Machine learning and cognitive computing are on track to become some of the most widely applicable technological movements we have ever seen. Together, they are set to be primary drivers of the fourth industrial revolution and will cause social revolution on a global scale. Individuals and organisations must begin preparing for this eventuality.
During the digital revolution, those who were slow to adopt, or those who eschewed the internet, suffered severely because of their indecisiveness. Since then, lessons have been learnt, and, more than ever before, society has become aware of the effect of technological change. We can easily see this technological revolution unfolding around us, and its advances are likely to occur at a surprising rate. Now is the time to educate, to consider how the applications of AI impact business, to experiment, and to understand how AI should be effectively progressed into production.
This does not mean retraining to become a data scientist. But at the very least, organisations should learn about the practical applicability of AI. This is especially important for business leaders, many of whom don’t have the appropriate knowledge to effectively identify applicability to their business. More importantly, this knowledge provides leaders with visibility into how technology advances will change their operating market, as well as the impact their employers, clients and competitors.
Organisations that fail to understand the impact of AI - regardless of industry sector - risk going out of business.
Experimentation should be a fail fast, no regrets undertaking, fed by hypotheses from across the business. While machine learning often requires a significant amount of data, leaders should not delay experimentation because their data is unclean and fragmented. Instead, an experimentation capability that is empowered to obtain data from across the business should be formed.
This capability should be able to utilise the tools they feel most comfortable with, enabling them to try new technologies and approaches. The major public cloud platforms offer an excellent environment for this with ephemeral compute and tools like Azure ML Workbench and AWS SageMaker, alongside a raft of excellent pay-per-use APIs. These tools have also been designed with the data science process in mind, building in the ability to promote production more simply.
Opportunities should be created for development and experimentation teams to work together. This exposes engineers to the data science process and data scientists to engineering rigour - an area of common complaint when progressing to phases beyond experimentation.
Those already experimenting with AI are discovering a vast range of potentially valuable applications. Unfortunately, in many cases, their ability to prove these out in production is limited by enforced dependency on slow-moving enterprise-wide initiatives; data and cloud strategy and data lake programmes, for example. Businesses should be cautious when undertaking such initiatives, particularly when it comes to the manner of their implementation and enforcement. The same can be said of poorly delivered cloud strategies that are simply ‘worked around’, thereby increasing security risk, among other issues. Removing these dependencies until the appropriate point is key to demonstrating the value of AI in production environments.
The progress of AI and its effect on society is exciting, but a measured approach is required. While many forward-looking organisations are already experimenting with AI, it is crucial that they do not create a backlog of successful experiments that cannot progress. They can achieve this by expanding their views and modifying their approach beyond experimentation. Those organisations which succeed in doing so will benefit greatly at the expense of those who do not.
