AI in 2026: Cutting through the confusion

If you find yourself bewildered by the proliferation of AI models available, you are in good company. We are in the midst of ‘Cambrian Explosion’ of programs, all of which claim to do a number of tasks, or a subset of different tasks, better than their competitors.

 Take ChatGPT for instance. As of March 2026, the program was available in five different versions. And this after retiring six other versions in February. Google has four core Gemini 3 models, with four additional specialized models, including Nano Banana 2, its image generator, plus four ‘work’ or developer models. Additionally, Google AI is integrated into its own apps, like Gmail. And as big as ChatGPT and Gemini are, they are just a small corner in the AI universe.

How do you find your way through such maze of options? First, understand the types of models or programs out there. Eric Stoddard, of Car Design Academy, has distilled the models down to their essence and has identified four different types of tools: 

1. Large Language Model text generation: ChatGPT, Gemini, Claude, CoPilot
2. Text to Image Models: Midjourney, Nano Banana, Dall-E, Stable Diffusion
3. Sketch to Image Models: Vizcom, Krea (iPad version)
4. Sketch to 3D Models: Vizcom

Knowing which tool best serves the task is important to develop a consistent workflow, which Stoddard teaches at his Car Design Academy. It is helpful to know what the core strengths of each tool and how that can fit into your workflow. Just as you curate the output of these AI programs, you must curate the programs themselves. Also, all these models rely on prompts, and creating a powerful description to communicate your ideas is a skill that will be required in the future. You must learn to curate the written language to achieve the design you want.

Designers we spoke with (some of whom prefer to remain anonymous), all pointed to Vizcom as the best all-around tool for automotive designers. Bart van Lotringen, design director at DAF, told us,” I am very much in favor of Vizcom, of which we already have quite a few licenses right now. I am positively surprised with how well Vizcom listens to designers and is able to implement feedback and features quickly, that we and other users advise to them. That advancement is never too rapid, as long as the interface stays logical and intuitive.”

And Eric Stoddard noted, “Vizcom is emerging as the clear winner in the automotive/industrial design field. It’s a designer-friendly collection of tools, addressing all of the major workflows within the design process. I like to think of the Vizcom team as the ones nerding out on harnessing the best technologies so we designers don’t have to. They also have the best legal framework in the industry for intellectual property ownership and data security, along with broad support from the automotive OEMs.”

If you could even take animated 3D-based storyboards into a VR-setting that would be a real game-changer

Another model that was mentioned by our respondents is Midjourney, which is a highly sophisticated text to image model. Midjourney is great for ideation and can render its output in a myriad of styles, from realistic to impressionistic, from anime to pencil sketch. And unlike many programs that improve on the fly and hope for the best, Midjourney is deliberative and methodical in their development and upgrades. Like Vizcom, Midjourney’s creators listen to their users and try to act on every reasonable request – and a few unreasonable ones as well. Midjourney has its limits, and though it can render in 2D, as well as in video clips, it is not a 3D modeling tool. It’s more of an illustration tool, but a very good one, and one that shows enormous potential for the future.

 And what of the Future? Almost all the designers we spoke with hope greater integration with other digital tools would be an essential development of AI models. Common interfaces would be very helpful as well. Some AI models are being considered for bespoke hardware devices (looking at you Midjourney), and others are considering hardware as well. Already some AI models have been downloaded and operate independently on Mac Minis, and one was even loaded successfully onto a Macbook Pro. The implications are enormous. Imagine each design studio operating their own bespoke AI model on their own secure hardware. Who will need a Data Center?

Bart van Lotringen gives us a glimpse of his wish list: “What I am eagerly waiting for is that Autodesk will integrate AI-technology into Alias. Much of surfacing work is repetitive, labor intensive and based on experience, and with help of AI that could be made so much easier. So I hope Autodesk is reading this!

 “Also, what I am still looking for in integrated AI tools is to make the step from 2D to 3D to VR/AR and maintain the texture of your own sketch, combined with a good quality 3D-meshing. When that would become available you could take a sketch into 3D, put your VR headset on and review the design with your design team. And if you could even take animated 3D-based storyboards into a VR-setting that would be a real game-changer! It would allow you to travel in time and experience the future.”