Introduction

Welcome everyone.

Thanks so much Marcus for the introduction.

Speaker and Session Overview

My name is Rachel Beck and I'm excited to present Orca AI to all of you tonight.

It's a generative AI tool and so I'm going to talk about a little bit about myself and my background.

We'll talk about what problem I'm trying to solve and then we'll do a demo of the tool and lessons learned.

Career Journey and Foundations

All right so my journey I think like most people was not a linear path to product and technology all right so my journey like many of you was not a linear one I spent a little more than 10 years

working internationally in social impact doing research analytics and business intelligence helping people use data to make decisions predominantly with global non -profit so in the social impact space and I was an early adopter of technology

at least and what was up and coming at the time it was ICT information communication technology right so we still use all of these things are called apps on your phones right one of the first uses that we used in South Sudan

was a survey design tool to be able to instead of getting rid of all that paper right being able to use mobile phones to collect data in remote areas in East Africa right and to be able to do it offline right with no no Wi -Fi no network right and then you can come back plug it in and download all of your data

Early Tech Adoption in Social Impact

so this is sort of one of the first use cases in a lot of the global world where

we first started seeing technology and so I use this that perfect example with it survey design for needs assessments and trying to understand what are people's needs in these different countries where I'm working, what are the problems that they're facing, how prevalent are those problems, is it, you know, and who's facing them, which demographics, right?

From Needs Assessment to User Segmentation

So this is where you see sort of understanding needs and the problems people are facing, who are the people that are facing them, who's the beneficiary, or when you see on the other side of the screen, who's your user your user segment right and then understanding the operating

Understanding Context and Operating Environments

environment right context matters so when we talk about living and working in another country it's how does their government work what are their policies what's the regulatory framework there right in many sub -saharan African countries you have customary law which is the tribal law and then you also have

the government law how do you navigate that right what are the social norms if If your target user group is women, can you just access women? Or do you need to go through the men?

So context really matters. And the way that you think about how everything fits together in that full system is really critical for being able to do work around the world.

Bridging to Product Management

Funny enough, those same three things are pretty important in product.

And it wasn't really until I'm automating dashboards and working on global data management systems in you know more than 30 countries that I'm seeing this connection and I'm like wait a second I am a product manager I am working in tech and so that was really when I

Embracing AI and Technical Growth

started to embrace this side of my work I did an AI certificate course and was really excited to sort of find applications where I felt like I could lean into that research social impact background although I'm not really going

going to talk so much about the social impact side, and to be able to bring that to life here today.

Framing the Product Discovery Problem

So just for those who aren't in the product world, when we talk about the product discovery phase and thinking about what products we want to develop, it's that right side, right?

We're thinking about what's the problem we want to solve, who are users or users or segments, and then who else is doing this?

What's the market like, right? Who are our competitors?

So we're going to focus

Why Competitive Analysis Is Hard

on that for today so why competitive analysis why market research I've told you context matters right but also in terms of understanding what it takes to do this well it is wildly time -consuming right it is a lot of information to find

right how do you even find who all of your competitors are let alone what they're doing their features their approaches right who their users are do their users even like what they do right what do they love what don't they this This all goes into that type of research, right?

And because there's so much information, it's also really expensive, right?

It's so time consuming to do proper in -depth research that it's quite cost prohibitive at times. And oftentimes companies will take shortcuts.

So we don't want that. We want true depth for insights,

Qualitative Research and Synthesis Challenges

but it also takes technical expertise to code, to know how to think analytically about information.

And this is qualitative information, right? We're not talking about spreadsheets of numbers, we're talking about documents and putting words together and synthesizing information, finding coding and finding commonalities across code structures.

Enter Orca AI

As you can imagine, trying to do this well is important.

I built Orca AI.

Live Demo: Orca AI Competitive Analysis

I'm going to now shift. We're going to see how this goes to do the demo.

Mo.

What Orca AI Is

This is Orca AI, the AI -powered market research platform.

It's going to be a comprehensive research platform.

Setting Up the Workflow

Today we're going to focus on competitive analysis, and that's what I'm going to show you is that workflow today.

So when we would become a beta tester, which I will encourage all of you to do, once you log in and become a user, then it's going going to take you.

I'm going to skip that. There's no need for that admin today.

So then it's going to take you to this page.

Let me just refresh briefly. So the workflow works.

I tried to skip it. It didn't like it.

Here we go.

