One Designer + AI. Hundreds of Deliverables. — Vincent Wendy, AI Engineer

AAI Engineer
Computing/SoftwareAdvertising/MarketingManagement

Transcript

00:00:00Hello everyone, hope you guys having a good time at the conference, so before we
00:00:19start, how many of you are actually designers, like a product design? Hey, one
00:00:24hands, and another, okay, and how many, I assume that the rest of you are engineers, is that
00:00:32correct? Yeah, pretty much, okay, so today's talk is a non-technical talk, but more of a
00:00:40real-world experience, how I created the design for AI engineers, this conference, and our
00:00:46other past conference as well, and how AI has helped me, and so the talk today is one
00:00:54designer plus AI, which is me as the designer, and hundreds of deliverables, right, let's
00:01:00start. So my name is Vincent Wendy, I am a senior creative designer at AI Engineer, and
00:01:07at AI Engineer, it's a very small team, so we only have around 12 people to 15 people at
00:01:15the moment, and everyone has been doing their own thing, and I think AI has been a
00:01:24really helpful way to, like, helping everybody doing everything, and for event at the scale, we
00:01:33have a problem, obviously, right, and the problem is the scale problem, or I would call the challenges,
00:01:40and how to overcome it? It's basically automation, and we get to that in the later part of this talk, so when I prepare this talk, we only
00:01:53expected 6,000 attendees, and now it's 7,000, well, good for us, and then we have 140 sponsors, more, 140-plus sponsors, and then 300-plus speakers, 600-plus sessions, and one designer.
00:02:10And everybody needs every design, right, like, every single thing needs design. Sponsor needs assets, speaker needs graphic,
00:02:19you need signage, so you don't get lost, and this is basically what we do, what I do. So from stickers, do you like your swag, your stickers?
00:02:31Well, I hope you do, because I create that design, too, and to a landing page, speaker announcement, track mascot, all the stuff that you see,
00:02:41all the stuff that you see, most of the stuff that you see here, from a signage to a digital signage, landing page, everything is a
00:02:51a deliverable, and 1,000 details means 1,000 ways to fail, right? Because a missing sponsor logos is going to be a huge issue,
00:03:04and speakers that have a wrong schedule also have huge issues, right? And it seems impossible to handle that many kind of deliverables, but, yeah, meet my design team.
00:03:16So it's me and Devin, GPT, and Figma. And right now, we are at the stage where tools isn't the, like, it's not a problem anymore,
00:03:31but having a real problem is our advantage. So, for example, when someone asked me, "What inspired you when designing in AI engineer?"
00:03:41I don't know the answer back then, but after I think about it, it's actually a problem,
00:03:46that inspired me to, like, designing in this AI engineer, and we'll get to that in the later part of this talk.
00:03:53So, have you guys seen the talk by Simon Wilson, like, in 2025?
00:04:00Yeah.
00:04:01Yeah, and it's pretty interesting, right? He asked to -- he asked every LLM to create a factor file,
00:04:11which is basically a pelican riding a bicycle, and it is basically to test,
00:04:16and I test it again, and it's still doing this for the basic model,
00:04:20and it's not usable for me as a designer. But as a designer, we have to think outside the box,
00:04:26and we could simply ask that GPT create a stale image, like a PNG for a pelican riding a bicycle,
00:04:33and then I can factorize in on Figma, and we can ship that now.
00:04:37So, we have to think outside the box here, regardless of the capabilities of the LLM.
00:04:44So, how to solve this scale problem, right?
00:04:52Basically, five things. So, foundation first, reasonable designs,
00:04:56automated workflows, validated output, and also removed frictions.
00:05:00The foundation is definitely the core part that we need to set up right,
00:05:06like the design system, typography, colors, components, like other stuff.
00:05:11And once this is set up, like, for example, when we create the website,
00:05:17it's all set up within this thing. And, yeah, this is just an example.
00:05:22Like, we have the colors, primary, and then also the accent colors,
00:05:26the typography, and also the tech and all the other stuff.
00:05:31And, also, have you guys -- are you guys familiar with the atomic designs?
00:05:37So, yeah, my previous background is I'm a product designer,
00:05:41so I'm pretty familiar with the thing where we need to create a user-centric design,
00:05:46and also, like, atomic designs, right, where we create the smallest part possible,
