Picture this: You're racing against the clock on a crucial steel building project. The pressure is mounting, specifications are complex, and missing even one critical check could have serious consequences. Every engineer has been there - that mix of excitement and anxiety when facing a challenging design task.

But what if you had a brilliant mentor by your side, one that could instantly point you to exactly the right code sections and specifications, available 24/7?

Enter Clark, a groundbreaking AI chatbot from the American Institute of Steel Construction (AISC). Unlike typical AI tools that might give unreliable information, Clark is your trusted engineering companion, drawing exclusively from AISC's vast library of peer-reviewed resources - specifications, manuals, guides, and technical articles. It's like having the entire AISC knowledge base at your fingertips, ready to guide you through any steel design challenge.

During a private meeting with AISC leaders, I learned that Clark was developed through 18 months of intensive work and fact checking by Professional Engineers to ensure it is reliable.

Unlike conventional chatbots, it doesn't calculate or replace engineering judgment—instead, it enhances decision-making through quick access to relevant technical information.

Let's explore how Clark empowers engineers, especially those new to the field.

Features and Benefits

While Clark is still not fully accessible for everyone (as of the publishing of this article), we know that it stands out from typical chatbots by specializing in delivering resource-based information from AISC's knowledge repository, including specifications, design guides, and technical papers.

Here's how it works:

• Specification Guidance: Ask Clark about design considerations, and it directs you to relevant AISC specification sections. For specific beam design constraints, Clark pinpoints appropriate chapters and tables, saving time of manual searching and possible not interpreting the code correctly.
• Resource Access: Clark provides comprehensive access to design guides, technical papers, and supporting documentation. From simple beams to complex structural systems, all necessary information is readily available.
• Check Reminders: Young engineers benefit from Clark's ability to flag potential missing checks. It analyzes queries and suggests overlooked considerations, helping prevent costly mistakes.
• Time Efficiency: For time-sensitive projects, Clark delivers information swiftly. Engineers receive answers with references instantly, enabling faster, informed decisions.

When I asked whether Clark could understand questions contextually, such as "What should I consider when designing a beam with specific constraints?" AISC explained that Clark was trained using relevant specification sections and related designs, allowing it to continuously learn from and understand user queries. This makes Clark an advanced search engine tailored for steel design—not just finding information but contextualizing it for practical use. AISC has mentioned that Clark will provide screenshots along with the answers as a way to “check” itself and minimiza errors.

Addressing Concerns

While powerful, Clark has important limitations that AISC acknowledges:

• Math and Calculations: Current AI technology, including Clark, isn't suited for calculations. As an AISC leader noted, "AI is not good at doing math. If you give it a math problem, it may give you a wrong answer because it doesn't actually do the calculations, it searches for it." Engineers must rely on their expertise and calculation tools for design tasks. Clark focuses on guidance, not computation.
• Hallucination: To prevent AI-generated misinformation, Clark provides references and screenshots of its sources. By drawing exclusively from AISC's curated knowledge base, it ensures accuracy and reliability. This transparency allows engineers to verify and trust the information.

<aside> <img src="/icons/light-bulb_green.svg" alt="/icons/light-bulb_green.svg" width="40px" />

In AI, a "hallucination" refers to when a model, especially a large language model (LLM), generates outputs that are factually incorrect, misleading, or entirely made up, even if they appear plausible

</aside>