I love a good fairy tale as much as the next guy. This explains why The Princess Bride is my all-time favorite movie, with Casablanca a close second. While fairy tales are great for movie plots, they don’t belong in marketing research. Yet a number of posts in this space present just such a rose-colored worldview when they start talking about automation and artificial intelligence. Take, for example, Taejin In’s post on 13 February 2018. He writes, “In a world of growing data, it’s time to embrace AI as a tool to free up market researchers to do what they do best – using intuition and expertise to extract deep human insights and provide real value.” Therein lays the fairy tale.
My post is not about whether AI or automation is good or bad; it’s not how one should describe either of them as a domain. We can, however, discuss whether they are good or bad for the research industry. The assumption to date has generally been positive on the score – so I’ll present the other side.
Let’s take automation first. Researchers (and unfortunately non-researchers) have the capability to open a piece of survey software, click on a drop-down menu for the type of survey we want (new product, pricing, etc.), put in some keywords (like the category, the product name, etc.), and voilá – a survey appears ready to go. If you did this with a panel provider’s software, you might be offered instant fielding of the survey – just enter your sample specs and off you go. This is great for the novice who has no other resource to call on. You can automate the backside too – there are packages/services that will automatically analyze your data and write your report. Yes, they know nothing about your business, but so what – that’s where you, the analyst comes in. You, the analyst, edits the report and fills in the business knowledge. The best part – this is cheap.
While this may be great for the user, as long as they are willing to accept a black-box system, it’s terrible for our industry. We will not be teaching young researchers how to write a good survey. We will not be teaching young researchers how to statistically analyze data. We will not be teaching young researchers how to put together a story for management based on the data. Why not, you ask? Because automation can do away with the need for all these lower-level researchers (no offense, junior people)! A company faced with the option of freeing up people from survey construction and fieldwork and data cleaning and analysis could let them become analysts. Or, more likely, they could let them all go as superfluous and get a better return for their shareholders. After all, we have people more senior who do much of the analysis anyways. The automation fairytale is that this will set you free – the reality is it significantly reduces staffing needs, which will eventually reduce qualified researchers.
Artificial Intelligence is a little more complex to dissect. Zappistore thinks of AI as an expert system that can script, recruit, collect data, tab, analyze, and create a visualization. In this conceptualization, AI is automation and will also eat into staff. If we conceive of AI as independently “discovering” something in a big data set or finding consistencies across media that are too hard for an individual to find, then AI changes the nature of what we do and may or may not affect staff.
The fairy tale that automation and artificial intelligence will set us free ignores the downsides to our industry. We will lose much of the need for entry level staff, or entry level staff will now need to start with senior-level skills just to get hired. We will give up the training ground where we taught these senior level skills to junior level staff as we adopt automation and AI. Remember – Grimm’s fairy tales were very dark and scary.