Build Your Own Software With Zero Code

Stay a step ahead.
The Foxwell Digital Download brings you what's working in paid social and performance marketing right now.


Question: How do you build your own software with Claude when you have never written code?

Direct answer: Act as the product manager and let Claude Code be the developer. Colby Flood, who has never written a line of code, built Brighter Click's whole UGC operations platform by starting with the single most urgent problem, describing the outcome the user needs with zero room for interpretation, and asking Claude for an easy, non-developer, step-by-step process whenever setup gets technical. Once the team depends on the tool, add a test environment, push every change to test before production, and have Claude run a full end-to-end test.

In this episode, Will Sartorius talks with Colby Flood, founder of Brighter Click, about how to build your own software with Claude when you have never written a line of code. Colby walks through the UGC operations platform he built for his agency, from a 15-step creator application and a database of more than 600 creators to AI quality checks on every video, and explains how he decides what to build, how he tests it, and when he pushes back on Claude.

Colby started Brighter Click in 2019. The agency began in media buying and now leads with UGC and influencer sourcing, creative strategy and post-production.

We discuss:

- Why Google Sheets, Canva decks and Drive comments broke the UGC process
- How a non-developer gets started in Claude Code by acting as the product manager
- Why the build started with the most urgent problem, not the first step
- The 15-step creator form behind a database of 600+ creators
- Using ClickUp time tracking and Claude to find the admin work worth automating
- Automating shortlists, client approvals, contracts, briefs and upload portals
- AI checks for aspect ratio and subtitles, plus a 20-point QC checklist
- Turning client feedback into a running checklist every creator sees
- How to set up a test environment and push to test before production
- How to challenge Claude when it suggests a different build

Show site: https://www.ai-dtc-wtf.com/
Connect the Foxwell MCP to Claude or ChatGPT: https://www.foxwelldigital.com/mcp

Guest: Colby Flood, Founder, Brighter Click
Brighter Click: https://www.brighterclick.com/
Colby on LinkedIn: https://www.linkedin.com/in/colby-flood/

This Episode is sponsored by Motion's very own Runneth. Runneth is the AI Tool that is quickly becoming the go to decision maker brands turn to when it comes to creative decisions within their ad accounts. From Creative strategist to hook inspirations and creative briefs, runneth can do it all. See the case studies and try runneth yourself https://runneth.com/

This episode is also sponsored by Alia, the AI-powered popup platform that turns more of your site visitors into email and SMS subscribers. Our listeners get an extended 45-day free trial: [ALIA LINK]

AI DTC WTF is the show for brand owners, in-house advertisers and agency operators who use AI every day and want tactical workflows they can implement now. Published by Foxwell Digital: https://www.foxwelldigital.com/

To connect with Andrew Foxwell reach him here Andrew@FoxwellDigital.com
To connect with Will Sartorius DM him here https://x.com/will_sartorius
To connect with Thomas Moen DM him here https://x.com/thomasmoen
To learn more about Foxwell Founders and conversations like this one, go here: https://www.foxwellfounders.com
Learn more about Foxwell Digital and join our newsletter: https://www.foxwelldigital.com/email

Chapters

00:00 How to build your own software with zero code
02:04 Why Google Sheets broke the UGC process
05:32 How to start in Claude Code with no technical background
08:17 Which problem should you build for first?
09:40 The 15-step form behind a 600-creator database
14:42 How to find the admin work worth automating
16:19 Automating shortlists, contracts, briefs and uploads
26:57 Demo: AI quality checks on creator videos
35:40 How to set up a test environment with Claude
42:37 How to push back when Claude suggests a different build



Transcript

Build Your Own Software With Zero Code - YouTube

https://www.youtube.com/watch?v=ShTXfcpyB3M

Transcript:

(00:03) If you work in DTOC and you use AI and you're  wondering what the f is going on every week, this is your podcast. The AI TTOC WTF podcast,  your home for tactical tips, strategies, and ideas that you can implement right now in your  AI workflows to make your brand or agency more money. All right, welcome to another episode of AI  DTOC WT book. Uh, today we have a special guest.

(00:28) I think every guest is special. You know, today  is particularly special because a good friend of   mine, uh, Colby Flood, who runs Brighter Click.  Uh, thanks for coming on. Yeah, excited to be here. So, Colby, I think as a starting point,  why don't you just give us some background on yourself, uh, you know, your agency, sort of what  you do, uh, and then we can sort of get in the   weeds a little bit on what you've been building.  Yeah. So, I started Brighter Click in 2019.

(00:54) I had been freelancing for some time before that and  then just made that transition where I needed   to have more team because client base started  building up. Um, you know, fast forward from 2019 to 2026. Here we are now. We were primarily  media buying as a service because creative didn't really matter as much back then. Over the last  three to four years, we've delved really deep into   creative strategy and post-production, but  more specifically UGC and influencer sourcing.

(01:21) Right now, UGC is the primary source of our  pipeline for new client opportunities. And so, there's been a lot of chance, a lot of need  for us to build out tech, to build out tools   that can help get our team out of Google Sheets,  get our team, you know, centralized, but also make sure that the creator and the client experience  is good without having to pay, you know, multiple thousands of dollars per month for some  tech that we don't even own. So that's been my   goal over the past uh year, couple of months  is just building out better tech for that.

