Digital-first companies like Google, Amazon, Netflix, Airbnb, etc. haven’t only disrupted, but downright transformed their respective industries. Consider the success of Uber, which revolutionized the taxi industry, reporting monthly trips in excess of 50 million, without even owning a single vehicle. How did they do it? This was accomplished by combining modern-day machine learning algorithms and analytics strategies with flexible business and pricing models, while offering unique benefits to both drivers and riders at the same time. In hindsight, it can now be safely said that the transportation service industry and a geolocation app are thick as thieves. Uber recognized this synergy and capitalized on it. It’s Economics 101: they created a dynamic demand/supply-based resource optimization and pricing model. 

Having said that, machine learning (ML) is no longer a prerogative or a secret ritual performed by organizations born and bred in the digital age. Indeed, there was a time when manufacturing, healthcare, telecommunications, and the oil and gas sectors would brush off the need, or even assessment, of machine learning and artificial intelligence technologies right at the outset. The times have changed!

Telecommunication, and in particular the cloud and data center networking sector, has had its own share of evolutions, with multi-vendor architectures, divergent platforms, varied business processes, and networks driven by distinct user requirements increasingly becoming the norm, rather than an exception. To enable a differentiated interworking and ensure profitable existence, a highly reliable, scalable, performance-optimized, and cost-effective solution suite is desired. Some of the near-term applications include network planning and optimization, assurance, control and management, network-related customer experience, etc. This is where ML-driven solutions can help augment and, literally, revolutionize the industry landscape. While savvy business leaders recognize the potential of these technologies, they are not poised to benefit from them. What’s lacking is a strategic roadmap, right from inception to integration, and eventual productionization of modern analytics-driven frameworks. 

Here’s a strategic guide to adopting and integrating machine learning solutions in businesses. (Occasionally it focuses on the networking sector, but in general it’s applicable to a wider audience.)

Strategic Foundations 

• Layout machine learning objectives based on internal opportunities to improve customer experience. How do you want the analytics-driven systems to behave? How do you intend to work with them?
• Invest in leaders with backgrounds in network design, cloud architecture and business practices, together with a solid grip on data analytics. These unicorns do exist, and their potential needs to be carefully harnessed.
• Elaborate a data strategy. While this may be initially based on your current IT practices, ML solutions necessitate completely different data management frameworks.
• The majority of ML is executed on open source initiatives. Developing an in-house tool chain and leveraging what’s publicly accessible is the only way to build and scale ML projects.
• Outsourcing parts of the analytics framework is smart, however a full-scale analytics subcontracting initiative is akin to handing a stranger the keys to your business.
• ML in itself is nothing more than a cool toy, unless deeply tied to a specific industry – with all its nuances. Think software architecture innovations and service-based business opportunities for a consolidated view.
• Real-time network insights, predictions, and eventually prescription capabilities are the flag bearers of analytics initiatives. A modular approach is required for progressive success.
• There are many pieces to the analytics puzzle, including ML algorithm skills, data science knowledge, IT expertise, etc., but domain-expert is central to this jigsaw and must steer the initiative to achieve quantifiable success.
• Investments in research and development and strategic M&As are essential to the adoption of analytics culture.
• Start small – applications like network optimization are low-hanging fruits with high impact factor.
• Evaluate ML-driven product and solution outcomes in the light of clearly identified criteria for success.

Executive Management Essentials

• Evangelize ML-first approaches to customers and cultivate new business opportunities by exploiting machine learning initiatives.
• Decision-making must evolve. Senior and mid-management must be adequately trained to augment their decisions using advanced analytics-based recommendations.
• Democratize the use of analytics – encourage open culture and set appropriate incentives. But be aware: this requires time.
• Analytics is not a singular product or even a platform; it’s a process change. Actively and openly communicate this within your organization.
• Value your subject-matter experts. All of the decision-making will come down to engineers and executives with domain knowledge to review, judge, and perhaps fine-tune the solutions addressing a particular problem.
• Separate IT service delivery from ML solution inception, development and deployments. CIOs are to be concerned with the former, whereas CTOs or even a CAIO (Chief Artificial Intelligence Officer) should focus on the latter.
• Communicate the change – ML will not kill jobs; instead it will transform them and fine-tune them for productivity.
• Do not perceive ML only as a fire-fighting tool to engage and retain customers.
• Openly aim to break the silos. Data and insight hoarding is analytics’ biggest nightmare and defies the fundamental foundations of analytics strategy.
• The results will not magically appear on the first day. Managing expectations and trusting in your analytics commitments is important.

It’s quite apparent that ML-driven solutions come with their own unique challenges, and the learning curve will most certainly be steep. Nonetheless, it must be stressed that ML will not create transcendent self-driven networks; rather an extremely productive environment, where expert input, and more importantly oversight, will be necessary. 

The long and the short of it is that corporations thrive upon innovations and consequent business and operational optimizations. While experience-driven decision-making will remain fundamental to any business, it must be organically augmented with real-time data-driven insights exposed by modern machine learning and analytics frameworks.

Setting aside machine learning, it may be argued that this has always been the case, the differences however are in the extent, pace, and more importantly the adaptability of these insights, which were previously impossible to attain and exploit. 

Indeed, it may sound daunting at first and it’s likely to necessitate numerous operational and business process evolution trajectories – dumb silos will cut organizations deep and hard. That’s why the time to act is now.