To view the figures and tables associated with this article, please refer to the flipbook above.
Widespread acceptance of any new technology or innovation comes after a period of wariness and circumspection, but that timeline from introduction to adoption is accelerating. Structural engineers are balancing how to take advantage of new tools that could save them time and improve their work while also being careful to ensure the integrity of the work remains intact.
With this in mind, STRUCTURE approached three experts from software companies that serve structural engineers in different ways to talk about developing trends in technology. They are: Matt Cooper, CEO of BQE Software, a provider of firm management software geared toward architects and engineers; Mark Riffey, CEO of ENERCALC, which contributes structural engineering calculation software; and Josh Taylor, Vice President of Structural Engineering at Bentley, which develops, manufactures, and sells engineering design software for a wide variety of infrastructure assets such as bridges, airports, skyscrapers, and power plants. Their answers focused on the opportunities in general for structural engineers and the advancements in technology, particularly AI. They also shared words of advice for judicious implementation of the evolving tools.
How will AI impact the future use and development of software technology for SEs? In what ways will software further utilize AI to aid structural engineers in their work?
Matt Cooper, CEO, BQE Software: When it comes to AI, lots of firms are considering how it will impact the practice of structural engineering. However, we anticipate AI will also have a substantial impact on how SEs manage their firms overall. AI will help SEs interact more effectively with clients, staff projects, forecast project outcomes, improve data analytics, and more. Ultimately, AI will improve business practices and lead to more efficient and profitable firms.
Josh Taylor, VP of Structural Engineering, Bentley: I'm particularly excited about AI's potential to accelerate the earlier stages of the design process by reducing the "overhead" in engineering workflows. By overhead, I mean tasks in which engineers aren't fully utilizing their critical engineering judgement. This would include things like model set up, preliminary loading and design criteria, and establishing the standards for reports and drawings. The real ingenuity and efficiencies that structural engineers contribute to a project come when they are able to see how all the systems in a structure work in unison. This only happens once they have a solid starting point that they can conduct "what if" studies on and fine tune. We can envision AI getting engineers to this starting point much faster than they are able to presently.
Mark Riffey, CEO, Enercalc: AI is already impacting the use and development of software for SEs.
Larger vendors have significant resources to apply to AI projects. We are already starting to see them roll out AI-based tools in their software (example: Autodesk AI, an Autodesk tool to augment creative exploration, automate tedious tasks, and analyze data to provide predictive insights).
Even starting from a blank screen, you can provide AI with a verbal description of what you want (faster than typing), and even with less than perfect speech recognition, you can get pretty close. You might say, "Pretty close isn't perfect"—and you're right, but that isn't the point. We use these tools because they get to 60-80% of the desired outcome much faster than we do when doing the work without them.
Imagine a programming task that normally takes 3 hours from start to finish, starting from a blank screen. For some types of work, you can get these tools to get you to 80% in 2 minutes. Maybe it takes you another 30 minutes to polish it to where you wanted it, but you're still ahead by almost 2.5 hours.
AI isn't about replacing engineers. It's about empowering engineers by accelerating the mundane work they do, so that they can expend their mental energy on the most important engineering tasks and make better decisions on the things that require engineering judgment.
In what ways will software further utilize AI to aid structural engineers in their work?
Riffey: It's important to focus on what these tools do better and faster than we do. People are pretty good at finding patterns in visual data. AI is great at finding patterns in high volumes of visual, written, or binary data that would overwhelm us.
AI might also tell you who is more efficient at designing this type of structure vs. that kind.
These tools can be merged into existing systems so they can analyze existing data that's difficult for us to assess quickly. "For this beam, tell me what section sizes we typically use for (your use case) over the last 20 years."
Vendors like Qnect already have software to analyze constructability of a designed structure, work that's related to steel connections. Imagine the types of analysis AI can do across an entire project build with a global analysis tool.
AI can look at project data for something you designed years ago that is now being expanded or upgraded. When you do this work, you have to dig around to figure out what changes are going to be required to meet updated governing codes and design standards. AI could do this in the time it takes to open your analysis and design software.
Is the structural engineering community using software technology to its fullest extent? Which areas of the profession hold the greatest opportunity for improvement through the adoption of available software?
Cooper: Our purview is firm management software, and within that domain, the SE community is not using software technology to the fullest. More than 50% of small-to-midsized firms continue to use Excel, Quickbooks, or one-off point solutions rather than firm management-software purpose-built for the industry. Most of the firms that are using dedicated firm management software are still on antiquated/legacy platforms and are well past due for a replacement.
In fairness, the software industry has underdelivered in this domain for SE firms in the past. However, there is now easier-to-use software with better functionality that enables firms to drive more efficiencies and insights to run firms more effectively and keep employees happier.
