AI will change your insight job more than you think

AIThis post was written in response to a session at the MRS Conference in London where, earlier today, a panel was discussing AI (artificial intelligence) and seemed to have a mind-boggling degree of complacency about the impact of AI on market research and insight. To summarise their views, machines are good at repetitive tasks but can’t be creative, the jobs that will be absorbed by AI will tend to be those that are already automated, or those done by more junior staff, and those that are repetitive.

However, I think the role of AI will be much more disruptive than the panel seems to think.

Assumptions

A couple of starting assumptions:

  • Computers do not just do what they are programmed to do. Computers are increasingly programmed to learn, not to do – that is why we call it machine learning. Computers are already analysing data and writing reports and have been doing so for a while.
  • A large share of current market research is poor; it too often utilises a poor design, it uses a poor questionnaire or a poor discussion guide, the analysis is not deep enough and does not access all of the available information, and results in a less than impressive report or presentation.

AI will not just make the repetitive tasks faster and cheaper; it will also over a period of time, replace the worst (approximately) 75% of market research, leaving just the best as having a significant human component.

The future?

I predict the following will happen over the next few years (perhaps 5 to 10 years):

  • Project Design. Machines can learn the best range of designs for all the common research problems, e.g. tracking, concept tests, U&A, brand equity etc. The programs will ask the users of the research some questions and offer some choices (such as quant or qual – spelling out the implications of the choices in terms of time, cost, meaning etc). For most research projects the questions (e.g. the questionnaires or discussion guides) should NOT be written from scratch; there is lots of research on which questions work best, a machine will tend to use better designs, better questions, and be less likely to miss key elements.
  • Project management. The best way to handle problems is to spot them early, that is a great role for AI. Some of the problems found by the AI project manager will require a human decision, but that decision will be asked by the system and the options and their implications will be spelled out. However, in many cases decisions to increase the recruitment, to add new regions, to modify a question or probe will be made by the AI project manager without having to wait for human intervention.
  • Analysis. There are already systems that can interrogate data and write reports (check out Intellection for an example). With AI project design and AI project management the AI analysis will be even more powerful; for example drawing on wider information, taking things like cultural values into account, and looking deeper into the data than is normal. It is likely that the recommendations and interpretations generated by AI will be followed by a discussion between humans, but I suspect that in many cases more than 90% of the recommendations and interpretations will be used without being edited.
  • Communicating insight and recommendations. An AI approach to communication would factor in (in a formal and analytical way) what forms of communication work best with the message and the target audience. This would include small things like colours, spelling, reading age, but also writing style, idioms, analogies and story structures.

Winners and Losers

So, given my view of the future of AI, who will be the winners and losers?

The Losers

Here are some of the jobs/tasks that will change or disappear.

  • Project design. In the future most projects will not require a researcher to determine things like number of interviews, questions, locations etc. AI will generate a design and the end user and an advisor will sense check whether it is right. So anybody designing the detail of a project will see the task diminishing, the placing of fieldwork and/or recruitment contracts will be automated, as will most of the testing of the study.
  • Quant data collection. The placement of data collection contracts, the creation of the data collection instrument, and the monitoring of the data collection will be 99% AI.
  • Qual data collection? I think focus groups and IDI’s will be largely moderated by AI, but not for a few more years. However, semiotics, text analytics, social media will all become increasingly AI quite quickly.
  • Analysis. I see most analysis to be AI-led soon, with the crafting of the recommendations and interpretation falling a little but behind.

Winners

Here are the roles that will, in my view, prosper under the rise of AI.

