2023field-io · commercial-client · creative-development

IBM watsonx

A conference experience built at FIELD.IO that turns IBM's watsonx AI platform into an explorable, assemble-your-own AI stack, making enterprise AI adoption feel legible rather than intimidating.

The watsonx installation's hero render.
01

Overview

IBM watsonx is commercial client work built atFIELD.IOforIBMin 2023, shown at major IBM conferences. Daan Rongen was credited forCreative Development, alongside Guido Schmidt and Andy Duboc. Full credits: Executive Creative Direction — Marcus Wendt; Creative Lead — Jack Grafton; CGI Creative Director — Antar Walker; Technical Direction — Jonas Otto; Creative AI Engineer — Casper Wortmann; Creative Development — Guido Schmidt, Daan Rongen, Andy Duboc; Executive Producer — Sinéad McCarthy; Producer — Kerem Demirayak; Lead Digital Artist — Julien Bauzin; Experiential Designer — Max Palmer; 3D Design — Alexander Hahn, Lukas Rafik Mayer; Creative Technologist — João Fonseca; Motion Designer — Marcel Piekarski; Developer — Owen Hindley, Toby Harris.

02

Concept & approach

The brief was to make watsonx, IBM's enterprise AI platform, legible to a conference-floor audience who might otherwise find "AI adoption" abstract and intimidating. The experience breaks the decision down into four business concerns visitors can actually reason about:capability,cost,custom solutionsandgovernance. Rather than a static pitch, visitors explore, assemble and analyse a potential AI stack tailored to their own organisation, watching the stack's shape and implications change as they swap components in and out.

Screen mockup of the watsonx stack-assembly interface.
Screen mockup of the watsonx stack-assembly interface.

Once a visitor picks a use case, the tool opens onto a stack view where every dataset, model and prompt they've added sits as its own composed form.

A customer-service AI stack composed within the tool.
A customer-service AI stack composed within the tool.
03

Visual system

The four-concern framework is rendered through a generative 3D visual language: abstract model, dataset and prompt structures rendered as physical-feeling forms, so an abstract AI stack reads as something closer to inspectable machinery than a diagram. A model reads as a dense, layered flower of ribbons:

Generative render of watsonx AI models.
Generative render of watsonx AI models.

A dataset unwinds into a single continuous coiled band:

Generative render representing a watsonx dataset.
Generative render representing a watsonx dataset.

And a prompt stacks as a loose, off-axis column of rings:

Generative render representing a watsonx prompt.
Generative render representing a watsonx prompt.
04

Reflection