buying comparisons that have been common on retailer’s sites will expand to encompass the broader behaviour of each customer. Taking a step is to use machine learning applications such as Google’s Duplex to make these interactions seem more human.
Using machine learning with your CRM is a powerful way to use the unique pattern-recognition skills that a machine learning tool can deliver. Companies such as Salesforce are already using AI and its machine learning component to enhance the services their applications offer. Their Einstein AI can be integrated into Salesforce Customer 360, to deliver intelligent data analysis and predictive forecasts.
Nico Acosta, Director of Product Marketing at Twilio, explained to Silicon: “There are two major considerations that should be top of mind for CIOs and CTOs thinking about implementing machine learning. The first is the importance of creating a universal data pipeline. Organising, standardizing, and mobilising data is a huge challenge for businesses, and it’s so important to set a good foundation before you set out on this task. Secondly, you need to ensure that the vendors and systems you choose to help structure, implement or actual leverage ML are flexible. It’s essential that collaboration can happen across your technology stack, and you can continue optimizing and iterating. Without that flexibility, you are missing opportunities to grow and innovate.”
Machine learning evolves
A clear business case for implementing machine learning is the key to a successful deployment. Currently, your business will have to some development work, as we are a few years away from packaged applications that can be integrated together to form an ML service.
Says Adam Hadley, CEO and Founder, Quantspark: “Machine learning is technique rather than a technology per se. That means it’s accessible to almost all businesses. Colleagues can choose the tool of their preference to implement ML algorithms. R and Python are the most common, and also Open Source.
“The key to success is not the ML technologies themselves so much as the foundational data systems on which they are deployed. While less sexy, investment in robust, automated and real-time data collection, data processing and data validation systems is integral to a good data strategy. ML solutions are only as good as the quality of that the feeds them.”
Your business should also pay close attention to the skills your enterprise will need to implement machine learning. Indeed, according to new research from Rainbird – the AI-powered automated decision-making platform, 81% of those surveyed revealed that their organisation planned to increase investment in AI over the next five years. Of those who plan to increase spending on automation technologies, 22% suggested this investment would be significant. Interestingly, the financial sector is set to be the biggest adopter, with 94% of those surveyed planning to increase investment in AI over the coming years.
James Duez, CEO at Rainbird, commented: “AI should be brought into organisations to help employees, not hinder them. UK organisations – and beyond – need to fundamentally change the way they are adopting AI and, think beyond big data and machine learning. ‘Data scientists only understand black box solutions, and there are huge benefits to be had by moving towards more transparent symbolic technologies which can achieve automation outcomes beyond those available with data-only approaches. Such accessible tools also have the added benefit of addressing the skills gap by making AI far more accessible to employees without a degree in data science.”
Machine learning is one of the fastest moving aspects of AI today. From search engines to quantum computing, this branch of artificial intelligence will touch every aspect of your business. Initially, your customer-facing communications, but later how you manage the vast quantities of data you are collecting and, how this can be used to reveal new commercial opportunities.
Silicon in Focus
Dave Coplin, Former Chief Envisioning Officer for Microsoft UK and an established thought leader.
Dave Coplin has, for the last three decades been at the cutting edge of some of the world’s largest technology companies focused on the intersection of our society and modern technology. Formerly Chief Envisioning Officer for Microsoft UK. He has written two books, worked all over the world with organisations, individuals, governments and government agencies – all with the goal of demystifying technology and championing it as a positive transformation in our society.
What are the leading use cases for machine learning at the moment?
The best areas to explore the value of machine learning are any parts of a business that follow an established pattern. This could be simple things like established processes (think call centre troubleshooting scripts, etc.) but also applies to more complex parts of a business.
For example, the volume of customers in a given retail outlet will vary depending on elements such as weather, time of day, time of year, and so on. If you could bring together the historical data from each of those areas, you should be able to build a rudimentary predictive model that allows you to predict future business performance. Start small and experiment, and always look for more and better data – the more data you have, the more reliable your prediction will be.
Can you point to any great examples of how businesses are using machine learning?
Most organisations are well on the way to put machine learning at the heart of their businesses. From retailers like Ocado using machine learning to provide a fifteen-fold improvement in detecting and preventing fraud through to big manufacturing organisations like Rolls-Royce fundamentally changing their business model by switching from selling products supported by separate maintenance contracts and instead of selling a “subscription” to a product as a service, inclusive of maintenance.
What are the pitfalls CTOs and CIOs need to be aware of when implementing a machine learning strategy in their businesses or organisations?
The IT industry loves a bandwagon, and right now, they don’t come much bigger than the one that AI and ML are riding into the world of business, and everyone seems to be scrambling to jump on. AI and ML are transformational technologies and probably the most critical technologies on the planet right now, but that doesn’t make them the right tool for every job. L requires data and lots of it. If you don’t have enough data, your models won’t be reliable. If your data is of low quality, then your results won’t be any better.
These are apparent pitfalls and offer no real change from the garbage in/garbage out principle that CTOs and CIOs have been aware of for decades. What is novel, however, is a new world of moral and ethical issues that will likely be new to most organisations. Data is the lifeblood of ML and with it comes a range of issues that really need to be understood by most organisations.
There are important decisions to be made about how the data you already hold could be used, remember that GDPR exists for a reason and you need to consider the implications of how you choose to use and store data as a result. Equally, data is inherently biased (because it is scoped, generated and collected by humans) and those biases will always find their way into your results.
How does machine learning integrate with the other technologies businesses are using as they continue to digitise?
Invented in the 1950s, Machine Learning is a relatively old technology, but it has been brought to prominence recently by advances (or even with the general availability of) the cloud. ML and the cloud go hand in hand, from being able to store datasets whose capacity would leave most on-premise storage managers in tears through to scalable compute cycles that enable ML algorithms to find the elusive needle in the data haystack.
ML and the cloud go hand in hand and need to be part of any organisation’s overall technology architecture. At one end, the reality is that you will need to be able to deploy various uses of ML right across your business and at the other, it’s inevitable that those ML algorithms are going to need different aspects of the same organisational pool of data. Creating a data strategy (and a supporting data architecture) are the critical success factors for any organisation seeking to put ML at the heart of their business.
How do you think machine learning will develop over the next few years?
Machine Learning is going to grow in prominence and mature over the next few years, but perhaps the most significant developments will be around how organisation deal with the more complex social, moral and ethical challenges that the technology brings with it.
Creating algorithms that are free from bias and that respect the data rights of the individual are all on the road map, but so too are issues captured beautifully by Jeff Goldblum’s classic “Jurassic Park” principle (just because you can do something, it doesn’t mean you should) – facial recognition is an excellent example of machine learning-based technology that is going through that principle right now.
The answers aren’t always going to be clear cut, but the only way forward is to make sure that you table the debate with a broad range of stakeholders inside and outside of your business to ensure you can come to an answer that is right for everyone.