00:05:50and then combining it into, like, basically a Lego pieces,
00:05:53and then into deliverables. And this is pretty useful in my job desk right now.
00:06:02So, once we set up all of those foundation, we basically need to create --
00:06:09for example, we use Devin a lot. At the office, we -- everybody use Devin.
00:06:14Everybody, like, abusing Devin, for example. Yeah.
00:06:18And so, in this case, I just need, "Hey, we use this desktop typography
00:06:25and this mobile typography," because we know
00:06:29Cloud or, like, any other LLMs love to, like, throw in some random font size, right?
00:06:35And if we don't define it, it just delivering a slope, like the previous slope.
00:06:41And yeah, typography, color, and stuff, and then it comes to reusable design.
00:06:50So, once we set up it right, like, the website is -- has the branding to it,
00:06:57all the other themes on the AI Engineer, like, for example, the marketing themes,
00:07:02can create everything, basically. Like, they can create an email design based on that.
00:07:08They can create a flyer, a document just based on the website,
00:07:13because it's already defined -- like, it defined early.
00:07:18And yeah, once you get the design, you can just rinse and repeat.
00:07:24For example, the mascot, it all has the -- pretty much the same design,
00:07:28and it's rinse and repeat. And if you're already defining those things,
00:07:34you can basically, like, create one design that works for all.
00:07:38And this is the part that I'm most interested to talk about,
00:07:41which is the automated workflows.
00:07:43Before, for example, if you take a look outside the room,
00:07:48there's a schedule, right? The schedule for each and everyone.
00:07:52So, we used to do it manually on Figma, but now we use Devin for it.
00:07:58And let me show you.
00:08:02Hey.
00:08:04So, right now, we just pull the latest data.
00:08:08I just ask Devin, like, "Hey, I want this room at this base."
00:08:14And then we can just export it, download it to PNG,
00:08:18and the data is accurate, and then we can just ship it to the flash drive,
00:08:22and then put it on the screen.
00:08:25And it was, like, impossible before,
00:08:28because the friction is just too much
00:08:31between the designers and the developers.
00:08:33We cannot make things, like, pixel perfect,
00:08:35because once we tell the designer,
00:08:38"Hey, this is the design," and then --
00:08:40sorry, the engineers that created the design, for example.
00:08:44"Hey, I need this to be delivered,"
00:08:46and then they don't create it pixel perfect.
00:08:49It's a lot of feedback, right?
00:08:52But with Devin, we just say,
00:08:54"Hey, can you make this more accurate?"
00:08:58We can just connect it to MCP,
00:09:00and then if it doesn't work,
00:09:02we can just always, like, give a spec sheet or something
00:09:06or something that can be defined, like what's the spacing,
00:09:11what's the font size, et cetera.
00:09:14And this is what we do for the speaker announcement.
00:09:20So we have 300-plus speakers,
00:09:22and it's impossible for me to, like, handle one by one, right?
00:09:26So we create this thing, which is called --
00:09:29which you can also access to speaker announcement,
00:09:32and you can also try it yourself, like this one, for example.
00:09:38You can select it right here,
00:09:40and then you can also change your name.
00:09:42Well, yeah.
00:09:44For example, this, you can change the name to whatever you want.
00:09:48And we also have the landscape mode,
00:09:51which can be also loaded.
00:09:53If the speaker also have the headshot and all the details,
00:09:56it will automatically export.
00:09:59And we also have the trading cards,
00:10:01which is surprisingly pretty popular.
00:10:05And we have a different theme.
00:10:07And this is all Pixel Perfect, right?
00:10:11For example, this one.
00:10:14This is inspired by TBPN, so yeah.
00:10:19And how do I deliver this in Pixel Perfect?
00:10:23Let's jump into it.
00:10:25So the process here is, before, when I started my career
00:10:34as a product designer, it used to be just,
00:10:37OK, we need to research.
00:10:39We need to build product, like design thinking in general, right?
00:10:42And then feedback loop and stuff like that.
00:10:45But right now, it's just outdated for me.
00:10:49Like, in my case, we just go to Slack, Figma,
00:10:54and then send it back to Slack, because our Devin lives in Slack.
00:10:58And then ship all the things that he need.
00:11:01Like, for example, if we can connect the MCP or also the spec document,
00:11:07which is, for example, the spec sheet like this,
00:11:10which is a plug-in in Figma, if you're interested.
00:11:15It's free.
