Most businesses recognize that machine learning can generate exceptional value but many still wonder how, in what specific areas, and if the time is right to integrate it within their data strategy. Today, we explore what questions you should be asking to know if machine learning is right for your business.
First off, is the time right for machine learning? Or is it a buzzword and marketing-driven tech bubble?
The time is now and here is why.
Look at all the signals out there:
- Advancements in computational power: increased performance in microelectronics are enabling businesses to process more data. With more and more of them switching out their CPUs for GPUs, time-consuming processes (like neural net training) have seen a significant increase in speed. Market leaders like Nvidia, Google, and Microsoft are even working on hardware specifically used for machine learning.
- The cost of storage: according to research site Statistic Brain, costs have dropped from $105,000/Gig in 1985 to $0.03/Gig in 2014. More storage for your dollar means more businesses are keeping and recording data without worrying about space.
- The rise of data: ScienceDaily wrote in 2013: “A full 90 percent of all the data in the world has been generated over the last two years”. Today, we can expect that percentage to be higher. With that much data at hand, potential machine learning applications become countless.
- Market size and growth: marketsandmarkets.com’s latest report states that “the artificial intelligence (AI) market is expected to be worth USD 16.06 Billion by 2022, growing at a CAGR (compound annual growth rate) of 62.9% from 2016 to 2022”. If you know about a market segment that exceeds the estimated CAGR for AI, please email me. I would be curious to find out.
- M&A Activity in Artificial Intelligence Up 7x Since 2011 is a sign of increased competition holding clear value propositions within the market.
- It’s everywhere: “Ten years ago, we struggled to find 10 machine learning based business applications. Now we struggle to find 10 that don’t use it.” as stated by Gartner’s research vice president, Alexander Linden.
- Machine learning isn’t new: Its origins date back to the 50s. Bernard Marr provides a nice overview of the machine learning timeline in his article: A Short History of Machine Learning — Every Manager Should Read.
As Raj Reddy, Former Co-Chair of US President’s Information Technology, put it back in 1988, “the field is more exciting than ever. Our recent advances are significant and substantial. And the mythical AI winter may have turned into an AI spring. I see many flowers blooming.”
With all this in mind, I think it’s safe to say the AI winter has passed, spring is here, and summer is on the horizon. Reddy isn’t the only one to have noticed the flowers blooming.
What business problems can machine learning solve?
To help you identify what situations can be tackled with machine learning, start with your data. Look for areas in your business that are capable of producing data (in large quantities) and what value can be derived from it.
Machine learning is different than other technological advancements; it is not a plug and play solution, at least not yet. Machine learning can be used to tackle a lot of situations and each situation requires a specific data set, model, and parameters to produce valuable results.
This means you need a clearly defined objective when starting out. Machine learning is making considerable advances in many fields and all functions within an organization are likely to see disruptive advancements in the near future. Nonetheless, some fields are riper than others to pursue its adoption.
At Arcbees, we believe there are two functions in particular that are trailblazing businesses’ adoption of machine learning:
- Logistics and production;
- Sales and marketing.
The reason why these two areas are leading to way to a more widespread integration within daily practices is simple. They promise a direct influence on ROI.
What ROI can you expect from machine learning?
Measuring the ROI of a machine learning investment within your business strategy is highly dependent on the context and the problem being solved. Most gains from its use can be categorized within two major fields: predictive insight and process automation. Both of which can be used in ways that can either lower costs or increase revenue.
- Predictive insight into customers’ behavior will provide you with more opportunities for sales;
- Anticipating medicine effectiveness can reduce time to market;
- Forecasting when a user is about to churn can improve retention.
Within this context, machine learning has the potential to increase your reactivity by providing you the tools and information to make decisions faster and with more accuracy.
Process automation and efficiency:
- Augmenting investment management decisions with machine learning powered software can provide better margins and help mitigate costly mistakes;
- Robotic arm movement training can increase your production line’s precision and alleviate your needs for quality control;
- Resource distribution according to user demand can save time and costs during delivery.
When machine learning is used in this type of context, your business becomes smarter. Your processes and systems augment your value proposition and your resources are used more efficiently.
As more companies move towards its integration within daily business practices, more case studies measuring its impact will be made available.
What do I need to start using Machine learning?
Whether your investment concerns either predictive insight or process automation, certain conditions can favor your success. TechEmergence polled 31 experts of the AI industry and three conditions for success emerged in order to obtain an observable return on investment: sufficient data, qualified talent, and picking the right problem.
Machine learning is dependent on data. Essentially, other than the software and talent to select and train the models, all you need is coherent data in sufficient quantities as dictated by the problem you are trying to solve.
As easy as this might sound, it is not.
Data entirely generated through software processes (think Ads at Google for instance) are generally coherent, but when you have people involved, things get trickier. A shift in mentality amongst the people involved in the process of producing data is usually required. Within some organizations, using artificial intelligence involves as much change management as it does new technology.
To produce intelligible data, consider your ability to collect it, its source, the required format, where it is stored, but also the human factor. Both executives and employees involved in the process need to understand its value and why taking care of its quality is important.
For example, in a business tracking warehouse inventory, a virtual mismatch due to human manipulation is possible: the software thinks that there are 10 jars of honey in stock, but there are actually only 9 (or they are in the wrong bin, etc.). This might not significantly impact daily operations, but it could impact data quality. Comprehension on how daily activities can impact data production, for all people involved in the process, is the first step towards better data practices.
Originally published on Arcbees blog.