How Artificial Intelligence Does Half My Work
by leading social media consultant, Alex Sass
A leading social media manager has created a tool driven by artificial intelligence to cut his own workload in half, having studied 90 million Facebook datasets to create “instant persona” technology for small and medium-sized enterprises (SMEs).
Social media is an astonishing industry that I’ve worked in for a very long time. When I started in 1990, Mark Zuckerberg was still in high school, and it would be 16 years until Twitter was launched. It wasn’t even called social media back when my team launched some of the first technologies to host personal profiles and forum messaging. Our first real online community was driven by what we called an online social engine, and with 50,000 users, it was one of the biggest in the world.
Now, with more than 2 billion active users on Facebook alone and more than 70 million pages created by businesses worldwide, my little hobby has become more of a behemoth. My own statistics are crazy too, having moderated more than 1 million individual posts, hosting 150 events, and traveling the world three times over to talk to people about the theories, practices, and applications available for corporate belonging in the new world.
Within this chaotic and exponentially expanding business, certain things have remained the same throughout my career. Whether I’m guiding global brands or start-ups, the primary challenges their social media teams face are a constant. Every social media manager I’ve met claims to be both under-resourced and over-exposed. What they mean by that is they face the daily challenge of creating or curating new content on deadlines that would be inconceivable in traditional marketing, while also having to filter and make sense of the massive noise that social media creates.
Curating content that aligns with the voice of a brand and the needs of an audience is the single most important skill a social media manager can have. With most of their time spent on this each day, every minute counts, and each post is instantly measured in terms of engagement, meaning wins or losses are rapidly exposed. Back in the day, social media managers wouldn’t receive much of a brief. They would use their natural empathy to translate a two-dimensional brand into an entity that communicates and shares. Today, we do the same, either through constant crawling of content that might resonate with or influence our fans, or a reliance on a whole suite of automated technologies.
Gone are the days when we’d stay at our desks through the night to ensure a post was made at the right time. Now, we have Hootsuite and Buffer for that. Measurement has also come full-circle with Facebook’s own tools and the overlords of reporting, the ever-powerful Sysomos or Brandwatch. However, the one area that still lacks a genius approach is sourcing the post itself. Some companies have whole departments dedicated to listening to trends and creating unique stories. The rest of us still rely on Reddit, a good old Google search, or listening to competitors to find those diamond pieces that will stir our audience into conversation.
That’s why I created PostDiva. When playing with AI technology a few years ago, attempting to use neurolinguistic programming to predict fan behavior, I stumbled upon something incredible. Systems such as IBM’s Watson could be trained to study the fans of a Facebook page and turn the data into a usable, 51-trait personality/needs/values chart. Other AI systems like the Alyien New API could be taught to understand the numbers and return content from thousands of sources, automatically matched to the profile. In essence, I had found a way to link instant audience persona creation with the ultimate content library. To create PostDiva, all we had to do was make it easy to use.
The first part of the social media journey, where one must understand both the brand and the audience, requires immense amounts of time and investment. I’ve known brands that would “listen” for years before they would be confident to step in with a vocal page. They’d host workshops, forums, and study groups to decipher the noise their customers made.
That strategy works, but if you don’t have a six-figure budget for that sort of thing, I’ve found that it’s also possible to use my software to group the online population into a set number of studies and then, in only a few days, gather months of data from those samples, enough to create an entire online persona. All in all, we needed only 30 studies, each with a few thousand people, to be confident enough to represent almost any brand in the world, altering only key differences such as age, lifestyle, ethnicity, or interests. From these 30 studies (which contained 90 million datasets in total), I could produce any brand persona, simply by mixing them.