(01:54) That makes sense. So two questions piggybacking  off that you know the first what did your process sort of look like beforehand? Um and then second  like do you have a technical background at all?   Like what introduced you to Claude and like how  did you get started? Two good questions. Yeah. So, you know, our process beforehand was very  much fragmented in the sense that um we would reach out to a creator.

(02:23) You we'd post on  Twitter or something like that, post on Reddit, find some creators. Um we would pull all of  their videos that we could from their portfolio and then put them manually into a Canva deck. We  would send that over to the client. The client would review those, approve them, etc. And then  we would go into contracting. We would go into   product shipping, all that. All of this was  being managed in Google Sheets.

(02:45) Every client had a separate Google sheet that we had to  keep up with. We had to keep up with email   communications through um you know the creators  and the client. And then at the same time when we did QC, which is the longest  potential time drag of the entire UGC process, um we were doing that in Google Drive.  So a creator would upload their content.   We would then leave comments on that file in  Google Drive.

(03:12) The client would do the same and then we would send it back as either that or  a document to the creator. Everything was just so   scattered that there was no way. And it would  kind of make more sense as I explain the tool we built. There's no way to keep up with like what  creator is good at turnaround time and consistency of quality and what clients expect and all of that  type of stuff. And then technical background.

(03:33) Um, no, never written a line of code. I did um you  know I started to build out a software three years ago and hired a team of developers for that.  So I had a CTO from my hometown he's in his early 60s I believe. Um he worked at Wells Fargo for  like 20 plus years lead that project. And so I had a lot of opportunity over the past three  years to lead software development and learn a lot in terms of like how important it is to be  very very crystal clear on your communication and leave zero room for interpretation so you don't  build in the wrong direction for two weeks. Um   and just how that entire development process  goes. But no, I don't know Python. I don't know

(04:15) umnet or you know anything of the sort that way.  This episode of AID TOC WTF is brought to you by Runith from Motion. And you know, there's so  many interesting things about Runith that I   absolutely love. The best marketers I know right  now are not sitting there going, "How do I get AI to answer my questions?" They're thinking in  systems. It's I have taste. I have judgment. Um,   I know my customer better than anyone.

(04:42) And they're  putting that into Runith and having it work on it in a really sophisticated way. It's putting it in  there and Runith is amplifying what they can do   with that particular data. It's not a chatbot,  right? It's not just clawed. It's not just chat GBT. That's the thing I love about Runith is that  it pushes it further and it has your ad data, your competitor data, your creative strategy.

(05:08) It  has your product data and it helps you understand   all of these things together so that you can  genuinely train it on how you think and it becomes useful from an operational level. Saving you  time, giving you context and giving you important information so that you can make the most absolute  best decisions as a marketer today to help make your business run more efficiently. Motion can  do so much. Runith can do so much on top of it. So, I encourage you to give it a shot if you  haven't. runth.

(05:32) com and go and see what you can build. Got it. Makes complete sense. Um, do you  feel like having built out that sort of software beforehand sort of gave you the knowledge  base or the tools to sort of get started and,   you know, start building on your own? Or, like,  what sort of gave you the confidence to, like, get into cloud code and, like, what were the some  of the first things that you did? I think it did.

(05:59) I think it um, you know, it helped me navigate  the room better, right? When you're trying a new, um, let's say you're doing media buying on a new  channel or you're starting to try to do cloud code   whatever, you're in a dark room. Like, you need to  figure out how to turn the light on and navigate around it.

(06:17) And then, once you've done that for  a while, it's a lot easier to move between,   you know, uh, AWS Secrets Manager and versel  environment variables and all these different things, right? So, um, I had a little bit of  familiarity with AWS backend through the project, and really it was just taking everything  that I had done communicationwise with the actual human developers and then translating  that to cloud code. And at the same time, um, just really understanding like what my role was,  because my role was never with the software that I was building with the team to understand, um,  like what the developers were supposed to do tactically. My job was to make sure they  understood what the user wanted to achieve and   that the overall flow made sense for the user that  way. And that's all my job is now is to make sure

(06:58) that I understand the project management workflow,  how we can improve what we need to do for our team and that whatever cloud code outputs achieves  that. Got it. I think it's a really good frame of reference. So effectively like you are the product  manager, right? at the end of the day like you are orchestrating Claude. You have a hypothesis.  You have a thesis.

(07:28) You have a bedrock of   knowledge that you're sort of building on and  effectively using Claude as your executor, but like at the end of the day, you're sort of the  marionette, right? You know, pulling the strings.   Um, and so let's sort of dive in to, you  know, what you've built and like how you sort of got there. So, like it seems like  the problem was there's these really slow,   clunky systems with UGC.

(07:47) You know, right now  everyone's sort of like talking about Tribe, which is this sort of pay-to-play creator partner  network where, you know, I think creators make   1.5% of ad spend and nothing more than that.  Um, so if their ads take off, great. If not, then you don't have to pay for anything.

(08:07) Um,  and so like candidly on a lot of sales calls,   you know, we don't do creator sourcing, but one  of the first questions I always get is like we have this sort of dashboard that we've built  like couldn't you just like find creators with   this? And I was like, "Yep, maybe." But like  it seems like it's a pretty daunting task. So like can you maybe walk me through like sort  of soup to nuts like how did you start this? Yeah. So the you know the tool the build actually  did not start at the beginning of our process.

