In this hands-on UX research workshop, join Joe Formica as he shows how the Lyssna MCP and two skills from the Lyssna AI skills library take you from raw study results to a concise, stakeholder-ready presentation in minutes. Joe connects the MCP to Claude live, installs the skills step by step, and puts them to work on real research.
You'll learn:
How to connect the Lyssna MCP, so your AI tool has instant access to your research.
How to install two skills from the Lyssna AI skills library.
How to turn raw results into clear themes backed by real participant quotes.
How to decide what to highlight, what to leave out, and what to recommend next.
How to shape your themes into a quick team update or a stakeholder-ready .report
Chapters:
00:00 - Welcome and introduction
11:42 - Getting set up: Connect MCP & install skills?
17:08 - Case study: Understanding bidder confidence
30:05 - Synthesize research skill: Turn responses into reliable insights
37:52 - Share findings skill: From insights to action
47.51 - Q&A
Resources:
About our host
Joe Formica leads workshops and creates educational content at Lyssna, bringing over 10 years of experience in UX and education. His practical, hands-on teaching approach has helped thousands of designers develop their skills and create better products.
Transcript
Hello everybody, what's up? Welcome. I see we got some people here already. Let's see. Megan's here, Miriam, see Hannah, Jenny, Pavo, Lizzie, Salma. Welcome everybody. Good to have you guys here.
For now you guys can just hang out, say hi in the chat, let me know where in the world you're checking in from, and just open this link right here. It might show that little password thing in the thumbnail, but you should be able to see this page here, which is going to be our resource guide. Basically, this is where I'm going to put everything from the workshop.
You'll see there are already a bunch of links and promo codes and stuff in there. I'll be uploading the recording of this here. So nothing to do with it right now other than open it up, make sure that you're seeing what I'm seeing here, and keep it handy as we'll come back. So pop that open and let me know if you're having any issues opening that. Otherwise, we will get started in just a sec. Cool So let's get started, everybody.
My name is Joe. I'm the Design Advocate here at Lyssna. I've been a product designer and researcher for quite a while. And as I'm sure since I see some familiar names here in the chat for those of you who've joined workshops before, you know the deal.
For those of you who haven't my role here is to help you guys be successful with your research initiatives in Lyssna, to get the most out of it, whether you're, working on a big research team or a research design, everything else team of one.
My goal is to help you get the most out of Lyssna, feel confident with it, and start doing some good research that will hopefully lead to some good product decisions. So today we're going to be covering the MCP and some skills.
We're going to get set up on all of those walk through some examples, and have you guys leaving here with a really... a tool to really turbocharge your research without compromising it. We'll get into the specifics of that.
So as I mentioned, everything is recorded for this workshop. I'll be sending it all out after the workshop. Keep the chat open and active. We never have a problem with with a dead or a quiet chat in, in these workshops, so I'm glad to see you guys are already in there. If you got questions, if you can't hear me, if I've got, food in my teeth and should be aware of it, let me know in the chat.
I got it up here, and I'll make sure to get to your questions, scenarios, feedback, anything like that as we go. Other than that, stay comfortable, happy and hydrated, and I'll walk you through everything from here. All right. So I got a quick kind of icebreaker. Might seem random, but this is going to be... This is going to come into play for our example, our research, our scenario that we're going to set up and use some of the MCP, some AI tools, and some of the skills for.
So all I want you guys to do right now is go over to this website, which I will put in the chat. It's auctions of cool cars. This is... We're going to be, we're going to be taking on the role of a design/research/product team for cars and bids today to walk through walk through some of these examples.
So all I want you to do right now is take two minutes. I'm going to put a real official two minutes on the clock up here, and browse a little bit. Find your next dream car and click on it, drop the link to it in the chat. We had our company offsite last week in Vietnam, and we did a hackathon and a bunch of other stuff, and I used this site as my example for a few things that that we were demoing and working on. And it was fun.
So I figured we'd use it again here today. Check it out and drop a link in the chat. We'll see who's got the, who's got the best taste in cars. Ooh, I like how the thumbnail just shows up. That's nice. We can see them in the chat. Sweet. Good cho- I like... Good choices. Oh, we got the Supra. Okay, that Volkswagen van is sick. I've never seen that before.
There's a lot of cars on this site that I've never seen before, so that's not as surprising. Oh, we got the Hellcat. This is what my husband wants. Porsche Taycan, that's a nice car. That's a fast car too. The '83 Lada Niva. Wow, two minutes. You guys came up some good choices. The M2. I can see, Dodge Dakota, a little red express pickup truck. Love it.
All right, I'm going to have to go through these properly a little bit later. I won't get too sidetracked. But thanks for doing that, and you'll see how this ties into, research and all the important stuff that we're going to get into here today.
