When trust became the product requirement

A powerful AI system still has to earn scientific confidence

In 2026, Thermo Fisher Scientific was advancing an AI-enabled research environment designed for scientific work. By July, the product had reached a functional Alpha and moved into pilot validation with approximately 50 scientists, with plans to expand participation substantially ahead of a broader release.

The risk was not simply whether the underlying AI could produce useful answers. In scientific work, a system can be technically impressive and still fail if users cannot understand, challenge, and evaluate what it is telling them.

We rejected the default AI interface

The dominant pattern in AI product design was conversational: a prompt box, a fluent response, and an interface built to feel increasingly human.

G&CO. argued that scientific work required the opposite. Instead of asking scientists to trust an answer because it sounded authoritative, the experience needed to make rationale, comparison, validation, provenance, and confidence visible enough to interrogate.

In our efforts, we created an environment where AI-supported reasoning became more human, and more inspectable.

We separated immediate improvement from the AI-native future

G&CO. made two deliberate calls.

First, we kept near-term improvements to the existing digital experience separate from the new AI-native research environment. That allowed immediate commercial friction to be addressed without forcing the future product to inherit the assumptions of the current one.

Second, we chose not to make chat the primary interaction model. The product was treated as a scientific workspace, with AI supporting evaluation and decision-making rather than performing as an anthropomorphic assistant.

That distinction shaped the experience from the beginning: the interface had to help scientists assess what the system knew, how confident it was, and what required human judgment.

Build, test, then earn the right to scale

A functional Alpha was delivered in July 2026 and moved into pilot validation with approximately 50 scientists.

The experience was designed around clear workflow states and inspectable reasoning rather than AI spectacle.

Product decisions were pressure-tested throughout the engagement against the needs of scientists, scientific leaders, and the people responsible for bringing the platform to market.

The result was a working product progressing through real scientific validation toward a larger Beta and broader release.

The method began to outlive the prototype

The most important outcome was in G&CO. introducing a repeatable way to pressure-test product and leadership decisions before they reached scientists at scale.

‍

Frequently asked questions

What did G&CO. do for Thermo Fisher Scientific’s AI initiative?

G&CO. supported Thermo Fisher Scientific with product and experience design for an AI-enabled scientific research environment. The work focused on making AI-supported reasoning understandable and inspectable for scientists, with visible rationale, comparison, validation, and confidence rather than relying on a conventional chat-first interface.

How is enterprise software UX design different for scientific AI products?

Enterprise software UX design becomes more demanding when users need to evaluate consequential scientific information instead of simply completing a task. For Thermo Fisher Scientific, G&CO. designed around scientific confidence: users needed to understand the reasoning behind recommendations, assess uncertainty, and retain human judgment throughout the workflow.

Why didn’t G&CO. design the Thermo Fisher Scientific experience as an AI chatbot?

The familiar AI product pattern is a conversational prompt followed by a generated response. G&CO. deliberately designed against that pattern for scientific work. The experience was conceived as a workspace where scientists could inspect recommendations, rationale, comparisons, provenance, and confidence instead of being asked to trust the fluency of a chatbot.

What role did G&CO. play in Thermo Fisher Scientific’s digital transformation?

G&CO.'s role centered on product strategy and experience design: improving near-term digital experiences while helping define a separate AI-native environment for scientific work. Keeping those horizons distinct allowed current opportunities to move forward without constraining the longer-term product around the assumptions of the existing experience.

Was G&CO. responsible for Thermo Fisher Scientific’s enterprise AI architecture?

G&CO.'s work focused on the product and experience architecture surrounding the AI-enabled scientific environment. Our role was to determine how complex AI capabilities should become understandable, usable, and trustworthy within scientific workflows.

What should companies look for in an AI product design agency?

Enterprise AI products require more than a familiar interface wrapped around a model. An AI product design agency should be able to determine where AI belongs in the workflow, what users need to understand before acting on its output, how confidence and uncertainty are communicated, and where human judgment must remain explicit.

Thermo Fisher Scientific's work with G&CO. applied those principles to scientific research, where trust could not simply be assumed from the sophistication of the technology.

What makes G&CO.’s approach to B2B UX design different?

G&CO. approaches B2B UX design around the decisions users need to make, not simply the workflows software needs to accommodate. In complex enterprise products, that often means reducing cognitive burden while preserving the information, evidence, and controls expert users need to exercise judgment.

For Thermo Fisher Scientific, simplicity could not come at the expense of scientific rigor.