Data science, engineering and AI are all increasingly important for the future of communications. These technical skillsets and tools are revolutionising the way society engages with the world and our industry is not immune to those changes.
Still, for many PR teams, big data can seem miles away from the day-to-day.
So why should you start getting excited about data, machine learning and AI?
Let’s start with the statistics. According to SEMrush, 75% of executives fear going out of business within five years if they don’t scale AI. Digging into that, marketing and sales also prioritize AI technology and machine learning more than any other department.
Altogether, we have a tremendous opportunity to think about how we work alongside the broader marketing discipline to really earn PR’s seat at the table. To collaborate on data projects that span the marketing mix, or more simply to work with our clients more closely to understand the data they are collecting and how we can bring it to bear for the challenges a particular brand or corporate is facing in a particular moment.
Not to say that we’ll do this alone. Partnership with broader technical partners is going to be key too. There’s no sense in running in front of a speeding train – there is tremendous venture capital money being put into tech and AI companies. The question is how we structure those partnerships such that they are truly collaborative and supportive of the needs of our clients and industry.
Presuming you’ve bought the premise, let’s get down to the brass tacks. How can these technological revolutions make a meaningful difference to the analytics and PR work you’re doing, now?
Tools like Brandwatch’s Iris have been around for a while now, but there are multiple examples of simple machine learning and AI which help to surface trends within large data sets. You might not even realise you’re already making use of some of these tools. These might be supervised learning models doing supercharged regressions or more advanced AI. Ultimately these can be very helpful in understanding large swaths of data quickly.
Who, what, where
Machine learning and AI can help to classify a data set for you. Where you might have previously approached the problem of understanding how a brand or individual was mentioned across coverage using Boolean keyword searches, entity training can sometimes offer better ways of understanding the who, what, and why at scale because these models are trained to recognise the ‘entity’ not just by keyword but context. Anyone who has ever been tasked with monitoring a brand with a generic name (Apple, anyone) will appreciate why that matters.
Tools of the trade
Cluster analysis and other unsupervised learning models are some of the most exciting applications to me, personally, in the PR context. These tools can help identify the underlying structure of your data (rather than searching for pre-ordained outcomes as in the previous examples). Given the importance of insight to most analyst job roles, and the fact that PR operates in a space far less defined than some of our marketing colleagues, I hope you’ll agree they are quite interesting. It’s why we’ve built Space+, which makes use of Quid’s NLP to examine the interconnections of our audiences and thousands of online news stories. We then apply our own custom methodology to identify where a brand has competitive advantage.
So there are a few examples of how Machine Learning and AI are already helping PR practitioners do better work. But I’m going to push a bit further and say that data needs to do more than just tell us what’s already happened.
Or perhaps better put: we need to be doing more with data than stating what’s happened.
There is so much occurring in the world of prediction. Where we have feature-rich datasets we can – and should – be doing more to deliver better results to our clients. So I’ll leave this post with a few key questions we’re thinking about when it comes to doing more with machine learning and AI.
First: Where can we leverage machines to help us make better / faster / more accurate predictions? Where are we repeatedly asked the same question? Do we have tasks that currently require tonnes of manual hours which are largely predicted by past events? If that’s the case, machine learning or AI can probably do that job, better.
Second: What do we need to do to make the data we hold appropriate for the task? Or what data would we need to acquire to complement the data we already have? I think PR feels this more acutely than other disciplines as we haven’t had the historical rigour of collecting data. But we can, and we should.
Lastly: Where do we need the interaction between human + machine to occur? To quote Martin Schmalz of Oxford University – Marketing isn’t physics. We want to predict well with the help of machine learning and AI and have predictions that can be (imperfectly) explained with the help of behavioural science. It’s why we are continuing to integrate our analytics, data and behavioural science offering. And I can’t wait to see what we build next.