Machine learning is not a magic bullet, but it does have the potential to serve as a powerful extender of human cognition. In B2B and B2C businesses, this capability is proving to be particularly useful in identifying patterns across large swaths of customer and user data and helping drive better company outcomes: more influential content creation, a larger number of paid converters, saved marketing costs, and the list goes on.
In this article, we spotlight four machine learning vendors and explore how their applications are being used by four B2B companies to help drive business decisions and better reach and serve their customers.
Three B2B ML Case Studies Up Close
1 - GlaxoSmithKline (GSK), a global healthcare company developing pharmaceuticals and healthcare products, has used Luminoso’s natural language and text analytics technology as a non-invasive solution for gaining insight into parents’ growing concerns about vaccinations. They applied algorithms to help sift through and tease out patterns of parent fears, including presumed links between vaccinations and autism, as a primary driver for avoiding vaccinations. GSK used these detailed insights to help create new informational content that specifically addressed parents’ fears and provided incentive for following through with early childhood vaccinations.
2 - In the gaming industry, Wargaming has leveraged Cloudera’s advanced analytics platform as a solution for processing over 500 million daily events (equivalent to 3 TB of raw data), enabling real-time recommendations to better engage users and present relevant offers. A smart and more robust infrastructure allowed Wargaming to run 10 times the number of campaigns, resulting in up to three times the rate in customer responses.
3 - ZenDesk, by now one of the most popular CRM platforms in the sales and service industries, was looking for a solution to better target audiences ready to purchase their products. They felt their audience was too broad and led to excess costs for pay-per-click (PPC) and search engine marketing (SEM) leads. Using MarianaIQ’s social media engagement platform, the company was able to identify patterns in contact data and use the platform to help create categories of personas. ZenDesk claims that its lead volume increased by a multiple of four and effectively drove down cost-per-lead.
The Role of Man and the Role of Machine
Though all AI applications are unique, the specific processes completed by humans and machines are relatively consistent across industries. There are some tasks that machines cannot yet do (or do well), and other tasks that humans cannot do (or cannot do at scale), and companies selling AI services usually talk more about “what the technology does” and less about “what the humans still have to do to get the technology to do anything useful.”
Artificial intelligence tech vendors always want to describe the “done-for-you” experience of their product. “The humans just push go, and all this hard work is done by the machine, it really involves almost nothing from the humans at all.” This makes for a compelling pitch, but is almost never the truth.
In this section we’ll talk a bit about what the humans do and what the machines do, and how the two need to work together to glean any real business value.
Broadly (and this is a gross simplification as I don’t have 8,000 words to work with in this article) machine learning and AI can be seen to perform two core functions:
- Pattern recognition at scale: Identifying images (autonomous vehicles / other visual applications), identifying user behaviors on page (recommendation engines, various personalization applications), identifying most likely products for an up-sell (retail or eCommerce applications), etc. Much of the value of machine learning is identifying patterns that humans can’t detect, and doing so at scale.
- Process automation at scale: Whether it’s filing paperwork, buying and editing social media advertisements, answering simple customer support questions, or queueing up the best leads for a salesperson to call, the simplification and streamlining of individual business processes is one of the core benefits of AI.
Humans also have an essential hand in helping prepare and direct algorithms to solve pre-defined problems. The machine is the engine swimming through data and spotting mathematically-based patterns in text, actions, etc., but it’s the human part of the team that is doing the following:
- Goal directedness: Determining when and where to apply effort or technical tools (AI or otherwise). No machine learning application will automatically know what goal is strives for, nor will it be able to walk into your office, integrate into APIs, train your teams, and begin making business decisions (not yet, anyway). “What are we optimizing for?” and “How should we use this technology?” are questions currently answerable by humans.
- Distillation: Congealing insights from across the company, determining the meaning behind the results of various tools (AI or otherwise). Putting these insights into context of the business and the customer. A machine might be able to optimize a result, but humans will have to make sense of that data and - in many cases - make the relevant strategic and tactical considerations for that data.
