Pittcon 2026 Conference Report: Generative AI Enters the Analytical Chemist’s Toolbox

Through four complementary presentations, the James L. Waters Symposium moved from visionary perspectives on AI-driven discovery to practical considerations of validation, implementation, and scientific integrity.

AIMATRY—A New Field of Chemistry

The opening presentation, “AIMATRY: A New Field of Chemistry,” was delivered by Omar Yaghi of the University of California, Berkeley, who was awarded the 2025 Nobel Prize in Chemistry.

Yaghi introduced the concept of AI MAterials chemisTRY (AIMATRY), a proposed new discipline in which AI serves as an integral partner in materials discovery. The talk focused on how large language models and adaptive machine-learning algorithms can dramatically accelerate the discovery and synthesis of advanced materials.

A central example involved the crystallization and discovery of metal-organic frameworks (MOFs) and covalent organic frameworks (COFs). Traditionally, identifying new crystalline materials requires extensive trial-and-error experimentation. Yaghi demonstrated how AI-guided experimental design can analyze patterns in prior crystallization experiments and suggest promising synthetic pathways. According to his results, these adaptive algorithms can potentially double the rate of materials discovery.

Beyond the specific examples, Yaghi emphasized the broader implications of AI-assisted laboratory practice. Generative models can simplify complex experimental planning, assist with hypothesis generation, and help chemists navigate large multidimensional chemical design spaces. The AIMATRY concept therefore positions AI not simply as a computational tool but as a partner in creative scientific exploration.

Accelerating Innovation in Analytical Chemistry With Generative AI

The second presentation, “Accelerating Innovation in Analytical Chemistry and Measurement Science with Generative AI,” was given by M. Farooq Wahab of The University of Texas at Arlington.

Wahab began by examining the intellectual foundations of analytical chemistry, emphasizing that the discipline rests not only on chemistry but also on applied physics and scientific computation. For many analytical chemists, however, the complexity of advanced signal-processing and chemometric algorithms has historically created barriers to implementation.

Generative AI, he argued, has the potential to function as a Ph.D.-level assistant capable of explaining, generating, and adapting algorithms for chemical analysis tasks. Wahab presented several practical examples illustrating how AI can assist with common analytical challenges, including the following:

• Advanced feature detection in spectral or chromatographic data
• Information-theoretic denoising methods
• Resolution enhancement through regularized deconvolution
• Estimation of instrument transfer functions
• Rapid synthesis of literature and algorithm explanations

A particularly interesting aspect of the talk addressed how AI systems can translate complex code into conceptual explanations that are easier for analytical chemists to understand and apply.

Wahab also explored less conventional uses of generative AI, including diagnosing issues in scientific publications and translating non-English analytical literature. These capabilities, he suggested, could broaden global access to chemical knowledge.

Despite the promise of these tools, Wahab cautioned against uncritical adoption. He compared unverified AI results to the famous case of Clever Hans, the horse once believed to perform arithmetic but later shown to be responding to subtle human cues. Without careful benchmarking and validation, AI systems may generate convincing—but incorrect—answers.

Beyond the Hype—Lessons From Chemometrics

The third talk, “Beyond the Hype: What Chemometrics Can Teach Generative AI,” was presented by Rasmus Bro of the University of Copenhagen.

Bro provided a valuable historical perspective by comparing today’s surge in AI enthusiasm with the long experience of the chemometrics community. For decades, chemometricians have confronted the very issues now emerging in AI applications: data quality, model overfitting, interpretability, and validation.

Drawing on examples from multivariate analysis and calibration modeling, Bro argued that many of the methodological safeguards developed in chemometrics remain essential for AI-driven analytics. Without careful validation protocols, AI models can easily produce results that appear plausible but lack scientific reliability.

A key theme of the talk was that AI should assist—not replace—human reasoning in chemical analysis. Rather than serving as an automated decision maker, AI should function as a tool that augments expert insight. Bro emphasized the importance of maintaining transparency and scientific rigor when integrating generative AI into analytical workflows.

His message resonated strongly with the audience: While AI may introduce powerful new capabilities, the foundational principles of chemometrics—careful experimental design, validation, and interpretability—remain indispensable.

From Calibration to Interpretation

The final presentation, “From Calibration to Interpretation: How Generative AI Is Rewriting Chemical Measurement,” was delivered by Jerome Workman of LCGC–Spectroscopy.

Workman framed his talk around the increasing complexity of analytical chemistry. Modern practitioners must integrate knowledge from chemistry, physics, mathematics, and, increasingly, computer science. As chemometric models become more sophisticated, the mathematical barrier to adoption has grown correspondingly.

Generative AI, he suggested, may help bridge this gap by providing task-specific analytical assistance for chemometricians for a variety of spectroscopic data processing tasks. Workman described practical approaches for embedding advanced generative AI tools into laboratory workflows through carefully curated and structured applications within data-analysis pipelines.

A central concept presented was how current chemometric approaches developed over decades provide a solid foundation for current AI implementation. In this framework, AI is used to generate algorithms or analysis approaches, which are then rigorously validated against benchmark datasets before being deployed in routine analysis. This methodology transforms generative AI from a black-box curiosity into a dependable collaborator supervised by expert knowledge.

Workman also discussed strategies for overcoming common implementation barriers, including the following:

• Ensuring high-quality and well-curated data sets
• Selecting appropriate algorithms for specific analytical tasks
• Integrating AI tools with existing laboratory information systems
• Distinguishing scientifically useful AI applications from hype

The talk concluded with examples of how generative AI can assist with calibration development, spectral interpretation, and algorithm design in spectroscopic and chromatographic analysis. He cautioned AI users to rely upon their own knowledge of chemistry to implement these new tools, as he compared several traditional chemometric approaches to generative AI approaches.

From Curiosity to Analytical Colleague

Taken together, the presentations in the Waters Symposium illustrated both the promise and the challenges of generative AI in analytical chemistry. The session moved from visionary concepts, such as AI-guided materials discovery, to pragmatic considerations such as algorithm validation and workflow integration.

A recurring theme across all four talks was that AI should augment human expertise rather than replace it. When paired with rigorous validation, curated data sets, and strong chemometric principles, generative AI may significantly accelerate innovation in chemical measurement science.

The symposium ultimately framed generative AI not as a disruptive black box but as a new analytical colleague, one that must be guided by the same scientific discipline that has long defined analytical chemistry. As AI tools continue to mature, sessions like this suggest that the analytical community is beginning to define how these technologies can be responsibly integrated into everyday laboratory practice.