What We Learned From Our First 200 AI-Enhanced Design Projects

The X Factor: Bridging the Gap Between Creativity and Data

On the surface, designing with AI is simple. You enter a prompt, you get an image.

But to get a creative, high-quality, on-brand image that not only fulfills a brief, but elevates designs and drives performance? Not simple.

At Superside, we have a thesis: To achieve this level of design with AI, we need to augment data sets with targeted information about a brand. Though insanely powerful, foundational AI models are innately biased towards the mega brands they were trained on. So, unless you’re designing for Coca-Cola, Nike or Disney, you need to feed the model more information to be able to generate assets that meet your specifications.

This is why we’re doing things differently by starting with data across our workflows. Depending on a customer’s needs, we might apply that data at the prompt engineering or model training level—but our proprietary brand database ensures our creatives have the information they need at every level to best take advantage of AI.

Essentially, we’re building an AI design system that bridges the gap between creativity and data, enabling unprecedented speed and scale. Here’s what it entails:

• Brand database. Imagine all the creative data points (style guidelines, project specs, pre-existing work, final key art etc.) needed to deliver on-brand, on-brief designs, in one place. Now, imagine that data only gets better with every project as the knowledge from successful designs is aggregated, creating a positive feedback loop between human and machine. This is the secret sauce behind our ability to use AI at scale, while maintaining creative quality and consistency.
  
  
• Prompt engineering. Tapping into our brand database, our creatives can get a coherent picture of a brand’s content and move faster by generating prompts pre-tailored to their style, giving them a leg up in tools like Midjourney, Stable Diffusion and GPT-4. By using this data to engineer prompts, it’s possible to deliver on projects with less stringent specifications. For those with a very specific style, we take things to the model level…
  
  
• Model training. If a customer has a specific style and needs high volumes of creative, injecting information into prompts may not cut it. In these instances, we use low-rank adaptation of large language models (LoRA) training in Stable Diffusion to fine-tune a base model to specific brand needs. This extra bit of guidance is like the cherry on top of a foundational model—but it’s this decoration that takes the cake from bland to on-brand.
  
  
• Creative expertise. Before the bots can understand requests, a human needs to think through tenuous concepts like brand. After all, brand is both there and not there. Sometimes, it’s the inclusion of things. Other times, the absence. Sometimes you can’t quite put your finger on what makes it special. That’s why creative eyes, expertise and ingenuity are so essential to guiding our AI-powered processes.

The result? AI amplifies our creative talent and unlocks exponential value for our customers.

Through this combination of unique data sets, prompt whispering and fine-tuning, our creatives have been able to drastically speed up workflows, hitting the milestone of 200 projects completed with AI in three and a half months. In addition to refining our processes and offering our customers more and more value, we’ve learned a thing or two from these projects, including just how much time AI can save our customers and where it works best today.

The Efficiency Factor: Analyzing 200 Projects Enhanced by AI

As a subscription-based business where our customers get access to a certain amount of hours a month, we track everything. From years of data and experience, we can provide precise estimates of how long projects will take, whether you’re in need of a digital ad campaign, presentation or fresh brand assets.

With the beta launch of our AI-enhanced creative services, this historical data has served us well. For one, it’s allowed us to see exactly how much more efficient we are with AI, and how that translates into time and cost savings for our customers. Month over month, we’re analyzing:

• How long a project took with AI
• How long it would’ve taken without AI
• What the difference is
• How that translates into value for our customers

Looking at our numbers for Q3 and Q4 so far, we can see traditional estimates for how long projects would take far outstrip the actual hours logged by the AI-enhanced creative team, which translates into time savings of 39.55%—making us more efficient than we dared hope when we started this program.

How does this translate into value for our customers? It’s simple: The hours they save by choosing our AI-enhanced creative services go right back in their pocket. They can use those hours towards other design projects, and get more high-performing creative assets out of their Superside subscription every single month.

To put an even more tangible figure on it, looking at the last 200 projects completed with AI, we estimate we’ve saved our customers 3,744 hours, which equals $299,525 worth of design time.

Of course, not all projects are made equal. As we’ve built our beta program, we’ve gotten a sense for which use cases are well suited to AI enhancement, and where gains in efficiency are more moderate. In general, we’ve found AI performs best for accelerating mass volume creative projects, like generating multiple ad sets, tailored image libraries or custom illustrations.

