Introducing ‘Researcher-Driven AI’: Driving Agency-Quality Research at DIY Costs and Speed
Editor’s note: Editor’s note: This article is an automated speech-to-text transcription, edited lightly for clarity.
On September 25, 2024, Knit was one of three companies that presented on the subject of DIY research during Quirk’s Virtual Session series.
Knit’s Co-Founder and CEO, Aneesh Dhawan, unveiled the newest release of Knit's platform: Researcher-Driven AI.
Watch the session or read the transcript below to learn about this new offering and see how it was used in an actual study.
Webinar transcript:
Joe Rydholm:
Hi everybody and welcome to our session “Introducing ‘Researcher-Driven AI’: Driving Agency Quality Research at DIY Costs and Speed.” I'm Quirk’s Editor, Joe Rydholm.
Before we get started, let's quickly go over the ways you can participate in today's discussion. You can use a chat tab to interact with other attendees during the session and you can use the Q&A tab to submit questions during the session and the presenters will be answering them during the Q&A portion after the presentation.
Our session today is presented by Knit. Aneesh Dhawan, take it away.
Aneesh Dhawan:
Awesome. Good afternoon, everyone.
Joe, thanks so much for having us here today. My name is Aneesh.
Today I'll be presenting, on behalf of Knit, our presentation around researcher-driven AI, how to drive agency quality research at DIY costs and speed.
For those of you hearing about Knit for the very first time, we are an AI platform that helps researchers at some of the world's best brands run agency quality, quant and qual research at DIY Costs and Speed.
The way we do this is through our philosophy and technology of researcher-driven AI, which we'll talk about today as we walk through a use case of how one of our partners actually used their platform to get some really interesting insights around the topic of AI.
But before we dive in, I know we're all here to learn more about DIY tools and the evolution of DIY tools in today's AI native world.
So for most of us that have used DIY tools, they can be incredibly powerful tools to leverage as part of our research tool stack. You can get data quickly at a relatively low cost.
However, the downsides of DIY tools tends to be that you're getting really surface level insights.
In order to dig deeper and really get to the core of what's being said, you might have to spend a good amount of time doing the heavy lifting around analysis. That's why customers, that's why partners go to full-service vendors as well.
Full-service vendors can do a lot of that heavy lifting for you, helping you get to those deep and actionable insights. But that can take weeks if not months, to turn around and cost an arm and the leg. That's why for researchers today, there's a bit of a compromise when running research. You're always compromising between rigor, speed and cost.
At Knit using AI, we've developed a platform that allows you to run research without compromise. We think that AI makes it possible for you to get that agency level rigor in your insights, but at the speed and the cost of DIY. And today we'll walk you through how we do that.
But before we dive in, we also want to be realistic about some of the downsides that come with using AI and applying it to market research.
AI tends to fall short in a couple areas as well. AI tools today in market research can feel a bit out of the box. The insights that come out or the AI outputs can feel a bit generic and not really custom tuned to how you or your company runs the research.
AI can also feel like a bit of a black box. You don't really know how the magic is happening behind the curtains, how the AI is actually synthesizing some of this data and turning it into actionable insights.
And finally, AI tools can have a non-native experience. They can have an experience that's not intuitive. A lot of these AI tools don't really meet you in the way that you run research.
That's why at Knit we have developed a technology and philosophy called Researcher-Driven AI, and this is really what's powering our ability to help you run agency quality research, but at DIY costs and speed.
There's three key threads to researcher-driven AI.
The first is Knits AI. This is AI that is trained on our best practices when it comes to research. So it knows how to do things like helping you develop a questionnaire, an analysis plan in a top line report.
The second is our researchers. Our researchers act like any full-service agency research team. They really take the time to understand how you and your company uniquely runs research and make sure that's reflected in our AI and how it outputs.