I like the choices though. I'm going to have to browse these in a little bit more detail later. All right let's get into it. We're going to cover the Lyssna MCP today. We're going to cover quick little overview about what it is and what the AI skills are, and then we're going to focus in on a use case, get set up with the Lyssna MCP and connecting it to our LLM or whatever AI tools you're using, and walk through how a few of these specific skills work. Basically, we're going to get you set up to integrate Lyssna and your research with whatever AI tools you're using.
And I'm also going to share how you can get the most out of that, not just with, the technology and the tools that are built but also just from a research perspective, from a day-to-day perspective and how it can integrate into your workflow and how you can take things beyond what we do here today with some of these quicker examples. Okay. So MCP super quick.
I'm sure most of you guys are using or connecting stuff to your AI tools, your LLM. The Lyssna MCP is no different. Yeah, Kayla, I agree. We- you guys do have some fire taste. So the Lyssna MCP, as we'll see, connects to your AI tool, and it's got access to your Lyssna study.
So if you're running research in Lyssna, if you're doing usability studies, interviews preference tests, any kind of research that you're running, including, recruitment, it's going to connect to that and really open up a lot of new avenues for not only uncovering and I'd say discovering interesting insights in your research, but also just creating a new level of access to research.
Which I think is really important because a lot of the times where research kinda dies is in a Notion doc that no one sees anymore. MCP connects to your research and as we'll see, makes it very accessible and I'd say more useful than ever you can ask questions, right? As we'll see when you're connected to Lyssna, you can say, "Hey, have we ever tested the home page?"
And you can get answers for that. "Hey, when's the last time we ran a usability test on that onboarding experience? Hey, why did we choose this banner over the other one? Oh yeah, this was the research we had."
Or maybe get something back like, "Wait, that's why we chose it? We should do that again." So you're going to be able to ask questions across your research in plain language and get answers to it that's based in real research. And yes, these answers come from real participant data. I don't know how... I'm sure everyone here is at a different maybe level of experience using Lyssna.
I'm sure we got a couple of you here who've run tons of studies, are, pros in building, recruiting, reviewing results to studies. Maybe some of you here who have, are just getting started or who have not launched a single one.
But a reminder that this connector is really to just access data that is, real feedback from real people, either recruited by yourselves or through our panel in Lyssna. The Lyssna MCP is so fire. Yes, we got fire cars, fire MCPs going on. Amazing. So that's the MCP. Connects to your Lyssna account. You can type something into Claude and it'll tap into your research through Lyssna.
We're also going to cover some AI skills, which is something I worked on a lot put a lot of time and care into, and I'm excited to share. Skills, as I'm sure most of you have some experience with, give in general, give your AI tool guidelines rules, ways of working, and further refines how you interact with your LLM in certain contexts. And in this case, of course, we're talking about research.
So the leading line-- kind of the, the tagline and what I had front of mind was creating skills to give your AI tool a research brain. So we're going to cover some of these, but they're basically expert instructions that your AI follows for a specific research job built by researchers, I'm one of them. For researchers, designers, product people people who don't want a flimsy, hallucinated ChatGPT answer about their research, but really care about the speed, but not at the price of integrity or reliability of the research. And it's a free library that we're going to install.
I can talk about some of the other skills. There are five of them in there right now, and I'll sh- I'm going to walk you guys through how to install all of them. But we're going to focus on kind of the two big ones today, which are Synthesize Research and Share Findings. So basically connecting to the Lyssna MCP, that's going to allow you to access your research data.
The AI skills that we're going to set up are going to add that level of research expertise, know-how, integrity to how your, you and your LLM interact with research and deliver you answers that you can trust and make product decisions on.
So I already said this, but a reminder, since we're talking about a lot of AI stuff, and we're already into MCPs and skills and LLMs. Before we get too much deeper, I just wanted to pull back to remind you guys that this is all just... Everything we're going to cover today are ways of accessing and using research more efficiently and effectively, but it's all still from real people.
If you're using Lyssna and self-recruiting with your audience, if you're recruiting from the Lyssna panel it doesn't matter how many AI tools or skills or things that get us away from human interaction and human feedback even if those things are layered on top of it and the stuff we're going to cover here today a quick reminder that, the core is still real, reliable feedback from real people.
Something we prioritize heavily here at Lyssna, which is the quality of panelists and responses for your research. All right. Let's get set up. So I'm going to share a link with you guys, and I'm going to give you guys a little time to just get set up on this. Let me walk through it first.