(08:28) Um so it's okay with you. I'll kind of start like  where the build started and then go like linearly   through the process from start to finish. But um  the tool actually started with like what's the most urgent problem that we need to solve, right?  And at that time it was um two main things. One a frustration that when creators would upload  content into Google Drive they would not notify us about it.

(08:57) So our team was always on standby  to try and figure out when the content had come in from the creators and at the same time keeping  up with client feedback so we could minimize the opportunities for a client to have to repeat the  same thing multiple times for multiple creators. So the tool actually started out dead in the  center of the overall UGC process which was the QC phase. Now, taking a step back and then we'll get  to like what that dead in the center looks like.

(09:24) What we've built out now has been slowly  added on to week over week for the past   couple of months. And uh our team will test  out a new feature, send me some feedback in Slack. I'll screenshot that, give it to say  fix this. But we we try to own the entire start to finish. What I mean by that is  we've built out a form that is a 15step form.

(09:49) When we see a creator that we think might be a  good fit because we follow the mindset of ABR,   always be recruiting, right? We do that for our  talent for our agency, but also for our creators, we will send that creator the 15step form. It asks  them, of course, like name, age, gender, location, but a lot of other filterables because all of this  then once they get approved goes into a database.

(10:11) So, do you have a pet you shoot content with? If  yes, what type of animal? If dog, what breed and what age? Do you have kids you shoot content  with? If yes, list them out. Girl, 11 to 13 years old, boy, 1 to three years old. Do you have  visible tattoos? Are you licensed or state board certified in any of these things? We ask all of  these questions.

(10:37) And then we also get them to tell us the industries they work in. And for  each industry, they're given the option to   upload up to three video examples. We  do all of that for these reasons. Uh we have a database at this point of over 600 UGC  creators. We have a client come on board. How can we go through 600 people very quickly and find  who might have a Labrador Retriever or who might have a four fouryear-old son if we're looking at  Canva decks? It's impossible.

(11:07) So you have a I I just want to zoom out real quick just because  I think this is a really important detail.   So you have 600 creators in your network. That's  insane. So like how are you getting people to even complete this form? Like is there like an  incentive? Like you know what is that process?   No, it's just, you know, people we've worked  with over time and also just, you know, the opportunity to um have projects.

(11:30) And at the  same time the, the negative feedback that we did get from some is that they don't want another  platform to keep up with, but that's not what   this is. This is not like a platform for public  facing. This is just a database of creators. So, you know, very well some don't fill it out. Some,  you know, have rejected it. But more often than not, we find creators are open to that. Got it.  Okay.

(11:51) And so is the system automatically reaching out to creators or like how are you curating that  list? Good question. So um when a client fills out their onboarding form which is type form, we  ask them some initial filterable questions, age, locations, so like cities, country, um languages,  gender, all these different things. and we have that directly routed to our platform to do a  pre-initial filter.

(12:22) The other thing that we have is we built out a feature called opportunities,  right? And imagine it's just like it sounds. Uh our UTC manager will go in there, she will  assign the name and then put all the filterables that we need for that specific opportunity. And  then once she has uh completed that opportunity, it will go into our database and autoshort  list people that it thinks might be a good fit   based off that.

(12:50) And then she can go through  and either keep the check box for them or uncheck them and then click a button that  says send invite. So we technically have three ways that we invite people to our datab or to a  project. One is they're already in our database and we just send them an invite. we already  have their videos, their card. They just confirm   through a button through email that they're  interested.

(13:12) The other is we have their email and we don't work with them. They're not in our  database. We will put a string of emails separated by commas into this field. And then we will hit  invite and that'll give them a front-facing form where they have to upload a headsh shot, videos,  and then all that. The third is that exact same form that's frontf facing but just a public link.

(13:36) So when we're posting on Reddit, on Twitter, in Hive House, which is Foxwell's community, which is  great for you to see creators, we can have people   just apply through that that way as well. All to  try and minimize email back and forth to make our teams a lot easier. Got it. Okay. So there is a  human in the loop that's effectively helping vet   this initial process. Of course. Yeah, of course.

(13:56) There, you know, there are some things when we get to the QC phase like there is some things that AI  can do. It can check noise levels. It can check um lighting quality in a video because those  are like those are um scales if you will, but AI can't tell you if a video just is good or  not. So they're definitely still has to be a   human to assume this. So the goal is that this  tech alleviates a lot of their their burden.

(14:19) Got it. So you you effectively just outsource  like a lot of that pain in the ass admin work to   the goal with everything we build this  tool, everything else is get all the admin work out. So our team can use their brains  for what their special is at. Got it. And   I do want to double click or take a brighter  click on what you're building.

(14:39) But you know, you know, I I think you raised a good point on  like the admin thing like how do you even figure   out like what are the biggest admin headaches?  Like how do you figure out what are the problems that you need to solve in this process? Yeah. Um  two ways. one is just asking the team and just kind of getting their feedback because normally  you know you can open the floor and and they'll   share you like what is the biggest pain point for  them.

(15:03) Um we have like a you know town hall if you will but just asking everybody to list out like  what are the things that are taking most of your   time. The other is time tracking. We use Claude  or Claude, we use ClickUp for time tracking, connect Claude to that and then have Claude do  a time analysis of everybody by client and by task type so we can understand what is taking  the most time for people. Got it.