In order to make these concepts clearer, we’ll use the example of the Luminoso use case listed above (which readers can explore in greater depth on the full case study page).
First, the humans (team members of Luminoso and GlaxoSmithKlient, in this case) likely were doing the following:
- Goal directedness: Humans at GSK determined what they wanted to find out, and where on the internet they were likely to find sentiment from consumers online (in this case, forums). It’s likely the team prepped the machine up to find certain keywords and provided some guidance on what patterns it wanted, before the machine started searching for prevalence of words, correlations, etc.
- Distillation - Humans determined what the uncovered patterns meant—i.e. which correlations indicated biggest fears associated with vaccinations.
Machines, on the other hand, likely were responsible for the following:
- Pattern recognition: Luminoso’s technology was tasked with finding specific terms (in this case, terms related to GSK’s drugs and the conditions that they treat), and find related clusters of words and phrases that might help inform the team about the opinions of their customers and potential customers (in this case, by crawling large message boards). Finding specific words - and their related and “clustered” terms - is the very definition of pattern recognition.
- Process automation: Both the crawling of the message boards and the repeated displaying of this information within Luminoso’s interface / reporting would in this case be considered “processes” automated by the technology.
ML Applications Helping Businesses By Getting Up Close with Consumers
The overarching theme throughout these case studies is the use of AI and analytics in identifying consumer behavior patterns and trends, ultimately helping drive business content creation and marketing efforts. While specific business customer objectives may differ — better marketing strategies, optimizing the purchasing lifecycle, influencing behavior change for the “greater good”, etc.—all deal with tracking and analyzing customer behavior on a more micro, personalized level.
While the idea is not new, AI and ML technology are allowing industries to execute on this behavioral information utilizing big data, and doing so in real-time and at scale and for far less costs and amount of time.
It’s essential to take note of where the benefits of AI and ML technologies end, and where human cognition takes over. In other words, while businesses are now able to uncover patterns and drivers of business that were all but invisible to the human eye before the emergence of these technologies—a veritable golden egg for customer-centric businesses—humans still serve an essential role in the equation: the ability to take action based on that information, apply strategic problem-solving and, develop an empathetic understanding that goes into consumer-driven behaviors and decisions.
In the statement above, the term “empathetic” is used in better understanding a person or group of people—the why behind a certain behavior or process—and in turn using that knowledge to design a better product or service according to the customer’s needs and desires. While machine learning excels at identifying and categorizing anything that can be quantified, humans still rule in interpreting and understanding how quantitative and qualitative data fits into a greater context. For example, an algorithm might observe that a particular group of consumers waits a certain number of days after scanning flight tickets to purchase, or that a certain group searches for particular flight routes during a specific time of the year. But the why behind these behaviors is something that can only be found out by contacting and listening to customer’s stories.
Concluding Thoughts on ML Use Cases
When it comes to creating better products and services, understanding the why of customer actions and reactions is essential and still not something that machines can tackle on their own. Companies must have a clear understanding of the problem that needs to be solved and frame the right questions for developing a better understanding of a problem, which—in addition to preparing and training the machine algorithm— sometimes includes humans reaching out to humans, and openly listening to concerns, desires, and motivations. This is why “subject matter experts” are often completely necessary for solving problems with machine learning and AI, and it’s a reason why humans still serve (a broad contextual understanding of a problem, a business, or an industry at large).
ML and AI technologies are helping shape expectations on both the business and the consumer side, and if we take the best benefits of both parts of the equation—machine-analyzed patterns alongside human understanding and strategic decision-making—then it seems we have the potential to do a better service for both sides. A ML-optimized approach also yields an ever-more competitive marketplace, and it seems safe to say that companies and organizations who don’t eventually make use of AI and ML technologies in tandem with human talent may find it hard to stay in the game.
I hope that the real AI-in-business use cases above - coupled with the breakdown of what the humans on the team needed to do to gain value from it’s application - have painted an informative picture of what AI looks like when it’s put to use.