Kailen Nordhausen Design
Kailen Nordhausen Design

 

Work/ZEISS EagleCase study · 2023–2024

An AI guide for choosing a research microscope

Carl, an AI assistant for ZEISS Microscopy that recommends instruments and software, answers support questions and turns the configurator into a conversation. From vision film to beta, led at Goods & Services.

Carl · motion study
Client
ZEISS Microscopy, through Goods & Services
Role
Associate Design Director, Experience Design
Dates
Summer 2023 – 2024
Team
Goods & Services experience, visual and motion design, with ZEISS digital and product teams
Scope
Vision film, AI assistant, instrument configurator, personalized journeys, learning platform
01The problem

Choosing a research microscope means matching an instrument, its components and its software to one specific experiment. The ZEISS site asked scientists to do that matching themselves, browsing a large catalog by application and product family, before a sales rep ever got involved.

The brief was to make that first step feel like talking to the most knowledgeable person at ZEISS, and to do it in 2023, when generative AI was new, unreliable and under a microscope of its own.

  1. Scientific buyers. A researcher will spot a wrong recommendation immediately, and will not come back.
  2. A global brand with strict standards and a long, rep‑led sales process the experience had to support, not replace.
  3. Leadership needed to see the future before funding it.
  4. A catalog built for browsing, not for answering a question.
02Three decisions
  1. 1

    Show the vision before building it.

    The project opened with a film, ZEISS Beyond Labs, in summer 2023, so executives could react to a working future rather than a requirements list. The first work package was scoped that October and staffed by December. The film did the selling; the team did the building.

    Turned down: a discovery phase that ends in a deck. A deck asks people to imagine; a film lets them decide.

  2. 2

    Give the AI a name and a narrow job.

    Carl recommends products and software and answers support questions, and nothing else. A scientist types “I’m looking for a solution that will help me count dendrites more efficiently,” and Carl answers with the LSM 900 with Airyscan 2 and arivis Pro, with the reasons. A narrow job kept every answer checkable, and gave the brand a character to design: idle, listening, thinking, notifying.

    Turned down: a general chatbot. The wider the job, the sooner it says something a scientist knows is wrong.

  3. 3

    Make the configurator a conversation.

    Instead of a form of specifications, the instrument page asks about the experiment and builds the configuration from the answers, then hands a quote to a rep. Around it, personalized journeys, such as a brain‑tissue sampling page with a 60‑day test drive of the Primostar 3, and peer masterclass training give researchers a reason to come back between purchases.

    Turned down: an AI layer bolted onto the existing product pages. The question had to come before the catalog, not after it.

03The work
The homepage asks a question. “What would you like to know?” replaces the catalog as the front door. Ask Carl is the first thing on the page, not a widget in the corner.
A recommendation, start to finish. Example questions get a first‑time user moving; the answer names the instrument and the software and says why.
The configurator as a conversation. The LSM 900 product page, from highlights to Build & Quote with AI: Carl asks about the experiment, assembles the configuration and hands it to a rep.
Carl has states, not a face. Idle, listening, thinking and notification, built as light in a glass cube so the assistant reads as ZEISS optics rather than a cartoon.
Personalized journeys. A brain‑tissue sampling page with a 60‑day Primostar 3 test drive: one page per research problem, not per product.
Peer masterclass training. Learn from the world’s best: training led by researchers, so the site is worth visiting between purchases.
04How it ran

Three views of the same work, kept live for the whole engagement.

Experience design boards for every screen and state, a service map of who buys a microscope and how a decision moves from budget owner to IT to the bench, and the engineering board the build ran on. The AI was the headline; the service design is what made it fit a rep‑led sales process.

Experience design. Page templates, components and annotations for the ZEISS web team.
Service design. Stakeholders from budget owner to quality control, and the journey from awareness to the ongoing relationship, with sales and marketing operations mapped underneath.
Engineering. The live board the ZEISS web platform ran on, so design decisions landed as tickets, not as a handoff.
05Where it stands
  • DeliveredVision film, ZEISS Beyond Labs, summer 2023
  • DeliveredWork package 1: Carl, the conversational configurator, personalized journeys and masterclass, scoped October 2023 and built into 2024
  • DeliveredBeta, with customer and user feedback sessions in 2024
06My role

I led experience design and design strategy from the vision film through the beta: the concept and story of Carl, the conversational configurator and the journeys, directing the UX, visual and motion work and presenting it to ZEISS leadership. Goods & Services strategy, motion and engineering partners, and the ZEISS digital team, carried it to the build.