Go ahead and click that link that I just dropped in the chat. It's a little kind of mini onboarding I made to to cover all the different scenarios of getting set up with the Lyssna MCP. So this is step by step. If you follow it, it will get you connected to the MCP, and it will install the skills, the ones that we're going to cover today.
Let me just go through it step by step here. Okay, so we got some just little overview, but I already gave this to you guys. Is it safe? Yes. It's actually not read-only anymore, but that's for some... We'll have some official documentation and announcements on that soon.
Here is where we actually start the setup process. So I tried to make this step by step because there are a bunch of different factors to make sure you guys get set up correctly with and connected to the MCP and the skills.
So just follow along with with the options here. If you don't have a Lyssna account there are some instructions to sign up for free and then as you'll see, a promo code to get a month free of our growth plan. If you do have a, you know, paid plan on Lyssna, your account is ready to go.
Now you just need to set up your set it up with the LLM that you use. I'm going to be demoing it with Claude here today. But you can-- There are all these little factors to that can disrupt getting it connected properly.
I'll stop talking through it. But basically, instructions for whatever scenario you're in, whether you've got no Lyssna account, a free account, a paid account, whatever LLM you're using I've got instructions here to set up the connector and the skills.
So let me leave it at that and let you guys work through it. I'm going to put let's put, Let's put five minutes on the clock just for a quick check-in. If you guys have any questions while you're walking through this and getting set up, let me know in the chat. If anything is not working for you, let me know in the chat.
If you've already got this set up then you can sit back and relax for a little bit. But we'll take five minutes for you guys to get set up ask any questions. I'll check in after that, and if we need some more time, we'll continue. How's everybody doing? Let me know in the chat.
Give me a, If you're set up, connected to the MCP and got the skills installed, let me know. If you need two more minutes, three more minutes, five more minutes, let me know. If you're running into any issues, let me know in the chat.
I want to make sure everyone gets set up on this. Yeah, let me know. Let me know what I can do, or if you just need a little more time. Cool. Okay, so let's see. Not the LLM owner. He's in a meeting, so I'll just have to observe today. Cool. Yeah, we will, I'll walk you through everything. I figured we would probably have some issues like that. I tried to make the smoothest possible clearest onboarding for everything, but still that can... Yeah, permissions, Claude accounts, stuff like that.
I'll put that stuff in the resource guide. But totally fine to just follow along today. I'm going to walk through it all on my account, and it should be a a quick fix there. But thanks for letting me know. Didn't see that option. Let me see if I can find a screenshot for you.
If you are up and running, let me know in the chat as well. The best way to... I think this is in here too, but best way to see if you're up and running is you could just ask your LLM if if you're connected to Lyssna. But if you've got studies in your account ask about them.
"Hey, can you pull up that preference test we did a couple weeks ago?" Start asking about your research. If you get something back, that means you're connected. Sweet. Download the skills. Got it from this link. Cool. And there is another place to download the skills.
I'll put the files, Yeah, all of these should be a quick, pretty quick fix, whether it's getting permission from the account owner or or I can share the files as well. If that's coming up for any of you guys, just hang tight and I'll- I will follow up with all that stuff. Connected.
All right, I'm going to give us we'll take two more minutes. Let me know if you guys run into any issues, and... Sweet. Good to see Amanda and Miriam connected. If you are connected, let me know. So let's do two minutes and then either way, we'll get into the important stuff, and you guys can follow along with me. Cool. Let's, Let's press on 'cause I got all this stuff ready to show you guys so we can look at it together.
You can follow along with me if you're set up. You can tap in, go crazy tapping into your own research. But we're going to get into a couple examples and gi- taking it for a test drive, showing you guys what and how MCP and skills can really contribute to a better research process. All right, so let's get into our scenario. Cars and Bids, you guys are already familiar with this site.
There's that nice... Oh, not the same Porsche up there, but a pretty, pretty sweet one. You guys have already done your shopping on Cars and Bids. We are going to use this as our example today.
So I'm going to transform from being the Design Advocate at Lyssna to, let's say, the, I don't know, researcher, product designer figure-outer at Cars and Bids, and we're going to use this as an example of a quick scenario that you can tap into using the MCP and some skills to highlight some of the features and functionality.
So first thing, like I always like to do with any kind of demo or walkthrough workshop is set a little realistic scenario here. Put us in the right frame of mind and set something up that might be a little bit simplified, but is going to feel, hopefully realistic and close to the stuff that we're working on every day.
We don't just go out and do research for the sake of research. Usually, it starts with something, a question, something that needs to be improved, and everything flows down from there. So that's what we're doing with this Cars and Bids example.
