## Episode Summary
Bo Bennett built DesignDraft.ai originally to solve a real frustration his daughter—an interior designer married to a general contractor—was having with generic ChatGPT image outputs. In this episode, he walks through how the tool routes design requests to different AI models (OpenAI and Google Gemini) based on the precision required, and why homeowners are driving the majority of usage. The conversation moves from a family project origin story to a live business with Pro and Agency tiers used in actual sales pitches.
---
## What You'll Learn
- **Why model routing beats a single AI**: DesignDraft sends precise edits (like "move the light switch 5 inches right") to Google's Gemini Flash and broader room redesigns to OpenAI—because knowing each model's strengths matters more than picking one winner
- **How tight UI constraints fix bad prompting**: Rather than letting users write free-form prompts and fail, Bo built the interface so users "can't really screw up"—the system guides them to the output they actually want
- **The sales pitch use case nobody expected**: Designers and contractors are using shareable before-and-after links to win clients before the first meeting—showing what looks like a photograph of the client's vision instead of a sketch or an "Atari-looking rendering"
- **Homeowners, not pros, are the volume market**: Despite targeting three user types, homeowners dominate usage—often creating images specifically to hand to their contractors so both sides are aligned on the vision
- **Text and image AI are more alike than different**: Bo's biggest takeaway from crossing into visual AI was that strong prompting discipline carries over—the medium changes, but the communication model with the LLM doesn't
---
## Notable Quotes
> "If you're presenting to a client what appears to be a photograph of exactly what's in their mind, they say, 'Yes, I'm going to go with this—this is exactly what I want.'"
> — Bo Bennett
> "Whosoever is not doing this has a major disadvantage for people who are doing this, like my daughter."
> — Bo Bennett
---
## About the Guest
Bo Bennett is the owner of Archieboy Holdings and a builder of niche AI-powered tools, primarily known for products aimed at authors and publishers. He built DesignDraft.ai after watching his daughter—who runs an interior and exterior design business alongside her husband, a general contractor—struggle to get usable results from general-purpose AI image tools. Rather than build a private workaround, he immediately opened it to the public, recognizing the broader market potential. He can be found at archieboy.com.
---
## Topics Covered
- AI Model Routing
- Precision vs. Standard Mode
- Interior Design Automation
- Contractor Sales Workflow
- Homeowner Use Cases
- Prompt Engineering for Images
- Before and After Sharing
- Building for a Niche Audience
Full transcript
HOST: It's good to have you back, Bo. Last time we talked about StarringMyKid, you mentioned your daughter is getting married soon, and you were building her childhood stories into books. Now you've also built her a business tool, DesignDraft.ai. That's a bit of a departure from your usual author-focused stuff. Walk me back to that moment—what was she running into that made you think, "I need to build this for her?"
GUEST: Well, she runs a design company, interior design, and she actually does some exterior design as well. Her husband's a builder, so they work together on these projects. And she has been using ChatGPT, trying to work with ChatGPT to get exactly what she want. And what she wants, and it's actually kind of funny, the uh the outputs and and what it does. I mean, it's really good, but it's just not the right tool to get the exact picture and the design that she wants. So, I decided to actually build a tool for her using not just the latest version of OpenAI image, but also using uh Google's processor, which does an amazing job, and using a combination of like the best ones. And more importantly, creating a system that has much tighter prompts in place where users can't really screw up, where it's very easy to get exactly what you want by using the user interface. Uh and that's what I did for her, and she's been using it ever since, very successfully.
HOST: Well, I was curious about that tighter prompt system because on the site, you have two modes now, standard, which is like a full room replacement, and precision, which allows for really surgical edits. How does the mechanism in precision work differently? And when does a designer really need that versus a full redo?
GUEST: Both of them use different LLMs. So I believe the I could be mixing these up, but I believe the precision uses the Google Google's Gemini. Image like flash image because it also known as Nano Banana 2. At least that's the best one right now. Cuz because that allows for very precise instructions. Like for example, you could say, move the light switch on the wall 5 in over to the right. And it'll it'll do it. And um usually ChatGPT has trouble with something like that. But a combination knowing which models where their strengths are and where the weaknesses are, we can make sure that it the command gets routed to the right model to give the user the best image every time. And sometimes users may be playing around with it and they just say, you know what? Let's just do a redo. Let's start from the beginning and that's their choice as well, but they have that option.