(15:28) And then you   feed that into Claude and Claude helps you say  like, okay, these are the admin tasks that we can alleviate the most pressure from. Yeah. Like  for example, we just like a senior media buyer,   we just found that like 30 35% of their  time was going to ad campaign setups. just not nothing t like nothing strategic. This is  just turning off optimizations and meta and all that type of stuff for their side campaigns. If  anybody listening has a tool that actually sets up   um campaigns that is not just duplicating  an ad set would love to see that.

(15:55) But because of that we just brought in like someone  outsourced that is just specialist at setting up campaigns. So that's all they're doing.  Right. So that's how we use time tracking.   Got it. Okay. That makes sense. Let's go back to  the process. So, we have a network of creators that are built in uh in this database. A client  fills out a sheet.

(16:15) You're using Clawude to do like sort of a best match in terms of like these  are the creators that would work. Um what's next? Yeah. So, once the invite gets sent, then the  uh creators will apply. We set a deadline for like the availability for that  opportunity. Once there's the   submissions, our team goes through and reviews  and then they will create a short list.

(16:38) There's another button in that little opportunity screen  that says uh generate short list. And what that short list does is it creates a public facing  URL. So the client can then go in and see uh four cards um horizontally or four cards beside  each other and then you know limitless vertically. Each card is a creator. They click into that  card, a modal pops up or a screen pops up.

(17:05) It plays the creator's videos and then it has  an approve or reject button. If they approve that creator, we originally built this with it  being front like a wrapper on top of ClickUp   because we were using project management and  ClickUp. We've transitioned away from that. When it was in ClickUp, approve would move that  creator to a status called client approved and rejected. client rejected. Similar in our tool  now.

(17:34) Um, and then once they've reviewed everyone, it notifies our team in Slack so they can go  into the portal, see who's approved and denied, and then from there, they will click another  button that automatically sends the rejected creators a notification that we're not going  with them. The client went another direction.

(17:56) And then for the ones that were  approved, they click another button that automatically connects to Panda doc, which  I think we're going to switch away from because   their API is extensive, and automatically sends  that creator a contract to start their engagement. Got it. Okay, that makes you go ahead. Yeah. Okay.  So, that makes sense. So, yeah, at the end of the day, like it sounds like a lot of APIs are sort of  hooked up in here.

(18:20) So like when let's say you know probably doesn't make sense to sort of distill  down every API that's being utilized but I think   one thing that uh a lot of folks struggle with is  just like the the Instagram sort of API sort of sucks right like it you can't really you know do  DMs very well from it. So like you know are your creators constantly checking that dashboard to see  if they got approved for something or not or like   how are they getting notified? Are you shooting  them a DM via Instagram? like you know how is that what does that look like? Yeah, they don't  have to check the platform. They it's all through   email. So every action that happens is sending out  an email either to the client or to the creator.

(18:55) Got it. Okay. And then so you're just you just  have like the Gmail uh API hooked up effectively. Yeah. Mhm. Got it. Yeah. We're using we're using  we're using resend. Resend. Perfect. Okay. Got it. So the creator would get an email. They can  hit like accept decline.

(19:16) that feeds back into   the platform and then what happens like products  get sent out like briefs like what is the whole process? Sure. Yeah. So, um just imagine like  there's a little you know card or a little um section if you will that is that creator, right?  We've sent them the contract, they've signed it,   I signed it. That API feeds back into our tool  to notify our team that the contract has been signed by both parties. And then that's when we go  into brief writing as separate. But once the brief   is ready, that creator has a card.

(19:43) Think about  ClickUp and you see ClickUp has like the custom fields in a task. So you know like let's say one  custom field might be um brief link and then the next custom field would be number of briefs.  The next one would be number of hooks per brief and then um the next one would be deliverables.  We have all of those originally in ClickUp now in our front-end tool. And so they'll go through  and they'll outline scope for that creator.

(20:06) So let's say two briefs, three hooks per brief, um  provide both of the brief links from Google Drive and then they will put the deliverables 9 by6, 4x5  and RAWs. When they do all that, they click send um send brief or send scope and it will  automatically generate them a Google Drive folder. It will automatically move those um we use Google  Sheets for our briefs.

(20:34) move those Google Sheets into their new Google Google Drive folder. It  will automatically create them an upload portal that has drag and drop areas to upload creatives  for every brief that they have. And then it will send them an email outlining next steps and what  they need to do to get started. They click through that email and there's a screen that says start my  brief or accept my brief.

(20:57) As soon as they do that, their clock starts and we start counting  down to the 3 to five day turnaround time we offer for them to produce their content. Got  it. Okay. And just to take a few steps back, the brief generation process, like what does that  look like? Is AI generating the brief like based on context like you know what? Yeah.

(21:21) What's that  process? We've done a lot to um to automate or use AI for like the research side of things. Uh, so  like, voice of customer, competitor research, and   all that, we are still very humanled when it comes  to the brief writing. I'm sure our, you know, copywriting team jump starts with claw and stuff  like that, but you know, there's a lot of, um,   there's a lot of opportunities still, I personally  believe, for AI to improve on copyrightiting, so a lot of that is still human le. Okay, cool,  that makes sense. And so, like, what about,   like, getting the product to the creator?  Is that automated? Like, how does that work?

(21:53) We, uh, we could probably improve this. I'd love  to know if you know anybody listening or if you have any thing that you do. But what we will  do is the client does receive that creator's card and it shows their address and then we have  a field uh in that card that says insert tracking link because another pain point that we've had is  either clients not sharing tracking links or just   not knowing where the product was.