We're going to say that there's a problem, general business user experience problem. Too many auctions are getting views, comments, attention but don't end in a sale. So Cars & Bids is an auction site for cool cars, and I don't know exactly what their business model is, but I am positive that they take a cut of a sale if you if you s- successfully sell your Ferrari or auction off your Ferrari or whatever car you're selling on their site.
We're going to imagine that maybe that's dipping a little bit or it's a metric that we want to improve. Auctions that actually end in a sale rather than a seller ending an auction or a reserve not being met. So the impact, pretty obvious here but no sale means no cut, no revenue for Cars and Bids.
And beyond just the bottom line aspect of this, this scenario and this the impact of this issue and what makes it worth improving yeah, beyond the, financial aspect, buyers also start to lose confidence, right? I've never had an auction site I've never run an auction site.
I don't think I've even ever participated in an auction. But I would imagine if you have a site with tons of auctions for cool cars, and most of them are ending without selling or the reserve not being met or the seller pulling out, that's probably not a great great starting point for, driving people to your site to seriously consider buying something.
So our job is to figure it out and fix it. We're seeing that some percentage or at least we want to improve the percentage of auctions that end in a sale. So we did some research on it, and the question kind of zooming out a little bit is to research and understand what bit- builds and erodes bidder confidence, right?
So when someone gets to, the landing page, the auction page for this pretty sweet Mustang over here, aside from the car and the price, what are the things that get someone feeling confident to actually place a bid on a car that they've never seen in real life?
And what are the things that maybe make them a little skeptical and contribute to some of those auctions not ending in a sale? So our goal, to increase the percent of auctions that end in a sale.
Again, a simplified example here, but I wanted to give you something that's easy for us to all understand and that has a real impact, right? Something like this would certainly be on the radar of the Cars & Bids team if they were seeing a dip in auctions that are actually ending in a sale and more of them ending with a reserve not met or a seller pulling out.
So a maybe easiest way to sum all this up, if you're bidding on a car, what are the green flags and what are the red flags of a car listing page, an auction page on Cars & Bids? That's what we were what we needed to find out.
And I'm speeding ahead a little bit here, but we did that with a Lyssna study, right? We want to boost that percentage of auctions that end in a sale. In order to do that, we have to understand what factors, what information helps lead to an auction ending in a sale, helps lead to bidder confidence, and what are some of those red flags that might make someone hesitant, back off, and ultimately lead to an unsuccessful auction.
So here is the study that I created in Lyssna as the sole researcher for Cars & Bids. I'm just going to walk through the preview of it right here, just to give you guys the gist of the research before we start diving into the results. So first thing, we got a little screener question here.
In the past 12 months, have you placed a bid on a car through an online auction site? A pretty light screener here that could definitely be tightened up a little bit. But the idea here is that we're actually targeting people who have bid on a car, not just browsed like what I do.
I would be the wrong person to survey for this 'cause I do a lot of window shopping on sites like this. But I wouldn't be the best person to actually answer what makes, builds or decreases confidence to actually place a bid. So you've got a screener and a little welcome message here.
We're looking at how people decide whether to bid on a car in an online auction. Some open-ended questions here. Think about the last car you bid on online. What made you confident enough to place a bid?
And we'll say, "Lots of photos." Now think of a car you were interested in but didn't bid on. What held you back? Not enough info on the condition. We have a ranking question here. Rank these from most to least important whether, when deciding whether to bid on a car or not. So we've got some factors here.
These are things that, you know, if I actually worked at this company, would probably have on my radar, probably have some subject matter expertise to be able to point to some of the common most important information that you might, that buyers might be evaluating.
So we're asking participants to kinda move these around. Anything important to you that wasn't on the list? And then we get to the next part of our Lyssna study, where we're showing an actual listing page for this really sweet Ferrari that you can get to... That someone won for $68,000.
And we're asking some questions on that. So we're asking participants to look at the listing page, this is a real listing page from Cars and Bids, and following up with a couple questions as they pick through it. How confident would you feel placing a bid on this car?
My results aren't going to count here, so I'm just kinda doing a run through. Explain your rating. What made you feel more or less confident about bidding on this car? So getting some more, similar to some of our initial questions, but now getting into putting a real listing page in front of them and getting feedback.
And I believe this is our last question. What, if anything, would make you feel more confident about placing a bid? Nothing and that's the end of our study. So without looking at it diving too deep into like question by question what I wanted you guys see- guys to see here was first just get a feel for the study.
Sometimes when I'm working in LLMs and, connecting to things in Claude, and it's all text and, summaries of stuff, it makes me l- feel a little detached from like the actual... Especially with research, right? Like the actual questions that were being asked, the actual feeling, participant experience of going through this.