Taylor: One aspect of engineering software that continues to grow in popularity, but I believe is still underutilized, is the use of programming interfaces (APIs) available in the software. APIs allow users with some coding know-how to automate routines that would otherwise take lots of button clicks and manual operation, as well as push and pull data to and from the application to other digital tools. What users are doing with these APIs has grown tremendously in sophistication. Entire digital ecosystems are being built off commercial software by utilizing them. As you might expect, engineers are starting to inject AI into these workflows. In fact, the winner of 2023 Bentley Going Digital Awards for Structural Engineering was Hyundai Engineering, who paired STAAD.Pro with an AI algorithm for optimizing the geometry and framing configurations for mechanical shed structures.
Riffey: Some great structural engineering software is available via open-source and from independent software vendors. Even so, there are significant gaps in functionality, workflow, and interoperability—collectively. For example, software A might talk to software B, but only via a 1990s style import/export of a CSV file. Software C might talk to software A but not to software B—or engineers are forced to cut and paste, or worse.
Other structural engineering software might use more advanced methods (like APIs) to communicate with certain other software. Despite that, there's no end-to-end solution for engineering firms. As a result, they're forced to cobble together a solution from multiple high-quality software packages that fill a specific, necessary purpose, augmented by manual processing.
Engineering firms need full-spectrum holistic solutions. AI is a part of that solution, but is not the sole cure.
Software is a powerful tool but like any tool, misusage can be damaging. What are some pitfalls to be avoided when it comes to employing software? Do you have words of advice or general guidance you tell customers when they are implementing a new tool or product into their systems?
Riffey: Just like with analysis software, when using an AI tool, it's important to approach it with a clear expectation of the desired outcome—you shouldn't simply follow it blindly. AI tools aren’t designed to think for you. Using them effectively is a skill we can all develop to better leverage their capabilities.
Public large language models (LLMs) like those used by OpenAI's ChatGPT were in large part built on public-facing websites and data on the internet. No matter how you feel about this, I'm quite sure how you would feel if your project data was used while working with a vendor's LLM chat interface and that data ended up as part of the LLM vendor's training database.
What that means is that your data would become public. Imagine that you asked an LLM to analyze your sales, or a new product you're building, or some valuable new technology that you're working on. Now imagine that the LLM uses that data for training and your competition runs a ChatGPT type query to assess the competitive issues in your market. The data about the new technology you built could become part of the response your competition gets when they run their query.
This is why it's essential to make absolutely sure that your proprietary data is not exposed to public facing systems. This includes your code, your spreadsheets, your databases, descriptions of future products, etc.
Cooper: We see three common pitfalls. First, software implementations require BOTH leadership support and an internal owner responsible for the project. Second, these implementations take time, and so adequate capacity needs to be allocated. Finally, firms often try to force-fit the software 100% to their existing internal processes, rather than take the opportunity to refine those processes.
It’s worth noting: any time there is a major project requiring change management, the initial excitement often gives way to temporary dip in sentiment due to the time and effort involved. This dip often occurs right before seeing that proverbial light at the end of the tunnel. It’s important that firms recognize that they are on a journey and commit to completing the implementation so that they can achieve the anticipated benefits.
Taylor: Fortunately, structural engineering is governed by some immutable principles, among them the field of engineering mechanics. Regardless of what an engineer uses as an aid (a calculator, spreadsheet, finite element software, or a machine learning tool), the results can be assessed using these principles. Whenever employing any type of new tool, very well understood benchmarks should be employed to gain trust in the tool. Use a past project or a simple verification problem that you understand well. There are design examples available in the public domain. We also ship verification examples with our products that can be used for this purpose.
As AI/ML is concerned, there are many lessons we can take from the emergence of desktop software in structural engineering circa the early 1990s. There was similar skepticism, and I would say healthy skepticism, about the possibility for overreliance on what it produces. For example, the emergence of auto-design functionality required an investment of time for designers to evaluate if they were comfortable with what the software produced. They accordingly made decisions on where they were willing to employ these algorithms and where they felt manual design methods were more appropriate. The game plan for getting confident with AI's possibilities and limitations will be similar.
Do you have any other thoughts on the current and future trends in software for structural engineers that you would like to share?
Taylor: There are an enormous number of possible applications of AI to structural engineering. As a provider of software solutions, conversations with our users are indispensable in learning how we best support the evolution of their AI/ML processes—particularly, what the common needs are across the industry. AI/ML is a means, with the ultimate goal of making our users more productive than they are now while retaining the confidence they have in our solutions.
Riffey: I believe an engineering software vendor's highest purpose is to take the mundane, tedious, complicated, unrewarding work off of engineers’ plates so that they are freed to do the work software can't or shouldn't do. We can do that while leaving critical engineering decisions in the hands of the engineer where they belong.