(22:21) And then once  they insert that tracking link and hit submit, it will go back to us so that we can keep up  with that. I'm actually I was trying to work   this week on building out a crime job so  that it could scrape the tracking page to monitor when it got delivered, but it  seems like it's consistent with USPS,   but not with FedEx and and uh UPS. So, that's  one thing we're trying to figure out right now.

(22:41) Yeah. Yeah. No, I've definitely experienced that.  Uh it says it's delivered and then you get it   like two days later or something like that. Yeah.  Yeah. Yeah. Um, okay, cool. That makes sense. So, uh, the creator gets a product,  they shoot the footage on their own,   all manual, no real AI built into that, and then  what? They just upload it back into your platform.

(23:08) Hey, this is Andrew from Foxwell Digital. This  episode is sponsored by Aliia. Quick gut check.   What did it cost you to get someone to your  site last month? Now, think about how many of those people left without giving you any way to  reach them again. That's money you already spent   walking out the door. Often, the fastest way to  grow your email and SMS list isn't more traffic.

(23:27) It's capturing the traffic you've already have.  And that's what Alia does. It's an AI powered   pop-up platform that turns more of your visitors  into high quality subscribers without touching ad spend. More than 5,000 brands use it like Nike  Strength, Pelaton, Grunes, Hollow Socks, Marsmen, and the brands commonly see two to three times  more signups in their first 30 days.

(23:52) So go to Alia popups and book a demo. Our listeners get an  extended free 45day trial. So make sure you   tell them that Andrew Fox will set you. They edit.  So we have creators edit, which I know a lot of agencies don't, but we have creators edit. And  then they go into their portal. I can actually um I can show you this one piece really quick.  Let me just try and pull up screen really fast.

(24:15) Um and I'm also curious after we sort of walk through  that like have you considered or explored you know using Claude as a sourcer uh in terms of finding  new creators? We have not yet. That is one area of opportunity that I think we would have. Um we  have not done that much. We did just actually bring on somebody um like outsourced so that they  could handle just focusing on review of things or finding creators. Um but no, we haven't. Have  you ever tried that like with influencer or UGC   in general? Not myself, but I have a client  that's built out a whole system that will like

(24:54) you know reach out to creators and like that I'm  something exploring as well, right? Because like   we have for all of our clients, we know like  the personas, angles, formats that are working. We know the net new personas that we want to  test. And so it feels like marrying that to   finding creators is like a logical next  step.

(25:14) But this, the question is just like, do you just like have Claude go out uh,  you know uh, maybe you use like Sonnet   uh, find creators, and then you use like Gemini  Flash to like analyze their videos and like try to match up, like, you know, if you're selling a pet  product, like, "Here are like 100,000 pet product   creators," or something like that? Like, I think  it's doable.

(25:35) I just like don't know how reliable We, so we have not done on the UGC side of things,  and I've learned from many sales calls good to   define actually: UGC is content creator, no social  following, not posting organically, they're just shipping you back ad content; influencer being  like, uh, someone that has a following and then goes that direction. Those two get like combined a  lot.

(25:56) On the influencer side of things, we did just launch last week, um, a tool where, so, human will  find potential profiles on this is specifically for a YouTube influencer client or a client that  wants YouTube influencers and then our UGC manager will verify that they're a good fit. We'll put  them in a Google sheet and then we'll click a   button.

(26:22) It will go through with Amplify, scrape  all their profiles, scrape um the most recent 10 to 15 video titles, descriptions, and transcripts.  And then it will feed all that to Claude and Claude will consume those transcripts and so  and titles and then generate a three-step sequence cold email based on the transcript. So it looks  like a human who watched their content is reaching   out to them to increase our response rate.

(26:44) So too  early to give you data data on that because we just started running it last week, but that's the  only thing we've really done for sourcing so far.   Got it. Okay. Yeah, that makes sense. It's uh it's  an area of massive opportunity, but anyway. Yeah, let's see. Let's see this. So, this is we have a  test environment for our tool.

(27:04) Anybody listening,   highly recommend once you actually have a tool  that your agency is using day-to-day and relies on it, don't just ship it to that. Don't ship  updates to that. Like, ship it to test first. So, we have a test environment. This is when I said  earlier that when we send a brief to a creator, it   creates their Google Drive and also their upload  portal. This is what the upload portal would look like.

(27:24) Um, and so just imagine, you know, this is a  brief where or a project where each brief just has one hook, but this would be multiplied. So, you  know, they uploaded the the 9 by6 content. When they upload the content, we do have AI check to  make sure that the aspect ratio matches what the aspect ratio is supposed to be for that asset. We  have it check the first 5 seconds uh in increments to make sure that there is subtitles and if it  fails there it checks the next 5 seconds and if it   fails there it checks next and if they don't have  subtitles and they're supposed to it autorejects their video and makes them re-upload it. And  then um we also check caption placement. This

(28:01) is something that we need to perfect a bit. Right  now, we're mainly just tracking to make sure it's   within safe zones, but we need to add it to make  sure it's not over face as well. Once they upload the content, it then goes to our QC section of our  tool that uh looks something like this right here.

(28:21) And our team gets notified. Once again, this all  started because our team had a painoint of not   knowing when creators uploaded content. So, in  our creator portal, which is one of the tabs, they'll go in, they will start.  Let's see if I have an example here.   Uh, no. They'll click the button that says  start QC, and then they watch every video.