So I wanted to show it to you so you guys have a feel for what the actual study looked like to participants and also to just give you an idea of a study that would be probably a pretty good one to answer this question and ultimately make this improvement on our listing pages that would lead to more auctions ending in sales. So that's our study.
We launched it. We got some results back. And what's next is everything that comes after, and in this case, hopping into Claude and our LLM to use the MCP and some of our skills to take all that raw feedback and understand it, synthesize it, and communicate it so that research doesn't just die in a Lyssna study, but actually leads to some real improvements to the product.
So let's, let me pull up my chat over here. So I've got like a pre-made one of these just because I know it can be annoying to dep- definitely to wait for AI to do stuff when you have a live audience. But I'm going to start with a new chat here.
So I've got-- I'm connected to the List MCP. I've got the skills installed, and I'm just going to walk through, taking on my role of the Cars & Bids researcher with a completed study about our auction pages what would come next.
So first thing I'm going to do is just pull up the study that tested out, asked about car auction pages and what builds confidence with buyers. If you guys are following along, you're not going to have access to the same study that I do, but this is a good starting point here. You can ask... Just start by finding the right study in your account.
So if you already have studies in your account for, specific things related to your day-to-day work, ask about it here. Let's see. Hopefully, this one works. And I wanted to-- I did my caveman typing on this one just to show that hopefully, you can ask about this in plain English. You don't need the link.
You don't need the study ID number or the specific name. And you will get... It will find your study for you. So here it used Lyssna. That means we're connected to the MCP. Still haven't used any skills on this one yet. But we tapped into our research and in Lyssna.
Yes, there's one study built for this, Cars and Bids. What makes bidders confident is complete. So it gives us a little summary kind of confirmation that, yes, this is the one I was looking for. It's designed this way. Now, I only have one Cars and Bids study in my Lyssna account, so this is probably, pretty easy one to find.
But it's designed to give you a little feedback, a little context about the study, just so you make sure you know you're tapping into the right one. Because I know from my other non-example stuff with Lyssna I have a lot of studies that are named like home page one, home page final, home page preference test.
So this one probably pretty easy to find. The important thing is here, we asked about this Cars and Bids study, and we got some results back and a link to our results over here. So step one is, find the available study or the multiple studies and available research just by asking for it.
Now I'm going to skip past, some of the stuff in here 'cause we're going to get into that with our skills. We've got the study we want. We've got results. Now let's start, making sense of the results and what came back. We're going to start that by using this skill that I'll come back to Synthesize research, and synthesize research is basically a skill to turn those raw results into sourced and even cited themes and findings with the evidence behind every one.
So you can actually read through the entire skill file and see, all the rules and guidelines about how this skill works and how it processes that research. But at a high level, it reads every response, not just a sample. So we only have 15 responses for this car auction study.
But in larger ones, that is... That can be a difference from just asking Claude out of the box without a skill, where it might just sample a few or be opaque in what it's, what it's actually returning. As we know as researchers, we don't want to just pull a little sample size.
No matter how many responses there were synthesize is going to pull all of them. It's going to group answers into themes, each with a count and some real quotes and evidence. It's going to break down every rating and ranking question, and it's going to flag tensions and data quality issues.
So this is one of the most important things that I wanted to build in from a research perspective to this skill.
The experience that I've had w- and that I see people posting about, asking about all the time is, especially when it comes to research synthesis with AI, is that it's really easy for AI to be confidently wrong with how it's synthesizing your research and in some cases, overconfident about a certain theme or direction, and that there's not a whole lot of transparency into, like, how it's arriving at that quick summary.
So an important part of this Synthesize research skill that I wanted to build in was not necessarily to s- evaluate how reliable it is, but to flag stuff that you would want to know as a researcher.
So let me come back to this example here, and let's say if you want to call this skill synthesize research And it's going to do some work on that.
This might take a minute, so I have a demo ready to rock if if it takes too long. But it's not only g- going to just synthesize the information, but it's going to give you useful context, stuff that I would want to know as a researcher regarding not just what came back, but where it's from, the raw results, and things that I might want to potentially look out for in terms of the quality, the sample size, before making a big product decision based on it.
So let me go... I'll go to this one because it's exactly the same, but here's here's the synthesis for this. We get an overview. I'm not going to read through all this, but a short, unmoderated survey with 15 people who had bid on cars in an online auction, so getting some context there.
Getting the shape of the study and a link to it. Another aspect of this synthesize skill, and really all the skills that I worked on for this, is that I want to be able to link back to the actual study, right? When so many things get lost in translation in AI summaries and so much is happening under the hood I always wanted to make it available to actually just go into Lyssna and see the results, right?
See the raw text, raw transcripts, raw videos that people were typing in. You can see here we've got some quantitative results. How confident would you be placing a bid? We're seeing our average of 2.87 not great there. What matters most?