And lastly, it's you. Our entire platform is designed around the researcher. That's why it's customizable. Each customer has a unique experience on knit one that is tailored to how they run research or how their company runs research. It's controllable. You are always in the driver's seat controlling the AI and the research team and how it's analyzing the data, how it's helping you with the questionnaire, how it's helping you build out the analysis plan and it's comfortable. The tool is designed to look and feel like the tools that you use and love today and it plugs into your research workflow so that it's a seamless experience going from our platform to the end of sharing this internally with your stakeholders.
So how does Knit work?
Well it starts off with our research team taking the time to really understand how do you uniquely run research or how does your company uniquely run research?
After that, we make sure that that's reflected in our AI, which then generates a questionnaire for you, an analysis plan program and fields your study. And within 24 hours of fielding, you have a top line that is stakeholder ready.
It has key recommendations; key action items and it answers your research objectives through the lens of a story. From there, you can further explore these insights and tweak these insights through our suite of different analysis tools.
At the end of the day, we are helping our customers run research two to three times faster by helping them get to that top line report within days, not weeks. We're helping them get to the what and the why and combine both quant and qual in one dataset and analyzing it as one dataset. So you can cut the quant by the qual and the qual by the quant.
We'll show some examples of that later today and we allow researchers to focus on what matters. You no longer have to spend time on incredibly time intensive tasks that might not move the needle for your research process and instead can focus on how do you take these insights and drive action with it in your business.
With all that context, I'm incredibly excited to walk you through a recent research study that we ran on the Knit platform that looks at AI usage on an AI platform.
It's pretty funny, it's a bit meta there.
What we'll walk through how our partner used Knit’s platform to actually uncover some really unique insights and how our AI helped them walk through that workflow.
So, before we dive in, let's talk a little bit about how the study was structured.
We talked to over 300 respondents. Our partner in this case wanted about a hundred of them to share video responses. So open-ended questions that they responded to with a one-to-three-minute video. Of course, if they wanted to have all 300, we could have done that as well.
The entire process from them coming on our platform and building out a research plan to getting that top line report. So, that includes fielding survey design programming took about four days from end-to-end and it all started off with them sharing their research objectives on our platform.
What they told us was a little bit about some of those key objectives, who they wanted to run research with and within 15 minutes they had a questionnaire designed for them.
Now this questionnaire was pretty robust. It included both quantitative and qualitative questions.
On the quant side, at Knit, we are compatible with over a hundred plus different quantitative question types that you can ask.
On the qual side, this included open-ended text questions as well as those videos that I mentioned earlier. One- to three-minute-long video responses to the open-ended questions.
We fielded this through the panel that we have access to. It gives you access to over 10 million consumers. We didn't talk to younger audiences in this study, but on Knit you do have the ability because we are COPA compliant to talk to consumers in the Gen Z Gen A generation. You can also upload your own panel or your customer list or you can upload existing data. So, maybe if you've already run research that you want to analyze on the knit platform, you can really easily import that into the platform itself.
As this study was fielding, what our AI also did was generate an analysis plan, which was quickly reviewed within a day by our dedicated research team. This analysis plan really drives the analysis that happens in the top line report and that's really where the magic happens.
Within 24 hours of fielding our AI and our research team have now reviewed this top line ready report that is ready for your stakeholders. It includes your key recommendations, key actions, answers, all of your research objectives through the lens of a story.
We'll walk through some of those insights from this study right now.
So, the first topic that we looked at in this study was looking around familiarity and favorability of some of the AI companies in this space.
We started off by asking what AI companies are people familiar with today?
And our AI went ahead and was able to code all of these companies. What's really interesting about this is in the research plan, our partner wanted to look at these companies through two different lenses; looking at more legacy major tech companies as well as some of the newer players on the block, companies like OpenAI, for example.
So, the AI was able to use that context and actually develop that insight based on these audiences that we marked in the research plan and the analysis plan.
So, you can see here that these companies were grouped in those two different audiences, those major tech companies as well as these newer AI specific companies.
The other thing is the AI was able to take into context, and actually grouped, the different brands under their parent company or different models under their parent company.
So, you see for example, Google and Gemini in that case, Google's AI model. And in Meta's case you can see the different brands that were kind of combined here as well. So, Meta, Facebook and Instagram.
