So we decided to let AI build our website...Here’s how it went

At a glance Every company will need to refresh their website and tweak their brand positioning from time to time. When the time came to revamp ours, we wanted to see how much of the work we could do with machine learning; after all, we spend a lot of time using AI tools on behalf of our clients.

Using AI to Build our Website

Every company will need to refresh their website and tweak their brand positioning from time to time. When the time came to revamp ours, we wanted to see how much of the work we could do with machine learning; after all, we spend a lot of time using AI tools on behalf of our clients, helping them to understand their audiences – it’s a sandbox we’re used to playing in. There was more to it than that, though … we were curious. How much of the work you’d typically entrust to creative agencies and freelance specialists could be conducted using AI tools? Just how far has technology come? The short answer is pretty far, though it’s still a long way away from replacing humans. We used eight different AI tools for different tasks, including generating text and images. Each had its strengths, each brought something to the party. The problem was that none of them talked to each other. We were left with disparate elements with no tech solution to connect the dots … for that, we still very much needed the learned skills of actual people. That said, the technology was still able to take us pretty far. 95% of our draft text was written using help from GTP-3, which we used as an AI copywriter, with the results tweaked by a professional to get the copy over the line. The majority of our images were created using DALL-E2 and then edited and enhanced by our in-house graphic designer. All of this was directed using key insights from our own Audience Intelligence to create the messaging and brand positioning, psychometrically profiling our audience to understand who they are and what they need. It was this last element that proved most useful. Social data and AI helped us make some key breakthroughs in understanding our target audience, and those, in turn, helped us to understand the directions we needed to go. For example, we learned that there was real platform fatigue around marketers. A big chunk of the people we are trying to reach are overwhelmed and overworked and the last thing they want is to have to learn a new tech platform or tool. It pointed us towards a more human, helpful tone, where technology combined with human help answer tough questions and provide context rather than creating more work. While that was likely the single most important insight in our AI experiment, the most impressive was watching DALL-E2 create images based on our prompts. When it gets it right, it’s the closest thing to magic I’ve seen! There’s a “photograph” of a woman on our homepage banner – you can stare into her eyes and imagine what she’s thinking or feeling … until you remember that she has never existed and has been conjured entirely by an AI. When you know that, it becomes just a tiny bit creepy. As good as the tech is, a human is still very much needed to write creative prompts for the machine, and that’s an art in itself, and after some trial and error, the results were genuinely impressive. Meanwhile, our graphic designer was kept busy fixing the often-ghostly faces and six-fingered characters spun out of the ether by the AI. For now, human expertise – what is often called “Human in the Loop” or “HITL” – is still necessary to direct the digital tools, though watching the process unfold, you can’t help but ponder what impact this impressive technology is going to have on creative industries as the AIs get more sophisticated. We even looked into using Open AI’s Codex to help us with writing the code, though given the relatively simple nature of this project, at least, in this case, it was easier to use good old-fashioned coding. It’s something we’re starting to use inside the Buzz Radar platform, however. It’s very possible our next revamp could be AI-generated from the inside out. AI-based apps can provide a fantastically smart toolkit that can shortcut a lot of human effort and uncover insights that are foundational to your strategy. What a machine can’t do, at least not on its own, is pull all those tools together to build a website and reposition a brand to communicate effectively to an audience. Not yet, anyway, and not easily. Building the site would have still been impossible without a talented team of strategists, creatives and coders pulling it all together. We’re still a long way from AI being a complete solution. Which, at least for the world we’re working in right now, is maybe for the best.Oh, and if you enjoyed this blog post? You have an AI to thank for that, too – the idea for writing this in the first place and the way it’s structured were all suggested by the machine after analysing our audience. You’re welcome. How effective were our AI-led efforts? Judge for yourself by checking the new Buzz Radar website here! #artificialintelligence #socialintelligence #socialmediamarketing

Patrick Charlton Published on September 22, 2022 9:01 am

Frequently Asked QuestionsFAQs

Can AI build a complete website without human help?

No, AI cannot currently build a complete website independently. While AI tools can generate about 95% of draft text and create images, human expertise is still essential to connect all the elements together and ensure quality results.

What AI tools are best for website content creation?

GPT-3 is effective for generating website copy, while DALL-E2 can create impressive images from text prompts. However, both require human oversight - professional copywriters to refine the text and graphic designers to fix issues like distorted faces in AI-generated images.

How much of website development can AI actually do?

AI can handle a significant portion of content creation, including most draft text and image generation. The main limitation is that different AI tools don't communicate with each other, so humans are still needed to integrate everything and provide strategic direction.

What are the biggest problems with using AI for web design?

The main issues include AI tools working in isolation without integration, and quality problems like ghostly faces or six-fingered characters in generated images. AI also requires skilled humans to write effective prompts and provide strategic oversight throughout the process.