So when we log in, you're going to see a competitive market analysis. So today we're going to do an analysis together.

So we're going to create a new workflow flow.

And you can see here, there's two options. Uh, today, the one that's ready for us today is the complete, the full analysis.

Um, and I'm working on developing the human in the loop version of that. So that's coming. Um, but let's see what AI can give us all on its own.

Defining the Test Idea and User Segment

So I'm going to ask actually for you guys to help me with this. So the idea here is that we would put in a product idea and user group or user segment, that would use that product, right?

And then we're going to give it to AI. We're gonna submit that and we're gonna let it see what it comes up with for a competitive analysis, okay?

So, who's got a product idea? Anybody, yeah?

Jewelry. Okay, can you tell me more? What kind of jewelry?

Can I actually argue this is that? Let's do it, let's see what happens.

Kintsugi jewelry, so in the style of. Okay, wait, speak slowly, I have to type it all out.

A jewelry company that specializes in katsu ... I should know the name. In the ...

Repaired ceramics. In the ... Repaired ceramic style.

Let's use the ... Can we use the actual term? Sure. Kintsugi.

Kintsugi, okay. Okay, in the style of kintsugi, repaired, yes, repaired ceramics, does that work?

In the style of kintsugi, repaired ceramics, the style originates, let's just give it a little more context, right? The style originates from Japan to use this style for creating.

I'm just going to be redundant, and we'll see what happens for a jewelry brand. I want to be abundantly clear here. Next steps, user segments.

so who do we think is most interested in this women is there a particular age group ooh women okay women who have experienced trauma or who might resonate with can't spell resonate still can't spell let's have the internet help us

one s okay resonate with over I was gonna say maybe like resonates with healing resonates with a healing journey as this style is a rapper as okay let's

leave it at that this should be really interesting I've only tested tech or like software type ideas.

I haven't done a physical product like this so I'm actually really excited to see what it comes up with.

Behind the Scenes: Process Flow in Hunch.Tools

So while it's creating so I've done this in a second it'll say it's and it's going to think it does take a few minutes and while it's thinking what I want to show you is here

on the back end I don't know if you can see here it says the Kintsugi jewelry company competitive analysis and it's actually going to go through this workflow. I wonder if you're going to be able to see it as it goes through it.

Well, just, I'm not sure. Let's see if we can, here we go. Oh, maybe not. Okay.

I don't know if we're going to see it run. Um, yeah, we are. Okay. Do you see it moving? It's actually running.

So while it's running, so what it's doing is this is the backend analysis. This is in hunch dot tools.

Hunch .Tools is a process flow AI app, and basically for each of the different segments of your user flow, you can select the model that you want to work in, you write your prompt, and then you can give it parameters and everything, and then it will actually do what you've asked it to do. You're able to link each step together, so you're able to build on your work.

So whatever output comes from this prompt can then feed into the next creating that process flow So this was great when I wanted to be able to layer different kinds of analytics, right? I wanted it to be able to pull who are the direct and indirect competitors right to then say, okay

Give me you know these ten variables I want for each of those competitors Build that out then I want to say okay. I want to understand though that sentiment analysis of their users, right? right? Then I want to do a SWOT analysis and there's a few more things in there.

Um, but each of those things are then connected and as this runs, the outputs will then feed in. It has multiple layers of analysis here, which then feed into that, that brief.

Waiting for Results and Design Rationale

So then if we go back let's see how we're doing on this it does take a few minutes so okay let's see if it's ready last saved never that's new I haven't seen that before

right I was pretty excited about my little my little orca swimming in the ocean there any idea yeah sure sure so when I thought about developing a tool

for a competitive analysis and market research from a textbook perspective like if you were to read a book on on market research it would talk about the market as an ocean a blue ocean means there's no competition right it's a it's

an open market and then a red ocean is where it's a completely saturated market it's very hard to differentiate yourself and then a purple ocean is where there's likely competition but you can find your value proposition and differentiate

yourself so the entire style of the website or actually I should say of the presentation and and the website to some degree has that purple color because a killer whale is a competitive right animal in a competitive landscape right

so it's it symbolizes that and then when we go back to the actual PowerPoint you'll see the spectrum of colors is that blue to red with the purple in the middle so everything was kind of intentional around that that sort of

idea around a marketplace great question thanks for asking that that was fun to talk about.

