There are few technologies that companies are more excited about than AI. It has the potential to completely reshape how companies operate across functions, including marketing, customer service and finance. But as with many emerging technologies, there are challenges, and AI presents no shortage of them. So what's standing in the way of AI realizing its potential?
There are few technologies that companies are more excited about than artificial intelligence (AI), and for good reason: AI has the potential to completely reshape how companies operate across functions, including marketing, customer service and finance.
But as with many emerging technologies, there are challenges, and AI presents no shortage of them. That might explain why, according to a new MIT-Boston Consulting Group survey, 85% of executives believe AI will change business, but only 20% of companies are using it in some way, and just 5% make extensive use of it.
So what’s standing in the way of AI realizing its potential? Here are five of the biggest challenges that companies need to address if they want to start making effective use of the growing number of AI-powered tools that are available today.
Data is the lifeblood of the digital economy and for companies looking to apply AI to any number of areas, access to data is going to be one of the biggest challenges. In fact, according to George Zarkadakis, digital lead at global advisory firm Willis Towers Watson, data is going to be the biggest challenge companies face.
“To train machine learning algorithms one needs massive and clean data sets, with minimum biases,” he told AI Business. “One needs also to keep in mind data privacy issues when it comes to harvesting personal data, particularly in light of the General Data Protection Regulation that is coming in effect in 2018.”
The good news is that most brands have been keen to the value of data for years. Thanks to the ad market in particular, companies have recognized the value of first-party data, especially in light of the increasing cost of acquiring third-party data.
As a result, many companies have been investing heavily in creating the infrastructure to collect and store the data they generate and to recruit talent capable of making use of it. Those that are further ahead in this area will find that they have a competitive advantage in integrating AI into their businesses.
Even when a company has ample data available on which to create AI applications, it’s important for them to recognize that the models their AI applications are trained against won’t necessarily work forever.
Take, for example, AI applications that are used to manage marketing campaigns. Last year, IBM announced that it would be using its Watson platform to manage all of its programmatic campaigns by 2017. According to reports, IBM reduced its cost per click on average by 35% using Watson and in some cases, that figure went as high as 71%.
As AdAge explained, Watson “uses advanced analytics to create efficiencies in the bidding process by ingesting massive amounts of data and assigning value to potential target consumers based on the time of day, what device they are using, what language they speak and what browser they are using.”
The level to which Watson can analyze data is “mind boggling.” For example, it can look at “whether a smaller size [ad] is more effective when shown at 3 a.m. for a $2 CPM, or cost per thousand impressions, than larger ads at noon at a $3 CPM.”
But the digital advertising market isn’t static and models that have worked for months or years aren’t guaranteed to work tomorrow. While AI can learn as it goes, its ability to do so depends largely on conditions remaining similar to those it was trained on.
Changing ad formats, the coming and going of buyers in the ecosystem and an increase in the number of companies employing AI to buy ads, for example, all have the potential to dramatically change market conditions, so that they’re very different from those that existed when the data the AI was trained against was gathered.
This means there are risks that AI models will decrease significantly in efficacy or break quickly, causing harm, so smart companies will probably always need make sure oversight and safeguards are in place rather than trusting the business to AI.
Companies are increasingly looking to employ AI technology to support their customer service efforts. For example, many are building AI-powered chatbots that customers can interact with on platforms like Facebook Messenger.
While early incarnations of chatbots for these platforms left a lot to be desired, natural language processing (NLP) technology is rapidly improving and AI-driven bots are getting better at understanding what the humans they’re interacting with are saying.
But even so, AI applications lack emotional intelligence, and most importantly, they are unable to demonstrate empathy, and this is a huge barrier to AI success in customer service applications such as chatbots. After all, certain customer service inquiries can make or break a customer relationship.
One way brands can address this challenge is to limit the application of AI to customer service where empathy isn’t necessary. Chatbots, for instance, can be designed to serve as front-line customer service, responding to frequently asked questions and handling simple, generally low-emotion requests. Where requests are more complex or potentially sensitive, AI-powered chatbots should be able to smoothly connect customers to human customer service representatives.
David Raab, principal of marketing consultancy Raab Associates, has noted that “AI systems of today and the near future are specialists.” They perform specific tasks, such as scoring a lead or determining the optimum price to bid for a display ad.
Of course, AI-powered technologies are currently better at some specialized tasks than they are at others. Take AI automated content creation, a dream of content marketers everywhere. By 2018, Gartner predicts that 20% of all business content will be produced by machines.
While there is evidence that AI is capable of creating certain kinds of content that is virtually indistinguishable from human content in terms of clarity and accuracy, machine-produced content is substantially more boring and less pleasant to read according to one study.
Since emotive content is critical to content marketing success, brands have reason to be wary about putting the entire task of content creation in the hands of AI software.
But that doesn’t mean that AI can’t perform specialized content tasks. Brands can use artificial intelligence to identify trends and topics that lend themselves to popular content, predict which human-written headlines will perform the best, or curate content.
An innovative example of AI-powered content curation was on display during the US Open this year. The United States Tennis Association (USTA) trained IBM Watson “to recognize player gestures and facial expressions, crowd noises and broadcaster reaction” and then use Watson to help its broadcast and content teams identify match highlights to deliver to fans.
As Raab Associates’ David Raab observed, a marketing campaign involves coordination of many specialized tasks, meaning that for AI to take over a full marketing campaign “will require cooperation of many AIs.”
In theory, this isn’t necessarily a deal-breaker. But theory and reality aren’t the same thing. He explained what’s involved in making this happen:
It’s easy – and fun – to envision a complex collection of AI-driven components collaborating to create fully automated, perfectly personalized customer experiences. But that system will be prone to frequent failures as one or another component finds itself facing conditions it wasn’t trained to handle. If the systems are well designed (and we’re lucky), the components will shut themselves down when that happens. If we’re not so lucky, they’ll keep running and return increasingly inappropriate results.
What this ultimately means is that it will be more complex and costly for companies to build the kind of self-driving marketing campaigns that AI promises. For that reason, in the interim, savvy brands will be strategic about which AI tech they invest in. For example, one company might realize significant value applying AI to lead scoring while another might realize more value applying AI to social media sentiment analysis.
Because the returns can vary so much depending on the brand and its needs, companies will realistically need to analyze AI technologies and determine which ones offer the most value to them.
