Why do I bracket AI and Automation together?

AI and Automation word cloudPosted by Ray Poynter, 14 June

I often group AI and Automation together, when I am talking about changes in how we work with information and insights – and that seems to surprise or even offend some people.

The people that protest at my views assert that AI (artificial intelligence) and Automation are clearly quite different. They stress a) Automation can be done without AI, and b) AI can be used for ad hoc tasks that are quite separate from automation.

However, at the moment the importance of AI and Automation are, IMHO, inextricably linked – and in this post I will outline why.

Throughout history, automation has been the driving force behind reducing costs and turning niches into widespread phenomena. Here are a few examples from history, William Lee’s stocking frame knitting machine in 1589, Jethro Tull’s seed drill in 1700, James Watt’s steam engine in 1778, Henry Ford’s car assembly line 1913, Alan Turning’s code breaking computer 1942, and the sequencing of the human genome in 2013. Most recently we see the use of computational immunology to generate vaccine candidates, only 6 months after the first identification of the coronavirus. In all of these cases, innovation was used to turn a niche into a widespread phenomenon, via the mechanism of automation.

Automation has changed research
Over the last 70 years, the history of the growth of market research has rested on automation. In the early, face-to-face days, the automation relied on printing processes for creating questionnaires, data punching and verification innovations, and eventually computerised techniques for tabulating data. As telephone data collection emerged, the automation of call centres revolutionised the process, reducing costs, increasing quality, and speeding up the delivery of results. With the development of online data collection, the role of automation has been the driving force, in data collection, analysis, and the creation of results. The breadth and affordability of research today are entirely dependent on the automation that has been achieved – but most of it had nothing to do with AI.

AI tends to be hard to define
One of the tricky things about artificial intelligence and machine learning is that they tend to be hard to define. They are also a bit like magic, in the sense that when they are explained people tend to say ‘Oh, is that all it is, it’s not Magic (artificial intelligence) at all’. Cluster analysis is an example of unsupervised machine learning, but we tend not to call it machine learning (we call it cluster analysis), satnav is a great example of AI (but we tend to focus on its errors not the successes), Siri and Alexa are amazing examples of AI – but they have become commonplace, so people keep saying when will AI arrive, but people should look around them and see just how much AI is in everyday use.

If we try to focus on AI and its impact on market research and insights, we can easily become side-tracked into discussing what is and what is not AI. For example, whether the use of auto-transcription is truly AI, or whether the use of a report generation system is real machine learning, or simply a list of conditions and dictionaries. However, if AI is to have an impact on MR it will be via automation.

The Future of Automation is Linked to AI
AI without scale is interesting, it can reveal great insights, but it is as limited as any other specialist technique (e.g. semiotics, structural equation modelling, or ethnography). Almost the only thing that changes market research is something that can be applied at scale, and in terms of AI that tends to mean via automation.

Things we expect to be able to do via AI include:

  • Coding open-ended comments
  • Testing questionnaires
  • Designing samples and checking the quality of data
  • Analysing survey data
  • Analysing text, images and videos
  • Writing reports

But these will only make a difference when AI delivers them via automation. There will be AI that in the hands of an expert allows a complex problem to be addressed. In these cases, the solution will be relatively slow and expensive, so they will have little impact on the wider world of insights.

There might be some automation without AI in the near future, but not much. Most things that can readily be automated without AI have been, and if you are automating a new process, why wouldn’t you embed AI?

AI and Automation are Symbiotic
For the foreseeable future, all useful Automation will use AI, and all significant AI will be useful because it enables automation. So, we can discuss AI or we can discuss Automation – but if we are talking about impact, we are talking about the interaction of the two.

Join Our AI and Automation Webinar – 17 June 2021
Click here to register for the webinar (or to access the play again if you read this after 17 June)

NewMR Survey about AI
NewMR are hosting a short survey about the impact of AI on market research. Click here to take part in the survey.

If you have already taken part in the survey, click here to see the results.