(28:43) We have a 20 a 20 point QC checklist. It  says 21, but the actual one is that some of these are automated. So, no important content,  bottom 20% of screen, um, you know, sound levels, echo, reverb, those types of things. Some of  those get pre-checked by AI, some don't. They   go through this checklist and then they also  leave timestamp comments on the video as well.

(29:08) From here, once they have done this for every  video, they hit send a client for review. The client gets an email with their portal. They  go in, they get a very similar screen. They see our comments, they can leave their comments, they  can approve a video or [ __ ] back with revisions every time that we leave content.

(29:33) And to me, this  is why it's so important to own the database of creators. Every time that we leave comments  on a creator's content or a client leaves comments on a creator's comment, two main things  happen. First, when the client leaves comments, all that gets fed through Claude so that  it creates a client learning QC list.

(29:56) Because everybody listening who sells a creative  service has definitely experienced where you ask, "What are the brand guidelines and brand  specifics?" and then first brief or second   brief in, there's new things that you never heard  before, right? Yeah, totally. We always wanted to solve that problem of how can we keep track of  that. So, this is how we do that. The second one is we want to make sure that we are keeping  quality creators in our database.

(30:19) Our first you know yes or no on them joining our database is  what their content they show us looks like right. So if if Jordan who this demo fake name is submits  content and our team and the client says poor lighting poor sound and then on the next project  we and that client say poor lighting poor sound. All of that gets aggregated into their database  card so it can flag that multiple projects they've had that issue and it'll move them into a um like  safety security flag area so we can potentially move them out of our database. So we keep we try  to keep track of all of these types of things and

(30:56) we also at the end of a project have multi-star  turnaround time video quality communication rating systems one through five so we can give them  star ratings as well. Got it. So, that's really,   really interesting. And I think, to your earlier  point, like we've all been in a position where, you know, there's been a 180 on brand guidelines  or maybe they're like, "Hey, like, you know,   we need 16 pixels between each of these things,"  right? So, like, it's helpful to sort of have that in the database. Like, is your QC checklist also  being updated, you know when, with that database

(31:28) information? that's why I said it's a 20point  checklist that says 21 at the very bottom there's   a list of all the client feedback that gets  aggregated every time new stuff comes out. That's awesome. That's really cool. I am definitely  stealing that. All of that also go All of that also goes the creator portal that I mentioned um  or showed you earlier.

(31:49) There was a little yellow box at the top. That stuff goes to the creators as  well. And not just one creator, but all creators. So, if it's not something that is that creator  specific, it gets sent to all creators. Got it. Interesting. Okay, cool. Um, and then you  know, let's talk performance, right? Like, okay, so you know, a client approves a piece  of UGC content after a revision or two, like how is the performance of that ad affecting  future content that you're generating? That's   a good question. So, that goes completely  outside of this tool and more so into the

(32:24) tool that we spent three years building, which  I wish we would have waited two two and a half   years to build the code versus you are in great  company for what it's worth. Minus. Yeah, could have saved six figures on software that I never  did anything well, we used it at the agency but, um, we built out like a full... what started as a  creative analytics software.

(32:46) So, think like Motion ,for play, Atria; um, and we tagged things across  nine key categories: UGC, creator, messaging angle uh, creative format, product or feature, all  these different things, right? And then... through naming conventions our software auto tags each of  those with the specific thing. So it might be uh John versus Rachel and two different ads and then  all of that gets moved into our central creative database or central creative dashboard where we  have a tab for every category.

(33:17) So UGC creator product or feature and every row is the summed and  averaged data for all ads that match those tags. So that's how we start, like especially with UGC  creator. And you know, we don't do it currently, but once again, this is a... the UGC tool is  something we've been building over time. Um, we plan to integrate the two so that we can loop  back, uh, KPIs into the creators database cards per industry as well. Cool. Okay, that makes a ton  of sense.

(33:52) And do your clients also have access to that tool that you've spent three years building?  Yeah. Yeah, they do. Yeah, we uh we're trying to   build that one out quite a bit more. So, it  started as a creative analytics platform. We just added like a full Shopify connection and a  full P&L customized builder so that we could like have a lot more um buyin from clients, but also  financial transparency for our media buying team.

(34:15) Um so, yeah, they do get access to that. Got it.  Okay, cool. Helpful. Um, from a UGC perspective, coming from a guy that is not a UGC guy or  influencer guy or whatever the [ __ ] it's called these days, uh, you know, like how much of a value  ad has this tool been for your company? I think it's been huge. I think, uh, value ad for the  company and then for the team and for the client,   value ad for the company.

(34:45) I think it's it's great  to be able to show clients no matter what service that you sell in the sales process that you have  some sort of tech built out to make your process replicable. So you're not so human dependent.  I don't mean that in any negative way, shape,   or form, but there's a lot more confidence  in, like, okay, you're good enough at this to where you've been able to build out a program  for it and, like, I can tangibly tangibly see it   and know what my experience is going to be like,  right? So I do a full I do a full demo from start to finish in most of my sales calls. Um just kind  of walking them through the process and that tool

(35:16) and then team I mean it's just made our team  a lot more efficient but at the same time just made their life easier. And I think there's a  there's a fun part of it where the team can send   me a Slack message, I can screenshot that, give  it to Claude, and within 15 minutes their bud fix is done and they can just keep using it, right?  Or the new feature that they want gets built out.   Yeah, that makes sense.

