ONLINE WEBINAR : How AI changed the way I handle Qualitative data [October 16, 2024]

Moderator: Jiten Madia, Founder, flowres.io

Experts:

Introductions

Jiten: Hi everyone, I have Lauren and Yogesh with me. Yogesh is founder of Y2S Consulting and he talked about some really cool custom GPT tools he built at Y2S. Lauren has been an avid AI tool user in her past jobs and she's great at converting feedback into actionable results. As for myself – I’m a qualitative researcher turned entrepreneur; focused on service and product domains of qualitative research. We’ve now built a tool flowres and we have been in the transcription service space as well, through myTranscriptionPlace.

AI tools – Current usage

Jiten: Can we start off with a quick poll… apart from ChatGPT, how many tools have you used and for what? So, Yogesh Lauren, what do you think of the results?

Yogesh: I'm not surprised with the results, to be honest with you. Lauren, what do you think?

Lauren: No, it's just makes sense.

Jiten: Yeah, everyone is in an exploring phase, right? Let's dive in. So let's start with you, Yogesh and Lauren. What AI tools that you have tried so far and what's your overall experience with AI?

Lauren: I think you're really getting two very different angles with Yogesh and myself. I'd probably say you're getting my perspective as more of a lay researcher using AI as opposed to a real expert who is not just using, but also creating. I consider myself at this point a little more than a dabbler, having used a range of AI tools, both pure LLMs, I'll say just, from GPT of course; to Claude, which I think is my personal favorite. I even have a name for Claude. I call Claude, Claudeette…. who I really view as sort of my ‘thought-partner’. I also am a big fan of Perplexity. Native research platforms that I use, range from, let's say, discuss.io's Native AI, or Recollective or any of those embedded in some of the research tools that I'm using to conduct either synchronous or asynchronous work, as well as tools like flowres, CoLoop, Yabble, easythemes. There really is quite a range of difference in terms of both functionality and accuracy and just enjoyment frankly in using them and the results that they yield. I'll stop there. Yogesh, over to you. 

Yogesh: Thanks, Lauren. If I was to add a couple more names, I’ve been playing with text-to-video tools, which is a big enabler for market researchers trying to communicate complex information that you can't necessarily do through text or just images by themselves. The last one I'll mention is Meta’s Lama 3.2… pretty sophisticated in terms of its capabilities. Even if you put in simple prompts that relate to market research, it gives you back some fairly sophisticated answers.

AI tools – Possible use-cases in Market Research  

Lauren: I probably should have led with this… as a researcher, the way I'm viewing AI and its role in the work I do is quite simply – How can it be applied in the pre, during, and post data-collection phases. And from our poll, it was clear that the majority of people here use AI for back-end analysis. But there's tremendous value in using it in that pre and during phases as well. 

AI tools – Addressing data-privacy concerns

Jiten: So what advice do we give to people who want to use AI tools and seek client permission to do so?

Yogesh: Yeah, so this is a pretty new terrain for everybody, right? There are companies that are very much into, you know, wanting vendors to use AI tools. There are others who are uncertain about many things. And then there are a few who will just outright reject it. So it’s a diverse spectrum out there. For companies that are into it and wanting you to use it; as a vendor, you want to make sure that you are doing a couple of things that protect yourself and your client.

• 1.     Put some legal language in your Master Services Agreement (MSA) that talks about your usage of AI tools
• 2.In particular, be clear about permission rights eg. does the client give you permission to input information into any AI tool for you to use on any platform or whatever platform you intend to use.
• 3.    Be clear about privacy rights. You're dependent on these platforms in terms of what kind of security and privacy they offer. You can't control that, but you can certainly share that and If the client is still willing to sign up for that, great. If they're not willing to sign up for it, then you know what your limits are right there.
• 4.     You also want to make sure that you are specifying your IP parts that you are sharing with your client, while retaining the rights. So for example, I've designed custom GPT tools on ChatGPT… I retain that IP, it doesn't belong to the client. In a way, I'm licensing that access to the client, even though I may be doing all the work. You want to just be clear about that part of it as well.

Lauren: I think you have a better chance of getting client compliance and buy-in with those tools. Because I feel like maybe those walled gardens are just a little safer. 

Yogesh: Yeah. So, when you're dealing with the user information, whether it's a quantitative survey or qualitative… you can actually do something prior to inputting anything into any AI tool. The first thing you can do is you can anonymize the names.

Jiten: As a tool provider, I can tell you that most tool providers are GDPR-compliant and check all the standards. Because even for our own certification, we have to go and check various things. So I think data security is generally not a problem, if you are using a paid version.

ChatGPT vs specialized AI tools

Jiten: Let’s move gears a little bit… there are so many tools and then there is ChatGPT. I think everyone starts their journey at ChatGPT. So, what's your experience on using specialized tools versus ChatGPT? And if there are times when you prefer ChatGPT or vice versa?

Lauren: I mean, at this point, anything can give you a summary of what was said, what we heard, etc. What matters is having the knowledge, wherewithal and comfort-level to go way beyond that, to ask creative and artful prompts to try to just get deeper and really play with some of that data and massage it. Another that's really been interesting and fruitful is to partner with AI tools as a reviewer. So I might write and create and do my top line and my summary and all of that. And then ask for comments and feedback… “What did I miss” kind of a prompt. Because we know as researchers, sometimes there's the recency effect there just are innate biases that we bring to the table.

Yogesh: When ChatGPT announced that you can create a custom GPT tools, that's when I started playing a little bit more in terms of creating a structure and system that helped me automate certain things. Let's take, for example, taking a transcript. You can summarize the transcript, that's great, but that's only one part of your job, right? The bigger part of the job is, can you actually identify an insight from the transcript? But then the question becomes, how are you defining an insight? So when I designed one of my tools, I actually gave a very clear definition for what I wanted to see as insights and what was not considered to be an insight. So when you uploaded transcript, It will identify insights accordingly.

AI vs Human role in insight-generation

Jiten: I think that's one of the things, right... AI is basically designed to give answers, not necessarily the right ones, right? And then this becomes a problem in the research reports when, so you, I think if you're saying, you know, familiar with data, where you have done the data yourself, I think that makes sense. But yeah, I mean, generally, I think the onus of checking things still lies on the researcher… because you’re putting your name on the report.

Lauren: I mean, there's no question… I think human oversight is essential. None of us here are cut-pasting from AI into a report; without truly digesting, processing, making it our own, turning it into your voice. I was cautious in the beginning, but have gotten over some of the fear that AI is encroaching upon or stealing, taking our jobs, et cetera. Very early, I bought into this idea that if we do not learn how to work with, then I think we will be left behind. I find that I'm much more creative when I'm talking with AI.

AI and Powerpoint

Yogesh: I was just starting to say that there's a learning for all of us here as whether you're a qualitative researcher or an independent doing quantitative research or whatever, which is that traditionally we have relied on PowerPoint as our primary way of communicating back to the client. our output, right? But with these kinds of tools that are available today, you suddenly have a broader suite of tools you could use, whether you can use the NotebookLM to create like a podcast, you could use, tools to text, video tools to create videos Those were things that were hard for us to do two years ago, but now they're, now they're doable today. So that's a big shift in terms of what we can do or can't do.