00:11:15And it basically gives an annotation to the PDF.
00:11:20And yeah, all designers don't name their layers.
00:11:24So yeah, this is just like some random frame three, frame four.
00:11:29But the LLM will get it.
00:11:31And it's basically defining all this spacing, all this font size,
00:11:37and then all the colors and stuff.
00:11:40It's definitely going to help you develop a pixel-perfect product.
00:11:44And we also have just recently, like today, have photos,
00:11:51which we have to create the thumbnail for its speaker, right?
00:11:55And then we asked Devin, like, hey, who is this person?
00:11:58And yeah, it kind of did.
00:12:01Like, I make a Tinder kind of, you know,
00:12:06detection if this is the same person or not.
00:12:09And I think it's pretty accurate.
00:12:11It's Jason Liu.
00:12:12Yes.
00:12:13And then we can use this to, like, for context.
00:12:18Like before, when we create the thumbnail,
00:12:19we have to search all the codes that photographer have
00:12:23and search it one by one and maybe by time, if possible.
00:12:27But now we can just, like, oh, this is Jason Liu.
00:12:30Download that photo.
00:12:31And then we can paste it into the thumbnail, right?
00:12:34And it's pretty amazing.
00:12:36I mean, the world that we live in right now is actually, like,
00:12:40the state for me as a designer is already at the peak.
00:12:43Because what else can you ask for, right?
00:12:48I mean, we already have things to automate.
00:12:50We already have things to create the design fast.
00:12:54Basically, all you need is a problem.
00:12:56Because once you have a problem that worth solving,
00:12:59you can basically solve anything.
00:13:03And back to my talk.
00:13:04I got sidetracked like that.
00:13:06Yeah.
00:13:07And then-- yeah.
00:13:09And this is also the amazing thing that we test.
00:13:11So as you know, we have, like, hundreds of sponsors, right?
00:13:16Like 140 plus.
00:13:18And as you can see on the-- at the lobby,
00:13:21we have the banner with all the sponsors.
00:13:24And I basically tell Devin, like, hi, could you compare--
00:13:29could you check if there are any missing logos in this graphic?
00:13:33And the accuracy is 100% based on the test that I do.
00:13:38So-- which is pretty wild.
00:13:40And we use the same thing for the T-shirt that you got for your swag.
00:13:48And, yeah, surprisingly, Devin knows how to, like, visualize things, right?
00:13:54Like, how to detect things visually.
00:13:56And that is very surprising.
00:13:58Because as a human, we can, like, give errors.
00:14:02Oh, turns out there is one small service missing.
00:14:05But with this kind of thing, we can, like, double check.
00:14:09So human plus AI, combine it.
00:14:13Well, you've got your own QA team.
00:14:15And then remove fiction.
00:14:17So this is just the way of thinking.
00:14:21So as a designer, we have to think as a user, not as a designer.
00:14:26Right?
00:14:27Because every user has its needs.
00:14:29You can walk through the, for example, the map plan here.
00:14:33So basically, I'm imagining myself as an attendee to go to the registration,
00:14:39go to the--
00:14:41see the wayfinding and the QR code and then all the stuff.
00:14:46Basically, everything needs to be connected so you guys don't get lost
00:14:50and knows how to find your rooms and other stuff.
00:14:55And the real job is handling exceptions.
00:15:00So for example-- oh, I have-- yeah.
00:15:03For example, there is a schedule update, all right?
00:15:09And when we create this thing, it doesn't have an edit button.
00:15:13And then one morning, it just, hey, this schedule needs to be updated.
00:15:19And we don't have those edit buttons.
00:15:21I could just ask Devin, hey, can you add me an edit button?
00:15:25And then it did.
00:15:26So we can change everything now and then ship it to PNG and replug it to the screen,
00:15:32which is pretty convenient, right?
00:15:34And those exceptions, right, it-- it's not possible before when we have to do it manually and stuff.
00:15:41But now it's just get easier.
00:15:43And so the takeaway here is that to solve the scale problem, you have to actually think small.
00:15:51Think all the smallest thing possible.
00:15:53Think everything that can go wrong and will go wrong and then try to solve it before.
00:15:59And also, like, yeah, right now, basically, you can automate everything.
00:16:04And at this moment, having a problem is actually going to benefit you
00:16:11because that's going to help you ship a better product, going to ship things that are good.
00:16:18And, yeah, I think that's all that I can share.
00:16:20Hope my talk has some benefits to you.
00:16:23And, yeah, that's all.
00:16:26Thanks, guys.
00:16:29Thank you very much.