We see our average rankings here. So we're getting a really full, thorough and, unbiased readout of what happened in this research. We're seeing some themes here, one of them being a seller who answers well and quickly. This was the most common theme.
People gained confidence when the seller was active and specific, and pulled back when answers were vague, slow, or missing. On the Ferrari listing, seven of 15 people pointed to weak or missing seller answers in their comments as a reason for lower confidence.
So we're seeing some of the big themes that are emerging. And then I'll come down a little bit more to some of the lesser, less frequent or maybe less lower impact themes or maybe even just less themes that were not as tightly connected or maybe even just a little less clear in how people articulated them.
The idea again here, synthesis, it's making it easier and more scannable and giving you a starting point to start to break down this research without just glossing over the details. We see some tensions here, great photos, middling confidence, reserve didn't matter, seller reputation ranks number two. And finally down here some data quality notes.
So we're seeing a small sample size of 15, which I would actually argue is not... For something like this could be good for directional stuff. Many responses look implausibly fast. 11 of the 15 finished the whole study in 13 to 16 seconds.
I was really happy when I saw this get pulled up in Synthesize because I did all the study. I filled-- These are all just demo responses here, and I did them really fast. So this this is good that it picked up on that. If I didn't know that this was a demo and I didn't know that I did all of these by myself really quickly, I would certainly want to know that these were done in 13 seconds before, making my next big... Deciding on the next big feature improvement to our auction pages.
The listing only, right? Narrow sample no other demographic data, again, 'cause it was all coming from me here. Synthesize. Again won't read through the whole thing, but want you guys to see what it looks like with this example and really wanted to highlight some of the things that I focused on building into this skill that are truly meant for research integrity, right?
Not jumping to conclusions about things, and more importantly, just presenting you with important information about the research, about the reliability of it, about the sample size that you would want to know, or that I would want to know at least as a researcher to make decisions going forward. All right.
This we already covered here. Let's get to our next skill. By the way, if you guys have any questions on that one, drop them in the chat. I'm always slightly worried when I don't see anything popping up in the chat. I hope you guys are just following along or maybe synthesizing your own Lyssna studies.
Hopefully, that's the case. But if you got questions or want me to run it back on anything just let me know. Moving on with our workflow here to the last skill that we're going to cover share findings. So as it stands right now, we had a problem. We launched a study to get some feedback and answer some questions, and we've got a pretty legit synthesis of it.
Now, we could adjust this. We could push back on some stuff. We can certainly go through every response one by one. But for the sake of kind of moving on in the workflow here, we've got a good handle, and our LLM Claude has a good handle on what this study was about the purpose and the data, what came back with it Any researcher, designer, everyone here, I'm sure knows this and has experienced it.
This is not the end, unfortunately, of the research process. It's a big milestone to get back your results and see what was happening and see some common themes and pain points in the feedback, but there are still a lot of things that need to happen before that feedback turns into a, new feature, an improvement, or going all the way to our end goal, a 10% increase in auctions that end in a sale.
So with this next skill, Share findings my goal with this one was to facilitate everything that comes after the results are in. I originally built this skill as a build a stakeholder deck skill. And if that sounds appealing to you, don't worry, it can still do that.
My idea was take raw results, turn them into a deck that presents the big items, distills it down to the key points and, create something that you can present to your team across all, areas of your organization that is going to focus on the right things, get buy-in, give the proper context, and ultimately keep the momentum going on this research and the decisions that are going to flow down from it. But as I was working on this, I realized, yes, a stakeholder deck summarizing research results, yes, that would be really useful for me at least.
But the real skill and the real thing that happens after the results come in is not as narrow as just a deck. Really the bigger thing that all researchers have to do after the results are in are, is to communicate those findings for your audience and the intent, right?
So in some cases, that could be a stakeholder deck. You got 20 minutes. Joe, what happened? What came back in the research? What should we do? Give me the big stuff. A deck is going to be a good one for that.
But it could also be a quick Slack update to your team, right? It might not be like a super formal research report or something that you're presenting on a team-wide call. It could just be a, "Hey guys, we just just ran this study. Here's some of the stuff that came back."
Keeping your team in the loop and updated with the research and the feedback that's going on. It could be a FigJam board to get your design team going, right? "Hey, here's what happened in the study.
Let's jam on some possible solutions. Here were the big things that we learned that are, contributing to people not bidding on a car. The big things that are eroding trust in, buying a car sight unseen.
How can we fix these with the goal of, increasing that auctions that end in a sale metric?" So this skill is not just about creating stakeholder decks or any specific deliverable.