Findings from the Competitive Brief

All right, we have a brief. So I am not going to read this verbatim to you, but I am going to scroll through it and pick out some highlights. And let's see what we think.

There is an introduction, as you can see. It's going to talk about the most threatening competitors first.

Please let me scroll. Okay.

And the idea here is these are the competitors that you need to pay the most attention to. These are the ones that may be most similar to your idea and are the most threatening to your business idea.

Top Competitors and Threats

So we have SoCo. They don't, it says they don't focus specifically on kintsugi, but they have a significant market presence.

They're a series B, which is interesting. And they are scaling ethical jewelry with meaning is their value proposition. So we can see how that would connect, right?

right, to our idea and what we're testing. I'm just going to keep going. I don't want to take too much time to read everything.

There is an actual Kintsugi atelier, and they are actually embracing that philosophy. They pose a significant threat because they've established their positioning around authentic storytelling, premium craftsmanship.

It talks a little bit about how they've captured mindshare in the kintsugi jewelry niche they have a 12 ,000 Instagram followers so they have a starting to build a following around

that it says however their operational weaknesses and limited selection they only have 50 items available reveals vulnerabilities that a more execution focused competitor could exploit so you see it gives you both of those pros and and cons right and gives you ideas for how you might be able to position yourself to compete

Direct and Indirect Competitors Overview

in a market even with similar brands so then it talks about who are the direct competitors you can see here oops i scrolled a little bit fast but it talks about it has their strengths weaknesses and competitive benchmarks this will be presented for all of them as well as for for indirect competitors.

So I'm not gonna read all of them. We talked about Kintsugi already. So here's Broken and Beautiful.

They're upcycled ceramics through sustainable jewelry practices. So they may be actually using ceramic pieces for eco -consciousness.

Their weakness is higher price points, a limited market accessibility, and a small customer base. They also have 15 ,000 visitors of monthly traffic,

5 ,000 Facebook and a customer rating of four out of five there's healing jewelry which has a healing value proposition so you can see the strengths weaknesses benchmarks and then it goes through the same for indirect and then it gives you

Market Summary and Key Pain Points

a summary of the current marketplace overall so this is what the competitive landscape looks like they're strong customer demand for jewelry that combines physical beauty with emotional resonance.

In the specific Kintsugi -inspired niche, Kintsugi Atelier has established early leadership but shows significant operational weaknesses.

Across all competitors, the most consistent pain points relate to operational basics, shipping delays inventory availability website usability and customer service responsiveness that's key information right if you wanted to start a business

in this area so let me scroll down and then it talks about what your opportunities are and recommendations and this is your value proposition this is the way to structure your business to be able to differentiate yourself and then have a viable a viable product or a viable business so here it recommends recommends operational excellence as a differentiator.

Recommended Differentiators and Models

This makes sense based on what we just heard, right, as those weaknesses.

Have an exceptional digital experience that educates customers about Kintsugi principles, making product discovery intuitive, sizing information transparent, right? So this is really embracing that Kintsugi philosophy and making that front and center.

Modular collection strategy. Develop a core collection that allows for personalization without creating fulfillment challenges.

This balanced approach solves for both limited selection issues as well as inventory management challenges of larger competitors.

And then it recommends viable business models here. So we can see direct -to -consumer approach, which makes sense, strategic wholesale partnerships,

hybrid model addresses limitations of Kintsugi Atelier's direct -only approach while avoiding inventory challenges and maintaining tighter control over your production.

Revenue diversification should include core product sales, limited wholesale partnerships with select boutiques, digital content, and education.

So, and then, I mean, obviously it goes on a little bit, but that last piece is those recommendations of how you can differentiate yourself

and really capitalize on where your competitors' limitations, right, and then how to borrow what's really working and what customers love.

Takeaways from the Analysis

So that is an example of what you can get out of Orca AI.

I love that it works for all kinds of products. It is industry agnostic, as we can see. So yeah, that's pretty great.

So let me now, uh -oh, we're going to play this game? Okay, here we go. We'll get there eventually, guys.

Thanks for your patience. Oh, thanks. All right.

Lessons Learned Building Orca AI

So what did I learn?

As we can see today, technology is not perfect. Embrace it, but be prepared for challenges, right?

So that's not on this list, but I think worth mentioning, right, every time.

Prompt Engineering Matters

So the first thing that I want to talk about is prompt engineering is a thing.

