Integrant

SLAS 2026 Short Course | SOLD OUT

From Modular Design to AI Agents

How Chemistry and Lab Automation Teams Identify, Design, and Prototype AI Assistance

This short course explored how scientific teams design, validate, and deploy AI assistants and agent-based systems in real environments.

The session focused on how AI systems can be built and adopted across science, automation, and IT without sacrificing rigor, control, or trust.

Scroll to explore the ideas, or continue the conversation with our team.

From Modular Design to AI Agents

What We Explored in the Session

The course centered on practical questions teams are already asking:

Where can AI assistants meaningfully support scientific work?

What's the difference between copilots, assistants, and agents — and when does each make sense?

Why do so many pilots stall, and what helps teams move forward?

How do modular systems enable AI adoption over time?

What roles do science, automation, and IT each play?

Rather than promoting a single tool or approach, the session provided a framework teams can adapt to their own workflows and constraints.

What It Takes to Make AI Work in Real Systems

Beyond concepts, the course addressed the technical foundations behind production-ready AI assistance, including:

How LLMs behave, what they're good at, and where they fall short

How skills, connectors, and integrations give assistants access to real tools and data

Why orchestration, validation, and monitoring matter for agents

Where configuration ends and custom engineering begins

How security, governance, and ownership shape adoption

This helped attendees understand why many demos don't translate into durable systems — and what's required to bridge that gap.

A Practical Way to Identify AI Opportunities

We introduced a repeatable approach teams can use to surface and design AI-assisted capabilities:

1

Map real workflows at a high level

2

Identify cognitive steps like research, reasoning, or decision-making

3

Capture subject-matter logic and edge cases

4

Apply the right AI patterns for each task

5

Validate value early to guide prioritization

This approach helps teams move from ideas to systems that fit their environment, rather than forcing AI where it doesn't belong.

Who This Is Most Useful For

The session was designed for teams working across scientific and technical domains, including:

Scientific leaders responsible for research workflows and decisions

Scientific leaders responsible for research workflows and decisions

Automation and technology enablement teams

Automation and technology enablement teams

IT and data teams supporting integration, security, and governance

IT and data teams supporting integration, security, and governance

Platform and product owners exploring AI-assisted capabilities

Platform and product owners exploring AI-assisted capabilities

The content assumes mixed backgrounds and emphasizes shared understanding across roles.

Want to Go Deeper With Your Own Team?

Whether you attended the course or missed it, we often continue these discussions with teams who want to apply the ideas to their own environment.

These conversations typically focus on:

Reviewing a real workflow or challenge

Identifying where AI assistance may help — and where it shouldn't

Exploring what's feasible in weeks versus months

This is not a sales pitch or formal training. It's a working session to pressure-test ideas and explore next steps.

Have an idea or challenge you're thinking through? Let's explore what's possible.

This page summarizes concepts presented during the SLAS 2026 short course. It is intended for educational and exploratory purposes and does not represent a product specification or commercial offer.

Let's Continue the Conversation

Whether you attended the SLAS 2026 short course or are just getting started with AI in your lab, we're happy to explore how these ideas might apply to your team's specific challenges.