When Google’s AlphaGo machine learning program defeated a human champion in the insanely complex board game Go, many dismissed it as yet another computing publicity stunt – a purpose-built parlor trick, more Barnum than business case.
They were mistaken.
Research conducted by the Economist Intelligence Unit (EIU) and written in discussion with SAP shows that many organizations are moving ahead now, some aggressively, to integrate machine learning into their operations. For example, the survey of 360 organizations shows that on average 68% use machine learning to at least some extent today to enhance their business processes.
Our analysis of the data suggests that the rapid progress of machine learning is much more than an overreaction by management to another bout of technology hype. There are clear indications that organizations are using machine learning to signiﬁcantly improve performance across the breadth of their operations. Some are aiming even higher: to use machine learning to change their business models and oﬀer entirely new value propositions to customers.
What is machine learning?
A major discipline of artiﬁcial intelligence (AI), machine learning uses sophisticated algorithms to enable computers to “learn” from large amounts of data without being explicitly programmed. The more data the algorithms can access, the more they can learn. This enables software to adapt and improve the execution of tasks and processes autonomously and continuously.
These signals are clearest in the responses to our survey from a group of organizations that have implemented machine learning and are already beneﬁting from it. Most of this group have C-level leadership and support for their machine learning initiatives. The beneﬁts they are seeing – such as increased proﬁtability and revenues, more competitive diﬀerentiation, and faster, more accurate, and more cost-eﬃcient processes – are substantial, numerous, and span the entire organization.
We call the organizations from the survey that are already seeing beneﬁts the Fast Learners. As a group, their success demonstrates that organizations that aren’t actively investigating machine learning’s strategic relevance to their organizations are taking a big risk.
What are the traits of fast learners?
How have the Fast Learners leapt ahead of the competition? Five key traits emerge from the research that are important to their success – and serve as a template for organizations that have not yet developed an machine learning strategy or are just getting started.
1. C-Level, strategic priority
Machine learning is not just a technology; it is core to the business strategies that have led to the surging value of organizations that incorporate it into their operating models – think Amazon, Uber, and Airbnb.
Fast Learner organizations get this. They are beneﬁting from their senior-most management seeing the strategic value of machine learning.
“Such initiatives should deﬁnitely be managed at the most senior level,” says David Halliwell, director of knowledge and innovation at Pinsent Masons, a global law ﬁrm headquartered in London. “But it’s not just about good management. It’s also about understanding at a senior level what AI and ML can and cannot do.”
Fewer Fast Learners than other organizations suﬀer from a lack of strategic clarity about machine learning. And fewer are plagued by organizational resistance to change. The reason may be that machine learning is viewed as more than a tactical tool for simply automating away costs and people.
Some Fast Learners, for example, will certainly generate cost savings from workforce reductions: 30% strongly agree with a statement that productivity improvements will enable them to reduce headcount and another 31% “somewhat” agree that this will be the case. But 50% of Fast Learners ﬁrmly expect to retrain employees to perform more interesting and higher-value tasks and another 25% agree but with more moderate conviction. Most Fast Learners (76% altogether) also expect that some employees will be moved to other roles having similar skill requirements as their existing jobs.
Given Fast Learners’ far-ranging plans for integrating machine learning into the organization, it’s not surprising that cost savings is low on the list of beneﬁts they are reaping from the use of machine learning.
Indeed, Fast Learners are seeing a broad range of beneﬁts from ML, which signals that they are focusing on its transformational possibilities in both front and back oﬃce, revenues, and cost centers. Intel, for example, is using ML to improve cycle time and quality in its products and to reﬁne its sales oﬀers based on predictions about customers’ needs. Meanwhile, other Fast Learners report improved process accuracy (reduced frequency of errors) as well as increased speed across a range of processes. Indeed, speed has been more of a beneﬁt than cost savings.
2. Increased competitive differentiation
Fast Learners don’t see ML as a classic technology implementation, one that focuses on incremental efficiency gains. They see it as a way to stand apart. Thirty-one percent of Fast Learners say ML has benefited innovation of business processes or the business model.
The legal profession is a good example of how some organizations are looking to bring about fundamental, rather than just incremental, change with ML. For example, Pinsent Masons is contemplating using ML to completely restructure its relationship with clients. “The bigger opportunity is to become providers of knowledge-based systems to clients, which means moving from a services model to a product model – licensing your knowledge rather than providing services by the hour,” says Halliwell.
Firms in other industries plan to use ML to help change their revenue models in similar ways. UK-based Ocado, an online grocery retailer, created its own ML-based logistics platform for automated warehouses that it plans to license to other retailers.
Cliﬀ Justice, principal for innovation and enterprise solutions at KPMG, a consulting ﬁrm, believes machine learning’s potential in business model innovation is enormous. “AI and ML impact the business model in a much more signiﬁcant way than cloud or any of the disruptions we’ve ever seen in our lifetimes,” he says.
3. New revenues and profitability
Forty-eight percent of Fast Learners cite increased profitability as the top benefit they have gained from machine learning. Fast Learners have also realized that ML can have a positive impact on new revenue streams. Nearly half of them (48%) expect revenue growth of more than 6% in 2018–2019. In contrast, only 30% of other ML users who have not yet begun to generate benefits anticipate growth of this magnitude.
Intel is using ML and predictive analytics to identify its revenue opportunities more accurately. It has created a new ML-based sales platform that “helps salespeople interpret what is happening in the marketplace and better focus sales oﬀers to customers,” says Aziz Safa, Intel’s chief data oﬃcer. “This has generated signiﬁcant new revenue opportunities and increased our hit rate on revenue growth targets.”