So, the front end of Orca AI that you're seeing, the user interface, was all built with Replit. I was going to say I wrote zero lines of code, technically I copied like four lines today. But that was the extent of the code that I used to build this whole thing.

Otherwise Replit has a full agentic system where you type in, in natural language, what you're looking for, what you want to build, how you want to do it. and those prompts matter right in the same way when we were writing that product description

when we did the demo right the more information the more specificity right that you can give the better your output the more consistent that will be so prompt engineering is a thing all llm based tools want to please you replit included how many of you have used chat gpt or anthropics clod

and it just tells you what it thinks you want to hear and you know the information is not right everybody everybody right so replete is the same okay there are many times when I would ask a question to be like no it's this and I'd be like can you check this and it would check and be like oh you're

right and then based on the information it would write I was like mmm something else is missing you'd have to ask again right the point is anytime you're using using LLMs, please pay attention, right? Critically think about what that output is.

Is it spot on? How are you measuring that quality of a response to make sure that you're actually getting what you need and that you don't end up building bug on top of bug on top of bug and just trusting it?

API Integrations: Common Pitfalls

APIs are tricky, my friends. So APIs are how you connect, right? right? One tool to another, how information passes between them, right?

So I have API connections in two places, one between Replit and Hunch .Tools to be able to start that workflow in Hunch .Tools and then have that brief come back to Replit in the user interface.

The second is with Stripe for a payments integration.

So oftentimes when you use a tool like Replit, it'll add a new feature. you're like yes everything's working properly and then you test it and you're like haha my API broke right now information is no longer getting to where I need it to go and you've got to figure out how and why and it's a process so you've just got to kind of keep going and know it's gonna break but

that you can fix it and so the last one most importantly I could say this ten

Test Relentlessly and Iterate

times please test test and test again as you can see tonight right you got to to keep testing. You got to keep going. You got to keep working on it.

And so what I actually really enjoy about this is it's an iterative learning process. Every time you come across something that doesn't quite work, you're like, now I know I got to do that.

One thing I think I learned today, extra, we were trying to, I wanted to be able to give a coupon coupon code to some of you to be able to try this for free. I want everybody to join beta testing.

And so I was trying to figure out how to do that and the way that Stripe allows you to create a coupon to create a code for you to put it in is not actually compatible with the way that Replit built the payment structure. Didn't find that out until today.

we're trying to I'm testing like everything's built I'm just testing it and I was like oh no so eventually we were like wait I have the database I own the database and so I went in and was like I can just give whoever credit so we figured out a workaround right but so now I know

any time that I'm trying to connect two pieces of of software now it's like you know like wait I don't just need to know what I want to build I need to know is that going to be compatible compatible with my destination.

It's another thing to query, question, and plan for now when you're going through the process.

Why Replit Worked for This Build

One of the things that I actually love about Repl .it as a tool is it has when you... I'll actually show you. It's more fun to show you.

So here is where you essentially you're talking to various agents for specific purposes, but what I like here is at the bottom it has plan build plan and edit the plan feature is gold it is amazing you should be planning before you're building and

what I love about that is allows you to ask all the questions and get lots of information to be able to create a plan with your agents and then to be able to to deploy and implement and build. And it'll help you transition over to the other one.

So I really, really enjoyed that feature. And I think I haven't seen a lot of other tools that do that well, that allow you to interrogate.

And even as you're going, anytime I had a question was like, wait, what about this? I got to go back into plan and ask, is it gonna work? Does this, how does this work?

Tell me about the backend. What's the code look like? All that good stuff.

So Replit, I really am really happy with using Replit as the base and the user interface for this. So yeah, I think that's lessons learned.

Call for Beta Testers and Next Steps

And then you can actually go to the site and sign up for beta testing and become a user. I'm looking for early adopters.

You will see there are two websites. That's because the top one's not yet quite ready. So today I did the link for the domain name,

which is orcaai .site. that's going to be the domain where it's going to live but that transfer takes up to 48 hours I did that this afternoon not knowing it takes 48 hours so it's not quite ready but you can do the same thing and access the site at the second

website here at orca -ai -rachbeck .replet .app so if you go

How to Access and Early-Adopter Offer

there that's live and then today for the first five people from this group who sign up I'm gonna give you guys like 12 hours before I put it on LinkedIn for others to join in but the first five people that sign up I will I will give you a credit for your first use of the tool so it'll be free alright that's it

Closing

thank you so much