You can do all of those, but more broadly it's to package findings for whoever needs to hear them, whether that's a quick Slack update to your design team who already is already in the loop with this research, or a stakeholder ready readout that is going to, you know, operate at a higher level and focus on some of the less, less nitty-gritty design decisions and more about the big themes, momenta next steps in carrying that momentum into product decisions.
I'll show you what this looks like here in this chat. We have our synthesis, and I did a... My next prompt was, "Let's create a brief slide deck to present these results to stakeholders."
So this is calling the Share findings skill, saying Joe wants to take this study and the synthesis from it, and Joe wants to show it to someone in a certain format, communicate this.
And in this case, it's a slide deck to present to stakeholders. So here it's confirming hey, is this, are these the big points? Is this what you want to include in this? Is this the right goal?
And as you guys will see if you're playing around with this if it doesn't have this information, it's going to ask you, right? So when you're putting together a research report and talking about the impact or the next steps, an important aspect of that is knowing the goal of the research and being able to communicate the goal of the research.
So this one already had it. I must-- I maybe through doing this a couple times, gave it that goal or that impact of percent of auctions that end in a sale. But if not, it's going to say, "Hey, what is, what are you trying to present to your stakeholders?
What decisions are you trying to make? What needle are you trying to move?" And collecting some more information that's absolutely necessary to present your findings in a clear and compelling way. So I said, "Yep, that's the gist. We're looking to make up improvements to the auction experience page."
And again, getting a little confirmation here, some next steps, and then our deck put together. I did not modify this one at all, but you can see here it's taking that synthesis, confirming it, understanding the goals of your research, what you're trying to achieve with it, and turning it into not only a nice looking branded presentation, but a presentation that hits on the right stuff, that gives the right context to whoever you're presenting to.
And even if you end up making changes or tweaks to this, the idea is that you're not starting from zero, right? S- not starting from scratch. Making presentations like this from scratch takes me a really long time, and part of it is what to include, what not to include, how detailed should I go?
And then there's just like the grunt work of putting slides together. So this Share insight skill or Share finding skill allows you to craft it into how it... Whatever the output or deliverable is and to whatever audience you're presenting it to.
And I can share this with you guys. I'll put it in the in the resource guide. Last ones. I know we're over on time here, so let me just show a couple other examples. I said, "Looks great. Let's draft up a Slack message to..." Where is it? "A Slack message to our product team, giving them a refresher update on the study and the new results and highlights."
Get a little Slack message here. Again, just taking those synthesized results and not turning it into a slide deck for our stakeholders, but turning it into a quick and, more importantly, accurate update that we can share with our team. The last one summary of this prompt here is I said to create a FigJam for our design team to start jamming on possible solutions.
Give a little context about the study and give them... unleash our designers on this problem and this research, and set them up with a board where they can start thinking about possible solutions, possible improvements or maybe even further research to do. So use Figma here. Use our share ins- share finding skill, and let's see if I have this.
Created a little FigJam board For our team, let's see if this loads. Testing out some skills. All right, I'm glad... It seems like everyone's quiet in the chat because you're all just heads down and quiet, which is good.
That's what I want you to be doing. Mute me if you have to. But here's our FigJam, right? So the idea here is now taking those results from the study, which we can get right back to here, and not communicating it as a slide deck or a quick Slack update, but instead communicating it to a different audience, our design team, giving them the synthesis, the big pain points, the stuff that, we want to start jamming out on, coming up with ideas and solutions for.
So different in how it's communicated, different in the deliverable, right? A FigJam is going to be a lot more conducive to ideas, solutions, collaborations than a slide deck that you're presenting. They all have their value. They all have their spots, and as researchers, you have to, jump around from all of them.
This skill is all about taking that raw, re- reliable, synthesized research and being able to communicate it, craft it, share it in a way that is going to be conducive to action, right? To actually putting this into motion and continuing beyond the study.
We got some questions. So I'm going to I'm going to pause here. I know we're a little bit over, but I'm happy to answer your questions if if you guys have them. Let's check out the chat. Multi-skill and multi MCP workflows are so cool.
Really great to see our tools evolving so much and so quickly. Ways of working will always ebb and flow, and I'm excited to see what's next. I'm with you on that, and I know especially when it comes to research, Raymond, that there's I'm super aware of how nice it is to do things really fast with AI but also how dangerous might be, too strong for for research.
But yeah, like dangerous, risky, and flawed, it can-- the s- the downsides that can come along with it. So I'm with you. And I think through working on this a lot, I saw a lot of, both sides of that and a lot of the potential of it.
And I think what I saw most was like that it is possible to really maintain that, that integrity, really give researchers control over this but still use these tools that make things a lot easier, faster, and more likely to get put into action.