4. Key processes close to home
Fast Learners are already spending more today on business functions sourced locally, whether performed in-house or externally, than they are in low-cost regions. Organizations still awaiting machine learning benefits, on the other hand, spend considerably less on services close to home.
According to the survey, 58% of Fast Learners say they spend more than half their budget for business processes locally, compared to 39% of other ML users. This implies that Fast Learners have an edge over others in keeping their most strategic processes close to home.
This trend is expected to continue over the next three years. However, this doesn’t necessarily portend a sudden surge in reshoring. “Functions that are heavily offshored today may not necessarily be reshored in big waves, but when it comes to newer transactional processes, decisions are more likely to lean toward keeping them onshore,” says Arjun Sethi, partner at the consultancy A. T. Kearney.
For Fast Learners, this means that important decisions on sourcing priorities will no longer be based solely on cost. ML means that business relevance and customer value will increasingly take precedence.
Take Intel, for example. The company’s move into new markets and new products has produced increasingly complex sales engagements. With limited resources, business leaders at the company faced tough decisions about whether to focus only on certain customers and perhaps seek external vendors to manage more sales and marketing processes. In the past, this would have led to a classic cost-based outsourcing solution.
Rather than reduce coverage or outsource for additional help, however, Intel uses its ML-based sales platform to let employees cover a much larger number of accounts more eﬀectively while supporting the full sales cycle, Safa says. The platform analyzes large volumes of customer data and then applies reasoning to imitate the actions of a sales agent, initiating actions, such as creating customized sales oﬀers, to increase revenues. The system also delivers recommendations to online sales agents, and it includes a self-learning mechanism to improve its performance over time.
Stanton Jones, director and principal analyst at ISG, a technology advisory ﬁ rm, supports the view that ML will be a major driver of outsourcing decisions, however gradually it plays out. “These technologies will have a profound impact on the way that organizations decide what to outsource and who to outsource to.” Organizations will increasingly choose a do-it-yourself model of building their own internal ML capability, he adds, which will reduce the need for long-term outsourcing agreements.
5. Enterprise-Wide strategy
Fast Learners tend to look at what ML can do for their business in a holistic way. For example, more Fast Learners than organizations yet to realize benefits (36% vs. 26%) are implementing ML initiatives enterprise-wide – an approach that is more likely to benefit from synergies across different functions. More of the yet-to-realize-benefits group, on the other hand, are pursuing localized ML initiatives (47% vs. 39% of Fast Learners), which are often driven by individual business units or functions.
The Fast Learners’ broad approach to ML could help explain why 41% say that its use is translating into higher levels of customer satisfaction. Fast Learners have done more than other organizations to integrate ML use into key customer-facing and product development functions, such as contact centers, marketing, data processing and analytics, and R&D. In each of these functions, the Fast Learners’ integration of ML in business processes is considerably more advanced than that of the rest.
Is ML the great leveler?
The world’s technology giants may be leading the AI and ML charge, but organizations of all sizes have unprecedented access to online machine learning innovators and cloud-based computing power. These may help small businesses’ classic advantages of speed and entrepreneurialism count for more.
Fast Learners are diverse in scale, with the largest share (38%) falling into the midsize range, with between US$250 million and $750 million in revenues annually. Another 31% are smaller, with revenues of between $50 million and $250 million. This indicates that, when it comes to leveraging ML, size does not necessarily equate with success.
Large scale can actually pose some diﬃculties when it comes to business process change. For example, more of the largest organizations in the survey – those earning revenues of $750 million or more annually – cite organizational resistance as an ML implementation challenge than do smaller organizations. A reluctance to share data with external partners is also more common among large organizations than smaller ones. This matters because few organizations have the in-house expertise to develop and perfect ML techniques alone.
Scale does confer some advantages on businesses in their use of ML. Those engaging in proprietary ML research have greater cash and human resources to support R&D eﬀorts. And big organizations often have the computing power on hand to store and analyze the huge volumes of data needed for ML to do its work.
However, thanks to the cloud, smaller organizations now have access to substantial computing power at a fraction of the cost of maintaining such hardware on premise. They also have access to wide bodies of external AI and ML knowledge through roughly a dozen open source innovation platforms that big technology organizations and research institutes have been creating.
When it comes to ML, thinking big by no means requires being big.
Better learn fast
Based on the lessons from the Fast Learners, it’s clear that ML is more than just another technology wave. It is an integral part of a larger strategy to remake the business for competitive advantage.
Organizations that began testing the waters a few years ago are now moving ahead at pace and scale.
Moving forward, the gap between Fast Learners and the rest will widen. Further delay in putting an ML strategy into action will almost certainly spell trouble down the road.
It’s time for all organizations to start thinking about how ML ﬁts into a strategy for digital transformation.
Tips for starting the ML journey
Organize an ML boot camp. Plan training sessions for your executive committee to help business unit heads understand how ML can help grow the business.
Identify external sources of ML knowledge. Heads of business processes should canvass open innovation platforms where expertise and ideas are shared about applying ML techniques. Have them also gather and analyze examples of other organizations’ ML initiatives.
Pilot, but not for too long. The ﬁrst ML initiatives should be piloted in small sets of processes where risks are relatively low. Once proven, the scope of ML techniques should be steadily widened across business processes.
Manage the message. Have your organization’s marketing and communications teams produce a handbook for directors to use to answer internal questions about why ML is being adopted and what it will mean for their teams.
Review sourcing practices. Long-term oﬀshoring arrangements will need to be reassessed for business relevancy. Create a task force to understand which processes should be localized after ML applications are implemented.