HOST: Got it. I, I have to say, looking at the site, you explicitly target three different user types—the professional interior designers, the general contractors, and homeowners. That's a wider range than your usual publishing tools were for when you're laser focused on authors. Did that create any design tension when you were building the site, and which user would you say is driving the most usage right now?
GUEST: I would say the homeowners are driving the majority because it's a much obviously bigger market.
Then interior designers or general general contractors and that's to be expected.
Um but the homeowners will use it for all different reasons obviously if you're redesigning a room, if you want to redesign a room, you want to give your contractor an image of what you're talking about.
These work a lot better than sketches.
You could show them exactly what you're talking about.
Uh so it it tends to work well.
And no there were no real tensions when when building the site.
It was just you build it with the market in mind and you just you just do it.
And that's what we did.
HOST: On that note about homeowners giving a contractor an image, the Pro and Agency plans offer shareable before and after links. How do you see designers and contractors actually using that link in their sales process? Is it something they send as a pre meeting asset, something they iterate on live, or something else?
GUEST: Both, actually. My daughter is a good example of this, and her husband, the general contractor, they will take these images and use them in their their marketing in order to get a client. And it works extremely well, because if you're presenting to a client, like what appears to be a photograph of exactly what's in their mind, it's it's amazing. And they see it, and they say, "Yes, I'm I'm going to go with this because this is exactly what I want." Whereas the old way of doing it might have been a sketch, a rendering, um like a really cheap looking computer rendering that that looks like it's from an Atari game back in the 1980s. That doesn't work anymore. So, whosoever not doing this has a major disadvantage uh for people who are doing this, like my daughter.
HOST: Most of your tools so far have dealt with text—books, emails, press pitches. DesignDraft is fundamentally a visual AI product. What surprised you most about building in the image space versus the text generation space? Where did your existing instincts serve you well, and where did you have to learn something genuinely new?
GUEST: Working with images is different in many different ways, but it's ultimately the same in the way you have to communicate with the LLM. In order to get things done like prompting and good prompting. So there were while there are differences, there I would say that there more similarities than differences.
HOST: DesignDraft started as something you built for your daughter's business, and now it's its own platform in the Archieboy portfolio. At what point did you know this had gone from a family project to a real business, and is there a version of this story where she eventually runs it herself?
GUEST: और इमीडियेटली आई न्यू दैट दिस वॉज गोइंग टू बी फॉर एवरीबॉडी। एनीबॉडी हु वॉन्टेड टू यूज़ इट। आई वुडंट हैव बिल्ट इट अदरवाइज़। इवन दो आई आई लव माय डॉटर एंड आई डू एनीथिंग फॉर हर। इफ़ आई वॉज़ गोइंग टू बिल्ट अ वेबसाइट आई माइट एज़ वेल ओपन इट अप टू अदर पीपल एज़ वेल। शी डज़ गेट अ फ्री अकाउंट आफ्टर ऑल सो आई थिंक शी इज़ हैप्पी। एंड आई एम हैप्पी एज़ वेल बिकॉज़ आई सी दैट माय वर्क एज़ अ मच वाइडर ऑडियंस एंड गेट्स यूज्ड मोर। एंड दैट्स काइंड ऑफ व्हाट आई एम गोइंग फॉर। सो आई न्यू फ्रॉम द वैरी बिगिनिंग दैट दिस वॉज़ अ रियल बिज़नेस। आई डोंट सी हर रनिंग दिस। शी इज़ नॉट इंटरेस्टेड इन रनिंग द साइट। शी इज़ जस्ट इंटरेस्टेड इन यूजिंग इट एंड दैट्स फाइन।
HOST: Before we go, for listeners who want to follow up on what we covered, where can they find you and the work you're doing?
GUEST: They can find me at archieboy.com, which is my website, and they could also go to designdraft.ai. to find the site that we're talking about.
HOST: Excellent. Thank you so much, Bo, for talking with us today. I love that Precision Mode distinction with the light switch—it really makes it clear what the tool is capable of. And thank you for spending part of your day with us. Let's get you back on soon—there's plenty more we can cover. Until next time—that's a wrap.