Key Takeaway

Integrating AI agents like Devin with foundational design systems allows a single designer to produce thousands of pixel-perfect event deliverables at scale.

Highlights

  • A single designer handles 7,000 attendees, 140-plus sponsors, 300-plus speakers, and 600-plus sessions by leveraging AI automation tools like Devin.

  • Devin automates pixel-perfect deliverables including speaker announcements, trading cards, signage, schedules, and missing logo QA checks.

  • Scaling design operations requires five core elements: foundation first, reusable designs, automated workflows, validated outputs, and removed frictions.

  • Atomic design principles break complex deliverables into foundational design tokens like typography, color, and spacing to prevent LLM styling drift.

  • AI verification achieves 100% accuracy when checking sponsor logo presence across large banners and t-shirt graphics.

Timeline

Scaling Challenges for Large Conferences

  • A team of 12 to 15 people manages an event scaling to 7,000 attendees.
  • One designer produces every event asset ranging from stickers and digital signage to landing pages and track mascots.
  • A thousand deliverables create a thousand potential failure points such as missing sponsor logos or incorrect schedules.
  • LLM image generation tools require strategic workarounds like exporting PNG files for vectorization in Figma.

Managing massive multi-day tech conferences with extremely lean teams requires recognizing scale as the primary operational challenge. Traditional manual design workflows break down under the weight of hundreds of speakers and sponsors. Overcoming this friction relies on treating AI not as a standalone solution, but as an active member of the design and engineering team.

Foundational Systems and Reusable Assets

  • Setting up a robust foundation involves strict design systems, color tokens, and explicit typography rules.
  • Atomic design methodology breaks interfaces down into the smallest possible components resembling Lego pieces.
  • Explicitly defining mobile and desktop typography prevents LLMs from outputting unstyled text.
  • Marketing teams generate consistent emails, flyers, and documents once website design tokens are established.

Establishing design foundations early prevents stylistic drift when AI models generate layout components. By defining precise color palettes, font hierarchies, and component atomic structures, non-designers within the organization can spin up brand-compliant assets. This reusability cuts down repetitive creation tasks and ensures visual consistency across hundreds of deliverables.

Automated Workflows and Pixel-Perfect Output

  • Devin automates schedule exports directly from updated data sources into ready-to-print PNG files.
  • Connecting Figma annotation plugins and MCP gives AI agents exact spacing, font size, and color specifications.
  • Automated generation tools handle over 300 speaker announcements in landscape, portrait, and trading card formats.
  • AI agents perform automated visual QA checks to detect missing sponsor logos on large event banners.

Automated workflows bridge the gap between design intent and final developer implementation. By integrating AI agents directly into Slack and connecting them with design specs, teams bypass slow feedback loops. Visual quality assurance checks executed by AI eliminate human error when verifying massive collections of sponsor logos and graphics.

Removing Friction and Handling Exceptions

  • Designers must adopt an attendee-centric perspective by walking through physical wayfinding and registration maps.
  • Real-world event operations require handling sudden exceptions like adding unprogrammed edit buttons to live schedules.
  • Solving the scale problem effectively begins by thinking small and anticipating potential failure points in advance.

True operational efficiency depends on identifying and removing user friction points across physical and digital event spaces. Handling last-minute exceptions without breaking production pipelines proves much faster when AI agents can modify code on demand. Ultimately, having specific operational problems to solve unlocks the full leverage of automated design workflows.

Community Posts

No posts yet. Be the first to write about this video!

Write about this video