(35:44) I think uh there's  something that you mentioned that is probably slept on honestly by myself as well. Uh testing  environment. So like let's talk about sort of testing versus shipping live. like what how are  those set up? Are they they GitHub? Like you know,   how are these sort of set up? And then like  how are you the one that's going and doing all of this? Do you have a team member doing all  of this? Like how are you deciding like what   gets shipped and what doesn't? Um yes, right now  it is just me.

(36:10) I lean on to claude quite a bit for full end toend testing. What we generally  do, so if I'm building something brand new,   I'm not going to like day one build a test in a  prod environment. I'm going to build it to the MVP point to where it's actually going to go into  actual use or what we call actual production,   right? To where our team if something breaks,  it's going to impact how good our team can do their job. At that point, I will build out  the test environment.

(36:35) So, what I will do is honestly just tell Claude, um, you know, we  need to take this current product that we have   and duplicate it into a test environment. Give  me a clear step-by-step process of everything you need me to do and at the same time you do it.  I have claw directly connected to superbase AWS um uh versel everything I can and give it all the  permissions that I can.

(36:59) So it can essentially do all of that and very quickly for reference for  anybody listening MCP or API just real quick whatever the direct connectors are in cloud so  I guess it's MCP yeah but for reference like the   software that I built out uh with human developers  it took us probably like a week and a half to two weeks to build out our test environment.

(37:22) Um it takes Claude maybe like an hour to two hours. Like it doesn't take a ton of time.  And so what would I tell Claude? Like if I have and I'm doing this right now, but like if I have  a database for my team and I want to just like, you know, potentially test net new ideas in a in a  in a testing environment, what do I tell Cloud? Do I just say like, "Hey, can you like clone this?"  Like, how do you link the two? How do you link the   testing environment? How do you link the live one?  And like, yeah, what does that even look like? So, are let's just assume you're using like Superbase  or something in this example, right? Or AWS,   right? Um, Superbase is a lot easier. It will  start by duplicating everything structural wise.

(38:01) So, it'll duplicate your database. It'll duplicate  your front end. It's a completely different URL,   completely different um um like Versel is  the same thing, but completely different environment variables because you can choose  deployment, production, test, all of that. Um,   and then from there, everything is still linked  in the same git if that's what you're pushing to.

(38:22) It's just when you're doing when you're doing  builds going forward, I always tell it um push to test before we push to prod and it'll push  to test. And then I will tell um Claude run a full E2 or end toend test and make sure that  you test all potential fail points and then outline what needs to be fixed. So I'll let it  do that.

(38:50) What I've learned Claude is very bad at, unless you know I could just be missing something,  is the uh this makes sense for how we want it set up versus not, but it's good at checking all the  APIs, the documentation, all that stuff on test. Yeah. Yeah. It's also really good at leaking your  APIs if needed. Uh but no, but I think you you raised a really good point.

(39:18) Um and I guess just  like for like our power users that are listening,   like what cloud models are you using? Is it like  Opus, Fable? like do you have an orchestrator? Like are you just using this in Cloud Desktop?  Like what does your process look like?   Pretty simple guy probably. I uh I use Cloud  Desktop. Um I'm of the type where like if it's a simple fix, I'll just use like Opus 5. Um but  if you know, especially like on a Thursday when I need to max my credits out, I'm Fable 5, Fable  5.

(39:41) 1, like whatever I can do to try and get things built as as efficiently as possible. Got it.  Nice. Yeah, I'm I'm sort of the same way. Uh max out Fable ASAP. Like if I buy, you know,  and the weekend Fable is not maxed out, I'm   doing something wrong. Like I I'm going to go ask  it like what what should I buy for groceries this week? Yeah.

(40:02) I like to just make sure that like,  you know, I I I fully understand there's still a lot of value in in expert human developers because  I have no idea what under the hood looks like with my codebase for projects like this. Um, but I  like to try to use the best model possible to minimize that as much as possible. Got it. Uh, we  did an emergency pod this morning about Fable 5.1. I'm curious, uh, you know, as we sort of wrap up  here. Fable 5, Fable 5.

(40:31) 1, like have you tested   it? You know, what differences have you seen?  Like, where is it better? Where is it worse? I have not been able to uh I think it moves a  little bit faster but I have not been able to really give like a clean end to end. What  do you feel like is better in in those two? Yeah. Um twofold. One uh it's really good at QCing  your past skills and markdown files. Like I think one thing that was happening and this is going to  be redundant for the other pod but that's fine.

(41:01) Like one thing I was finding is that like Claude  like was taking so long to like QC static images that it was generating for like proposals and  I was it would take like 6 hours and then I'd   like go look at the statics ads and I'd be like  this is fine. This is ready to ship. And it's because like I had like two lines in the markdown  file that was just like putting it on this like   loop. And so what I had Fable do Fable 5.

(41:21) 1 do  is like I need you to go QC all of the skills I use more than twice a week and I need you to  clean them up. And it literally did that and like it made a massive difference. The other thing  I will say Yeah. Go ahead. Have you uh I told you we have um AI check the UGC for subtitles.  Claude told me that flash will be better than claw for that. Have you tried flash for that? No.  But 5.1 is it it's really good at subtitles.

(41:46) It can also do translation very well. If you want to  like go to a different language, you can upload a   video. It will translate the entire video and like  also adjust the subtitles. It's actually a pretty good editor. Uh okay. Yeah. Okay. I'll try it out.