These findings coincide with the types of projects where we’ve seen the most growth. Leading the pack, digital ads, static web and social graphics account for 47.3% of our first 200 AI projects, while illustration design (18%) is our next most significant segment.

Looking at some examples of these types of projects, we can see AI has helped our teams reduce design time by as much as 90%.

Examples of Superside’s AI-Enhanced Design Projects

D2L Brightspace’s Brilliant Ad Campaign

D2L Brightspace, a leader in educational technology, wanted to breathe new energy into their brand and bid farewell to the drab, predictable ads of their past. By using Midjourney image generation to replace traditional stock imagery, the team was able to develop a unique, head-turning style, while moving fast and maintaining consistency.

The results: 70% design time saved, 114 ad variations, 100% brand consistency

IPG’s Pawsitively Awesome Illustrations

Independence Pet Group (IPG), one of North America’s premier pet insurance companies, was looking for fun, whimsical illustrations to enrich their internal merchandise collection. To meet this brief, the team used Midjourney to quickly explore illustration styles and come up with the perfect fit for IPG, delivering delightful assets in a mere 11.5 hours.

The results: 90% design time saved, $15-20K design costs saved, 750+ images generated

Avalanche’s Fresh X (Twitter) Templates

Avalanche (also known as Ava Labs), a smart contracts platform in the blockchain industry, needed X (Twitter) space banner templates to liven up their socials. Using a combination of Midjourney and Figma, the team developed a suite of versatile visuals, suitable for various speaker compositions as well as dark and light themes.

The results: 37.5% design time saved, $1-1.3K design costs saved, 3,300+ images generated

Though we’ve been focusing on time and cost savings, it’s worth noting we’ve found many other benefits to using AI in our creative workflows. As a replacement for stock photography, image generation guarantees originality and enables brand consistency. In instances where a customer has a very specific style, it might be more time-consuming to train an AI model and engineer the right prompts—but it’s an investment that will pay off in the long run.

Take Amsive, for example. Given their specific brand aesthetic, we didn’t save on time or costs by using AI to generate graphics for their last social media campaign. But we did add value by providing a wealth of unique images to choose from. And next time Amsive comes to us with a campaign, we’ll be able to leverage all the data we amassed from this AI-enhanced project to bring the next one to life faster.

The bottom line? I repeat, AI is an amplifier. Of talent, creativity, and as we’re discovering alongside our customers, next level value.

The Value Factor: Getting More With AI-Enhanced Design

With over 200 AI projects under our belt, I feel confident saying, this is the way.

At Superside, we’re just starting to develop our AI-enhanced creative services, but even in this beta learning phase, the results are beyond promising. Not only have we reduced billed hours by 39.55% compared to our traditional model, our customers have been thrilled with the creative, giving AI projects an average CSAT rating of 100%.

So, this increased efficiency doesn’t mean we’re compromising on quality. On the contrary, AI has infused our projects with even more creativity, allowing our designers to engage in deeper exploration and opening doors that simply didn’t exist before.

And we’ll continue to knock on those doors…

On top of improving our existing workflows, we’re beginning to explore new creative use cases, unpacking the promise of AI for concepting, branding and motion. At the same time, our core focus remains unchanged—finding ways to offer our customers even more value with creative that performs.

Join Us in a New Era of Creativity at Scale

Superside has always promised to deliver high-performing creative faster, at reduced cost and insane scale.

With our AI-enhanced creative services, we’re taking that promise and multiplying it ten-fold.

And we’re not just dipping our toe in, we’re doing it all. From model training to craft truly custom solutions to data structuring designed to tap into Superside’s collective creative genius, we’re developing an AI-powered system that’s already providing unparalleled value to our customers.

But of course, we’re still in the early days of AI and our program...

In 2024, we plan to scale our use of AI by a whole new order of magnitude, aiming for 2,500 design projects completed per month. Following our innovation roadmap, we plan to effectively 25x our output of AI-enhanced projects by the end of the year.

With the rate of change in creative AI, I’m sure we’ll have a few more learnings to share before we get there.

Explore Superside's AI-Enhanced Creative Services

See how we can deliver creative that performs for your brand at faster speeds, greater value and larger scale.

Explore now