After that, the AI followed up with some insights around some quantitative questions that we asked for some of those newer companies.
As you can see, the key insight here was that while OpenAI was mentioned a lot in the qual. From the quantitative findings, you can see that in terms of some of the newer AI companies, it is, the rest of them are significantly lagging in awareness. So, companies like Perplexity.AI and Mosaic AI.
The next topic we looked at was around trust in AI.
What we found here was that when it comes to privacy, privacy of your data, consumers kind of looked at those legacy companies and some of those newer companies and equally trusted them. With consumers kind of directionally slightly more trusting some of those larger legacy companies in the space.
When it came to innovation, you see a similar story here except in reverse.
What you see is that there's about an equal amount of trust here when it comes to trusting these companies in their ability to innovate. But some of these newer AI companies, consumers directionally trusted a little bit more.
Another topic that we looked at here was the usefulness of AI in a consumer's daily life.
We found that 80% of consumers found AI to be useful in their daily life. And to get a little bit more color around this insight, we had a video question.
This is an open-ended question that consumers could respond with a one- to three-minute-long video response. This is where our AI tools, especially our video AI tools, come in handy.
As you can see here, what the AI has done is it's actually transcribed all of those video responses to this question, and on the right here has actually generated a summary of the key insights coming out of those videos.
You can see that under the summary tab here, if you look really closely, there's also these green boxes, which if you click into, you can actually see the specific videos that are being pulled and being compiled into this summary. This prevents hallucination in how we are analyzing the data. You always know where the data's coming from because the AI is forced to cite its sources underneath.
You can see a breakdown in some of these key themes. And as you can see in these key themes, the AI actually quantifies these themes for you when it presents the key insight.
So, what we found here was that the top daily uses for AI was around quick info retrieval. So, think about searching on Gemini or ChatGPT chatbots and then finally voice commands as well.
Now if you drill into any of these themes, you can again see the specific videos that are being cited to bubble up that theme.
If you look into those specific videos, you can actually select the ones that you like the most, highlight the portions of this transcript that you want to include and very easily generate this really powerful sizzle reel or show reel.
What we've seen is this really helps bring those insights to life and personifies some of the quantitative and qualitative insights that our AI is uncovering from here.
One of the last topics that we looked at was around current and future applications of AI. What we found here was that consumers mostly found AI to be useful in phones, followed by computers and smart home devices. This is for current applications of AI, but when we look to future applications what we find is a pretty similar trend except for vehicles.
So, vehicles are actually at the bottom of the list for current applications, but they jump up to a second ranking here when it comes to future applications for AI and what gets consumers excited about what AI can do for them when it comes to their AI's impact on their daily life.
What we found is that it really revolved around productivity. You see this with productivity at work and personal assistance being two of the most common answers here, and personal assistance, again, looking at things like Google's AI assistant.
Now, what's really great about the Knit platform as well is you can drive the AI to also generate new insights for you.
So, a question that we asked here around AI's impact. Let's say you wanted to get a little bit more nuance there and generate an insight around a specific generation. So, in this case, we can ask the AI, ‘where do Gen Z respondents find the most value from AI in their daily lives?’
Within just a couple minutes, Knit will actually generate that insight in the form of a slide or a card for you. And you can see here the key insight that it generated from your question.
What's really cool about this is a qualitative question that we just applied a quantitative cut to. That's what is really unique about how we go about analyzing this data, because we're asking both quant and qual in one dataset our AI is able to analyze it as one dataset.
So in this case, we're taking a qualitative question, ‘where does AI have the biggest impact in your life?’ and applying a quantitative cut. But the AI could also do the reverse taking a quantitative insight and applying a cut through a qualitative theme.
This allows you to really get to that core of the insight and combine both quant and qual in one dataset.
With that, we'll open it up for any Q&A. We'll keep it short and sweet today.
Again, really appreciate you guys joining and learning a little bit more about how Knit’s researcher-driven AI allows you to get some of that agency quality rigor but at DIY costs and speed.
Turn it back over to you, Joe.