  • Client success managers. These are people who get to know the client’s real needs, help shape the use of research, and help socialise the findings of research into the wider organisation. The people who will flourish are those with people, business, and AI utilisation skills.
  • Bespoke researchers. Some people want bespoke suits, some need them (if their shape is non-standard for example), similarly there will be clients who want and/or need bespoke research. The people who will flourish here are those who knowledge, presentation skills, and reputation mean that they can be more effective than AI research.
  • Entrepreneurs/Intrapreneurs. These are people who will identify new businesses and opportunities. These people will be ‘outside-the-box’ thinkers who can see new ways of doing things.
  • People creating the AI systems. This group includes coders, system analysts etc, but it also includes the people creating the best practices, the people helping the machines learn, and people creating the bridges from people to machine and machine to people.
  • The people driving AI systems. In order for AI to design a research project you need to know what it is you really want to achieve. For the foreseeable future the process of answering the questions from the AI systems will require humans, humans who understand what the questions are and who have a sense of when the answers the research user is offering are only partly correct. These people will also know when to override the AI, when to instigate changes in the AI solution, and how to best use the outputs.
  • Performers. This is a catchall term for writers, presenters, comedians, storytellers, artists, cartoonists etc who can bring results to life in ways that AI will not offer in the foreseeable future.

 My guess is that over the next 10 years 60% of current jobs in MR will go, and perhaps 20% will expand/emerge.

Note, these changes will also impact HR, sales, marketing, IT, finance and indeed every aspect of the orghanisation will also change, with the same sorts of ratios.

How does this picture compare with your thoughts about AI?

8 thoughts on “AI will change your insight job more than you think

  1. I could not agree more with your view. It is difficult to be very specific on the impact details though, but it is clear to me that the MR industry, once again, it is reacting very late to trends happening around it.., which makes me think that perhaps some of the big players are the first one interested in denying the reality as it will impact their business models in particular…

  2. This article is very much in line with my thought process right now. Researchers need to integrate backwards (learn the technology) and/or forward (become ‘bespoke’) to be ready for that future which is already happening now.

    Google can search billions of data points in secs, and test questions among millions of LIVE sample across all demographics; If I’m a client, I would really ask myself why I need to spend dollars and wait 4-6 weeks for humans to tell me what I can get from a few keystrokes.

  3. Another treat for poor researchers! 🙂

    Just curious, how the coming of the AI will impact consumer/sensory research?
    It is clear fact that people make their evaluations not only using their immediate sense but they give their responses using the whole sensory experience the have had during their lives (from the childhood etc.).
    This is what machines can not replicate yet. Or…

  4. All points are very valid, Ray, thanks for sharing.
    The Turing Test will be an online discussion board, or focus group or community, run by an AI. Honestly, I think we are not far from the time when an AI passes it (unless it already happened).
    As for reporting, I remember developing myself an automated sensory 60/40 test where the commentary was automated too (that’s easy – and not particularly intelligent – as you’ve got 2 options: difference or no difference and you randomly pick from a pool of sentences). I thought it as a spoof of market research, meaning that when research is conceived to tick a box, rather than adding value, then a computer can do it. Never had the courage to submit it to a conference to avoid a precocious end of my career; now I think I should have.
    I agree with your view that survivors will be what you call client success managers, as in most cases an agency is selected because of personal relationship rather than because of methodology, (the other driver being cost): I cannot see – at least for now – a customer insights manager talking over the phone with an AI and discussing their company’s internal struggles. I also agree that performers will survive, hence the importance of storytelling and presentation skills. This gets me thinking about young (wannabe) researchers. They have not developed any of those skills. Currently if you enter the MR arena you are head down into (manual) charting, running tables, data checking, qnaire writing, which a machine can do better. These first 3 years of your career will disappear. So, what will the career path of a graduate be? What will be the requirements to enter the industry? I think the industry should start re-thinking their skill development programs or we run the risk nobody will enter the arena.
    One point that also got me thinking is “people creating the AI systems”, in particular people creating the best practices, helping the machines to learn. Now a poor teacher will seldom have brilliant pupils. How many case studies are actually substantiated by extensive research rather than covering vested interests? I can see an AI crunching best practices and finding contradictions. I can also see an AI analysing retrospectively the history of thousands product launches and unveiling idiosyncrasies.

Leave a Reply

Your email address will not be published. Required fields are marked *