The hype around artificial intelligence (AI) has been unmistakable (almost deafening) for months. Aside from the huge deals being made by players like Microsoft and the big investments by other MAMAA companies (Meta, Apple, Microsoft, Amazon and Alphabet), for entrepreneurs there is suddenly the promise of greater efficiency and faster answers to difficult or previously unsolvable problems. At the same time, headlines are emerging that artificial intelligence will steal jobs and make humans redundant, especially in the creative world. If computers can create art or solve complex problems in seconds, where does that leave humans?
For web and creative teams, AI is primarily a tool: it can accelerate research, drafts and technical routines, but it does not replace professional judgment or design accountability.
Code, drafts and the repetitive parts of a project
1. AI as a starting point for code
As developers creating websites and applications, we are often faced with tricky programming problems. We believe that artificial intelligence can provide a starting point for brainstorming a solution to a programming task.
In the example above, one of our developers wanted to learn about a new tool we were starting to work with. Instead of searching through the documentation to find out how it could be used for our specific use case, we could simply ask ChatGPT, a chatbot with artificial intelligence, to give us an example.
ChatGPT is also able to track the progress of the conversation and make adjustments based on our input without having to start from scratch. In this example, the AI provides our developer with 1:1 coaching on how to use a new tool that is tailored to our exact needs. ChatGPT not only gives us the code we need, but also explains it, which makes it a great learning tool. When our developers are able to quickly understand complex code issues, they can save hours - not only for themselves, but also for our customers - and gain valuable time for more innovative activities.
Artificial intelligence can help developers quickly free themselves from known and tedious code so they can focus on more complex and unique problems. The less time we have to spend recreating already established patterns, the more time we have to develop unique, interactive and user-friendly extensions.
Images, sound and video: where generated drafts help
2. Using AI to explore visual and audio elements
Artificial intelligence can also do amazing things with images, sound and video. When it comes to web design, one of the most important requirements for a unique and well-designed website is good visuals. It is important that the imagery and video is of high quality, that it is unique to the brand and that it is not too reminiscent of a stock photo. This means that a company must either have an in-house photography team or invest time and money to find, buy and edit stock images so that they are unique to the brand. Artificial intelligence has the potential to simplify this process by creating unique and royalty-free assets based on a prompt. A.I. will never replace a custom photo shoot under artistic direction or assets created from scratch, but it's a great option for quick blocking and tackling or fast concepting.
A consistent visual system can be defined for editorial illustrations and explored with AI-generated drafts. Final selection, editing and approval remain human responsibilities.
Could an artificial intelligence do this for us? We tried to train an image-generating AI, Stable Diffusion, on what this style looks like by feeding it some examples and then asking it to create its own style.
The style isn't perfect, but with a little tweaking it comes pretty close. We edited one of the line drawings we received so that it had a high resolution and the right background color and used it for the cover image of the article.
We also tried to develop some unique elements with DALL-E 2 to convey a concept in a presentation - someone talking loudly on their cell phone during a boat ride. Building on this, the cognitive image generation needs some refinement to get out of the uncanny valley of not-quite-right images. The results here are a little comical, but fine-tuning the prompts can lead to some amazing images.
Companies such as Adobe Stock rely on generative AI content and allow participants to submit content and get paid for it as long as they own the rights to it. However, the question of who exactly should own the rights to generative AI art is still the subject of controversy.
How AI supports user research
3. Using AI in quantitative user research
When we work with quantitative user research data or data that is number-based and contains large amounts of information, we often have to open a spreadsheet program and start «number crunching» to analyze our results. Artificial intelligence has helped our research team to remember formulas and apply them in context.
Again, ChatGPT is a great learning tool to determine the correct function and give instructions on how to use it in context.
A language model can help recall a formula or suggest a starting point, but its answer still needs to be checked against the relevant documentation and the underlying data.
4. Using AI in qualitative user research
Qualitative user research is often seen as more «human» and empathy-focused, so using AI for this method may seem strange at first. However, as with our other techniques, we try to use AI to speed up our process and expand our creative thinking.
Part of preparing for qualitative user research, such as interviews, is creating a list of questions to guide the conversation. Could AI help us with this?
This is not the exact list of questions we would go straight into an interview with. But asking ChatGPT to ask us questions about different scenarios can help us brainstorm what questions we could add. Maybe you haven't thought of asking questions about privacy and security in this context. Thank you AI!
What if AI could teach us empathy? It seems far-fetched, but ChatGPT has been trained on a huge amount of data that probably contains written records of people's wants, needs, goals and frustrations from all walks of life. Could artificial intelligence help us to put ourselves in their shoes?
Data like this can help us get into the mindset of our target audience before we interview them, or it can help us write helpful website copy later. Of course, we should always check that the information we get from AI about a target audience is accurate, and this is no substitute for human-to-human interviews, but we believe it can be a really helpful start.
What does it look like when the interviews are completed and the analysis begins? We believe that AI can really speed things up here. After we receive a large amount of interview transcripts, we have to sift through the data and search for themes. This is very time-consuming. Artificial intelligence could search through these transcripts in seconds and show us useful patterns and themes. Again, it's not a replacement for the human mind, but it could act like a second pair of eyes and identify areas that we may have missed.
A short, de-identified interview excerpt can be used to test summarization, provided the research participant has consented and the chosen service is suitable for the data involved.
The resulting summary can support an initial review, but it should be compared with the transcript before themes or conclusions are accepted.
Services like Notably.AI, which claim to be able to analyze user research faster with the help of artificial intelligence, are emerging to take advantage of this opportunity.
Testing ideas and building small internal prototypes
5. Using AI to generate and test ideas
ChatGPT is really good at taking all kinds of information and providing a concise summary, explanation or processing. There are endless applications for all kinds of business, such as summarizing a long email exchange.
Many AI services provide application programming interfaces (APIs). These can be used to prototype narrowly defined functions such as summarizing long email threads, provided privacy and quality controls are in place.
Gmail itself is likely to introduce these functions soon. The development of AI products is progressing at a rapid pace and everyone is trying to jump on the bandwagon.
Small internal experiments can also help a team understand where generated output is useful, where it fails and which review steps are necessary.
Will the AI trend continue?
An explosion of AI-based applications and debates
We predict that the artificial intelligence trend will not go away anytime soon, as it has already proven to be extremely useful in its early days of widespread availability. More and more applications will integrate AI as part of their offering and we can certainly expect a lot more excitement and discussion.
AI products, plans and usage limits change frequently. Before building a workflow around a service, review its current terms, privacy controls, cost model and options for exporting or deleting data.
The ethics of artificial intelligence will certainly remain an important topic of discussion. It is our duty as early adopters of this technology to reflect on the impact it will have on the world, to be alert to the possibilities of its misuse and to use our human capacity for critical thinking to assess whether the technology is providing accurate and useful information.
Bonus use case: Artificial intelligence for the closing paragraph
It's only fair to end an article about all the great applications and shortcuts we can get through artificial intelligence with artificial intelligence wrapping it up for us. Goodbye, ChatGPT.
AI can shorten selected research, drafting and development tasks, but useful results depend on a clear brief, suitable data and human review. For a web agency, the practical question is not whether to automate everything, but where automation improves the process without weakening originality, accuracy or accountability.