So I'm excited too. I'm with you. Christina, do you have any more information you can share about the Lyssna MCP not being read-only? I will get you the exact information. We-- there is write access, and I know we are sending stuff out about that, I don't want to say announcing.
You'll be getting an email about that soon where you actually are able to build, recruit studies right from your LLM, which all these examples were just read. I don't have an exact answer, Christina, because I know it's available in some accounts that are flagged for like beta releases.
The main thing I can say definitively is that you will hear about it in the next week, and it is going to be available for I believe all accounts with MCP access. So you will be able to write a study in Claude and launch it through Lyssna. Fab, write access beta.
There you go, Hannah. Does Lyssna support native AI analysis and summarizations for tests conducted in other languages like Portuguese, or is the AI synthesis features currently optimized only for English?
That is a good question. I believe some of our, I can get you a more definitive answer on that one, Lucia. But the, a lot of the native like I can go to these. Let's see like summarization stuff, we've had question, whole study summarization.
Those are those work. We have clients all around the world who use those all different languages, and we have a lot of functionality just for lang- like launching your study in separate languages. So I know that the stuff in, like the in-app AI summarization tools work for any language that you can launch a study in.
Though I don't know if that or how rigorous that, or if there's any difference when you're using it through an LLM. A good thing to flag though that I will pass on. Raymond, how do you see the mindset of researchers and adjacent roles service design changing when a lot of the compiling, synthesis, and planning can be handled by an agent?
What would you prioritize if you are looking to streamline research with these tools? Damn, good question, Raymond. Let me think. We'll make this... I'll make this my last answer for the day. Supposed to be in English.
Seems to quality in the translation. Cool. Thanks for that, Anna. All right, Raymond, your question. Let's see. What would you prioritize research with these tools? So what I would say is the, a lot of the compiling, synthesis, and planning can be handled by an agent.
These skills I'm showing you here, you could you could just as well plug in a link to your Lyssna study in just regular out of the box Claude or ChatGPT, and it can find patterns in there. It can find patterns in there.
Are they going to reflect the actual sentiment of your users? Maybe not. Could they appear overly confident or overly certain in one way or another? Definitely. I've experienced that before.
So I think the w- so when the compiling and synthesis and planning can be handled by an agent, I think that's that's true, but there's still a level of that gets you into risky territory about the quality of your research.
And then when you get it, it's becoming faster than ever, like faster and looser than ever. That gets you into m- further in the deep end of now, not only like conducting flimsy research or letting an agent conduct flimsy research, but then using that or taking it at face value to make product decisions and running that over and over again. So I think the... What would I prioritize if I were to streamline research within these tools?
I would prioritize always having a human being always h- being able to have a transparent view of the research, the synthesis, the reviews that are being done, the context needed to say, "Wait a second, that yeah, that looks good on paper, but that's not actually how we ask those questions," or, "That's missing some of those key themes." The other things I would focus on are, like, grunt work aspects of research.
I showed you that stakeholder presentation, right? And that was part of that is research expertise to put that together and to know what to focus on. Another part of that is dragging stuff around in Figma slides and, adding your logo to each slide and, pulling the right, right screenshot there.
So what would I prioritize? I'd prioritize streamlining the aspect, like the more grunt work, time-consuming aspects of research that don't necessarily affect the, or, hit on the quality of the research, and I would focus on spending my time as a human, asking the right questions, looking into data to see if there's something that, AI would skim over.
Asking good follow-up questions in an interview, really using, like, where a lot of the process is going to be not as human as it once was, using that as an advantage to make the hu- to really put a lot of focus into the areas that do require a human, like a good follow-up question.
Yeah, Finnish is not considered an easy language. Ask participants to answer in any particular language going forward. Hannah, I'm going to leave that as the the cliffhanger question here. Hey guys, thank you so much for joining, for hanging with me today.
Thanks for staying, staying after and for all your good questions. I'm sure more will pop up. So I'm going to send out a follow-up email to you guys with everything from the workshop. I'm going to update this resource guide.
I'll put the recording in there. If you guys have any questions, let me know. Set up stuff. I know BF, to follow up with you, so I will get on all of that stuff. But a lot of this stuff is new for me too. I'm excited that you guys are excited about it.
Best way to get into it is to get into it and, start looking at some of your studies, running them, experimenting with it. But I'm available if you guys have any questions. And like I said, I'll send out- I'll update this guide and send it out after.
So you can check it out and start getting busy, continue getting busy with the MCP.
We've got some upcoming workshops over here as well. So check this out, keep an eye out for an email from me, and let me know how I can help on your MCP journey. Thanks, guys. Have a good one. See ya.


