PR is critical for startups. The profession is always been one part relationship manager, one part storyteller — a sales and marketing gig rolled into one.
But with billions being poured into artificial intelligence (AI) and machine learning over the next 30 years, the tech PR industry is next in line for a serious AI upgrade.
For the scoop, I reached out to PR and marketing tools vendors and PR agencies big and small.
Here are the big takeaways:
According to Jeff Hardison, VP at Lytics, a customer data platform and a major player in analytics, “Machine learning is already helping marketers make more efficient use of customer data, and complementing what they’ve had for centuries: intuition and experience.”
The keyword is “complements,” not replaces, he said.
Machine learning today. Tomorrow, AI
“Two years ago or even last year the technology that could calculate and develop the algorithms weren’t as efficient or open as they are today,” said Saif Ajani, CEO of Keyhole, a hashtag analytics company that uses Google Tensor Flow to tap into the capabilities of machine learning.
“Google and Amazon have put up AI clouds dedicated to machine learning just in the past 12 to 24 months,” he said. Keyhole can now plug its huge data set for social into the AI cloud and get results for clients quickly and affordably. Our data is 80 to 90 percent accurate in predicting what is going to happen after 30 days. Within 3 days we can predict how big the trend is going to be after 30 days,” he said.
The next #DeleteUber PR nightmare can be avoided
“What brands or PR agencies fail to understand is how big a crisis is going to be,” continued Ajani. “The #boycottunited crisis started in January, and it’s now September and still happening. The challenge for PR agencies is how do you know what’s going to go massive? Within 24 or 72 hours we can actually tell you how massive a crisis is going to be in the next 30 days.”
Soon vendors will be able to combine hundreds of different factors and billions of social posts to make predictions with an incredible degree of accuracy. Once your data tells you that a crisis is going to be massive, your agency will be able to confidently put the brakes on pre-scheduled posts and respond more appropriately to the situation, unlike what happened with #deleteuber and #boycottunited.
Some PR agencies are lagging
Surprisingly, several multi-national agencies I spoke with do not currently use machine learning or AI with their clients, and they asked not to be mentioned in this story.
That’s not good news for them.
Shift Communications, however, is ahead of the curve and poised for growth.
Shift uses AI and machine learning in predictive analytics, text mining, and advanced attribution. During a recent client crisis, Shift was able to crunch more than 15,000 content-rich blogs for a medical client in just 1.5 seconds to identify insights, trends, and keywords in hopes to identify the root cause of a situation. Through the process, Shift was able to uncover an entirely different reputation issue that the client is now able to address.
None of that would have been possible without machine learning and large data sets, says Christopher Penn, Shift vice president.
“We use a blend of open source software [and] custom code, with the help of their partner, IBM Watson,” Penn said.
AI will touch everything marketing pros touch
BuzzSumo, the powerful tool that allows any user to find out what content is popular, uses machine learning in a number of ways across its various products.
“Our primary use [of machine learning] is classification of articles into topics and extracting phrases from questions to help classify the question,” said Steve Rayson, director at BuzzSumo. “A simple example is that an article may be classified as being about e-learning, even if it doesn’t mention elearning specifically.”
BuzzSumo estimates that about half of its 3,500 paying companies are PR and social media professionals involved in content production and promotion.
Rayson continued, “Machine learning works well with large data sets and helps us with problems such as classification. It also helps us to identify common elements of content that gains shares and links. … More generally, we are seeing the benefits of machine learning using large data sets such as in translations, image recognition, and spam detection.
“We have a database of over 6 billion content items and add over 100 million new content items each month. We use machine learning to help us classify content and to rank content. For example, we crawl hundreds of thousands of forums to identify questions being asked on any topic and we extract phrases to group these questions into sub-topics,” he said.
“It’s really about focusing on the problem you are trying to solve.”
For AI, size of the data matters
As a leading platform for PR media measurement and attribution analytics, AirPR crawls, processes, and analyzes billions of data points per day and uses natural language processing (NLP) and deep learning techniques to “teach” its systems to understand text and classify articles, as well as determine relevance and influence for any brand.
“We use AI and machine learning to improve filtering of the data our customers have access to, removing the majority of spam and non-relevant URLs that distort business impact reporting,” said AirPR CEO and cofounder Sharam Fouladgar-Mercer.
AirPR’s own content marketing and PR teams leverage the machine learning capabilities of the AirPR Analyst platform to research which topics and story arcs are trending for its target audience. The platform provides deeper insight into the articles, authors, influencers, and messages that drive actual engagement with a customer’s brand.
Another company to pay attention to is Trendkite, an emerging leader in PR measurement and analytics.
“Artificial intelligence and machine learning are engrained throughout the TrendKite product,” a Trendkite spokesperson told me over email. “With machine learning, Trendkite is able to provide insights and recommendations based on reviewing more data points than humans could possibly reason over. This enables us to do things like distinguishing earned media content from other types of content so that we can better attribute business outcomes to PR efforts.”
Social tools are leading the AI charge
Hootsuite sees itself as innovative in data science, which also requires a large amount of data.
Hootsuite uses machine learning for social marketing, social selling, and social support. A product to watch is Hootsuite Insight, which tracks mentions and brand sentiment.
“We figure out trends around your brand, which allows us to identify spikes,” said Mik Lernout, vice president of product at Hootsuite.
“Hootsuite Insights consumes a lot of data to create meaningful results for our customers. A lot of that initial work is pretty run-of-the -mill data crunching. The data science and machine learning is used when Insights classifies and prioritizes the relevant data for a specific customer. Out of all of the social conversations about my brand: Are those positive or negative? What are the topics and themes that keep coming up? What are the real-time trends you might need to act on? These kinds of questions can only really be answered using data science and machine learning,” Lernout said.
Expect more PR apps with machine learning
Finally, I chatted with Audrey Mann Cronin, the founder of Say It Media, Inc., who invented the mobile app LikeSo. LikeSo is a personal speech coach that helps you eliminate annoying verbal tics. Cronin says she’ll be adding machine learning and artificial intelligence during her next round of funding.
Personally, I could see so many uses for this app by PR and especially executive communications professionals, as well as for anyone who wanted to improve their speaking ability. The app is included in the next syllabus at the University of North Carolina in Chapel Hill by communications professor Rick Clancy.
So pay attention, PR veterans.
The next wave of PR tech is here, and hunches and experience aren’t going to cut it anymore. Will robots take over PR jobs? According to Willrobotstakemyjob, there’s only an 18 percent likelihood of that ever happening.
A version of this post originally appeared on VentureBeat; reprinted with permission.