With all that time and money already saved, enabling social media managers to spend their budget elsewhere, we found we could also speed up content creation in the same way. It’s easy enough to select topics for content, but it makes more sense for AI to do it for us. Having tracked 51 personality traits for each of the study groups, all the machine had to do was collect enough content samples to learn how to match the personality of the piece (the topic, approach, viewpoint, etc.) with the personality of the reader. My machine turned the audience into one person, the brand into another, and found common interests between them. A perfect relationship was born. On launch day, PostDiva was already able to support an infinite variety of brands and drive content from 10 thousand trending stories, ordering them by both the likelihood of engagement and the intricacy of “point of view”. The content curation battle was won — game, set, and match.
I know other firms have produced content solutions too. Indeed, we began negotiations to partner with some of them. However, we differ in several small but important aspects. We believe the social media manager has a unique skill that cannot be taught, not to another person, nor a machine. That skill is in taking the ingredients of a story as provided by software and knowing how to reword it to suit both the brand’s intention and its long-term strategy. We also use AI to prepopulate posts for our customers, but it’s up to the human to select which is best. I’ve always felt a social media manager should be able to adjust their brand over time. With new product releases or additional markets opening, it’s necessary to return to the persona and make alterations on-the-fly. We allow that — our persona settings are both informed by our own research and available for constant tweaking.
So, what else sets us apart from other content curation methods? As a social media manager myself, I knew our system had to be fast. Sometimes, I want to consider five or ten different posts to publish each hour. Other times, I’m looking for one killer story to focus on for a whole week. The software I use has to keep up with me. PostDiva gives me five preselected and alterable content feeds for every brand. That gives me a full 360° vision of not only the audience’s core interests but also their sub-hobbies, coordinated lifestyles, or even their favorite sports — things I could never know without the machine telling me. All that is required of me is to turn it on and work with it to achieve whatever content goals I have that day — and I know it’s working when my engagement rates soar.
Although PostDiva can handle any form of social media content on any platform, we decided to focus on Facebook primarily. Having that one process, that one game in mind, made us focus on every minute part of the journey. For example, we knew that the average page on Facebook has a moderately receding fan base in many sectors, and we knew that the cost of Facebook boosts is increasing, while click-through rates are inverting. Meaningful, highly-tuned content addresses both symptoms, giving the social media manager the means to take on the organic battle without spending even more money on ads. To put some numbers on it, the price of those ads, as an alternative to generating organic content, has risen around 35% in only 12 months, with resulting impressions only gaining 10% in the same period. It’s an issue that can’t be overcome just by increasing your budget.
Finally, in my 20 years in this business, I’ve learned one more thing that has gone into shaping PostDiva. There’s a myth that “social media people” are high earners. It’s true that some consultants can charge astronomical rates for a single day’s work, but the vast majority of us do the job because it thrills us. Here in Spain, where I relocated a few months ago from London, the average annual wage for a social media manager is around €21,000, despite some of the tools designed to assist with the role costing in excess of €16,000 a year. This might seem a little crazy when you also consider that more than half of us are freelancers and managing multiple brands, without access to big company assets. So, I priced my tool at £20 a month, simply because we should, even though, so far, we’ve spent a decent percentage of £1 million developing it. My reasoning behind that choice is that the social media industry needs to readjust if it is to provide a sensible employment option going forward.
So, there we have it. It took around three-and-a-half years to get from first ideas to a working website for the public. Most of that time was spent in the “study” stage, learning how posts aggregate into personalities and how they can be aligned with authored content or news. We left ourselves only a couple of months to actually launch the system. Hopefully, the end result is clean, easy to use, and flexible enough to adapt to however you work. Please forgive any bugs as we go live. Like social media itself, something that can look quite simple on a page often has incredible complexity underneath.
www.PostDiva.com launched on October 29th
Created by Oregon Technology Ltd (UK & Spain)
Notes for editors
Some of my stats come from these amazing places:
- Lack of resources the biggest issue.
- Aligning with the voice of the brand.
- Curating content, pre-populating the post.
- Sysomos can cost more than $2,000 per user, per month. Even SproutSocial costs $99.
- PostDiva was developed concurrently in Bristol, United Kingdom and Arrecife, Lanzarote.