(42:00) And then you know my newsletter tomorrow is on this but like it's also a much better copywriter.  Like it doesn't have like those sort of like bolds   and like you know paragraph or like if it's not  this it's this. It feels a lot more organic. Obviously huge asterisk here. Like if you are  going to be you know posting slop on any you know newsletter or LinkedIn 5.1 will have the  watermark. So people will be able to tell that   you're using 5.1 copy.

(42:27) So like pick that you know  and you know think about it carefully before you do anything. Um but you know to that end it is  a much stronger copywriter I found uh as well. Awesome. Yeah I'll have to I'll have to take a  look at that. Cool. Um any sort of final thoughts you know for our listeners like if you were to  do one thing today and you're starting from zero uh what would you recommend? I'm trying to  make this as as short- winded but it might   be longwinded.

(42:56) um take the pressure off yourself  of feeling like you need to know anything about code development. And if if you are trying to  build something, all you need to understand is what's your goal outcome? What's the problem  you're trying to solve? And how can you tr communicate that to Claude well enough for it  to build something? If at any point Claude asks   you to go into AWS and set something up, go  into Versell, literally say, "Give me an easy non-developer step-by-step process to follow for  this with links. Make it easy.

(43:25) " And if it doesn't say, "Make it easier." Make it easy for yourself.  Claude is there to, you know, really navigate the whole journey so that you can just focus on being  the architect of the overall build. I think that's so well said. And like I have one followup on  that. I feel like a lot of folks get swayed when Claude tells them to do something a certain  way.

(43:56) And I think you know to your point like being   the architect is very very important, right? So  like if Claude makes a suggestion to you like we should try this like and it differs from your  initial thesis like how do you sort of think about   that? Like if it differs from your initial build  and it says like we should try doing this instead like how do you approach something like that?  Challenge it just like I would a person else.

(44:14) And you know, it falls back to, you know, I'm  not going to challenge it on like what language to write in or what this or that because I  don't know anything about that, right? But,   um, man, I'm trying to think of a recent  example. I had one yesterday, um, if it will... Oh, so we just actually, uh, yesterday just did  the update, because normally hooks always have, uh, subtitles in the first 5 seconds of UGC video,  but one video did not have subtitles in the first   5 seconds because there's no speaking, and it  was causing problems with being able to upload because he kept taking back. Claude gave me a  recommendation on how we could fix it. And I said,   "Actually, shouldn't you just measure the next  5 seconds first and then do this?" And it said,

(44:51) "Yes, that's a great idea, but Flash is  already pull flash is already pulling all   of these screenshots because we're doing another  process over here. So, we already have what we need. What you want to do with duplicate  work." Okay, cool. Let's do that then. So,   I mean, just, you know, challenge it and see  what it says.

(45:08) Yeah, I think that's a really really important point that folks miss is like  you can push back on these models, right? Like   you are pulling the strings like don't fall into  the trap of just like saying what would Claude do, right? I think that's really really important. Um  and yeah, it's just so easy to just be like, oh, you know, Claude told me to do this, so I'm just  going to do that.

(45:30) Claude has zero context of um how your agency operates and what you're trying  to achieve. It has a lot of context on API   documentation, all that, you know, fun development  stuff, but it's it's our job to make sure that we are directing cloud to reach our goal outcome. Its  job is to do it with the best code build possible. Yeah, totally agree.

(45:55) And then like I guess one  more question then we can wrap is like how do you ensure that you aren't falling in your that vortex  like that that the claude is sinking for you rather than you being the orchestrator.  That's tough man. I, you know, I've uh,   I'm on the, like, I'm in such a on the fence spot  about this cuz, like, at at one point I'm like, man, you know, you think of, and not relating  myself to this at all, but like you think of,   like, a high up Fortune 500 CEO, like they have  a intern and they have a receptionist that does a lot of the stuff, like responding to emails  and stuff like that. I'm like, okay, that's   like what claude is for the common personnel.  But then there's also times where like no,

(46:29) I need to like use my brain and just respond to  this email instead of copying and pasting it. I'm   on the fence. It's tough. It It is tough. Yeah,  I completely agree. Uh, I completely agree, and like, I was definitely AI redpilled for a while,  and I've sort of come out of that the other side. Like, I write my own emails now; like, I write my  own newsletter entirely. Like, and it just, like,   it makes sure I I'm sure, ensuring like my brain  is, you know, being utilized properly.

(46:56) So it's easy to fall into that trap. Yeah. it becomes, you  know, I think it's very easy for anybody to be to come over reliant on it and it's like what is the  what is the balance? You know what I mean? Like   what I guess Yeah. It's probably different  for everybody, but it's it's a tough one. Yeah. It's like if you want to revert to the mean,  just do whatever the quad says. Anyway, all right.

(47:17) Well, thank you, thanks so much for joining us. Uh  stoked to uh get this live and uh yeah, that's it for today. Thanks, guys. Thanks for having me.  We have an awesome tool that's a little bit of   a secret and I'd like to tell you about it. It's  called the Foxwell MCP. The Foxwell MCP connects into Claude, connects into your ChatgPT Pro and  it gives you direct access to every single thing   we've ever said in the Foxwell Founders community.  Check it out at foxworlddigital.com/mcp.

Made With AI

Next
Next

Using Claude and Codex without getting locked into one platform | Dustin Thompson