Why ChatGPT (and Other Generative AI) Will & Won’t Revolutionize Construction

From?Elvis taking selfies to a seemingly sentient chat bot , it’s hard to get on the internet these days without seeing an AI reference. Artificial Intelligence has broad implications in our personal lives, but you have plenty of news outlets to tell you about that. We’d rather discuss what AI in construction management can and should look like now and over the next 50 years. The industry has begrudgingly progressed from paper blueprints and computers to SaaS and smart devices, and AI presents another huge technological leap that will leave many companies behind and catapult others ahead. Here’s why you should be aligning yourself with partners and technology that will ensure you stay ahead.

Can AI replace most construction jobs?

You’re probably reading claims that AI and systems like ChatGPT will replace [insert profession here]. Marketers, customer service representatives, even sales teams “better watch out.” We’ve even seen discussions on the feasibility of artificial intelligence lawyers. The reality is AI models are only as good as the data they’re built on.?

AI will never be enough on its own, it’s a derivative of human input and will therefore always need human input.?

Consider the GPT-3 chatbot and the newly released GPT-4. This is the OpenAI chatbot people are asking all sorts of questions from “How to write a letter of recommendation” to “When is the best time to fly to Brazil?” It’s a search engine on steroids that chats with you in natural language and doesn’t serve you spammy ads in the process (well, at least not yet. ) It’s more powerful than any AI chatbot we’ve seen available to consumers, but it has two noteworthy limitations:

1. The GPT-3 model is built from publicly available data on the internet.?
2. The data is only current up to September 2021.?

GPT-4 claims to include a broader data set, but OpenAI is not releasing what that is.

Any industry-specific questions you may ask ChatGPT, such as, “Write me a commercial construction contract” will only elicit answers based on public data up to 2021 or data that OpenAI will not confirm the source of. We can’t think of any contractors who publish their construction contracts publicly on the internet.? Even if they did, estimates indicate that nearly 90% of construction contracts are bespoke, customized per project. In addition, contract language and conditions have changed significantly since 2021. For example, material price escalation clauses are adapting to inflation and supply chain costs. If you’re using ChatGPT to produce binding agreements on multi-million-dollar projects, you can see how that would be risky. You’re going to need legal review, experienced contractors and proprietary industry data to guide that AI toward a useful, accurate result. This is the case whether you’re using ChatGPT to draft contracts, brainstorm design ideas, or shape communications to shareholders for your construction company - until that AI model is combined with human review, proprietary data and subject matter expertise, it’s not immediately actionable and may be critically wrong.?

That being said, OpenAI, and the other companies developing the data and models behind interfaces like ChatGPT, are setting the foundation for all construction software you’ll use in the future. If SaaS solutions don’t have an AI component now, they won’t be powerful enough in the coming? years. The answer isn’t “AI will replace everything” nor “AI can’t replace everything,” it’s somewhere in the middle.?

Technology solutions built on top of AI models, finely tuned with human input, will replace everything, but we’ll still need experienced construction professionals ensuring quality inputs and outputs.?

Solutions from some? of our friends like?Togal.AI are using artificial intelligence to improve estimating.?Mercator.AI is helping contractors identify project leads earlier than competitors.?Cutting Edge AI is analyzing job site video feeds for risks and progress. Even if you’re not using one of these AI-first solutions, you’re probably still using software with embedded AI. If your teams employ CAD/BIM software, jobsite cameras, document scanning, or site condition sensors, you’re likely relying on AI. Whether you see it or not, proprietary data models are processing your project team’s input and spitting out smarter, more useful data because of it.?

What We Will & Won’t Build with AI at Document Crunch

Our goal is not to be the best AI contract review software. Our goal is to empower EVERYONE to understand their construction contracts. That goal is tech-agnostic. Our philosophy on AI comes back to two of our core values: Be Inspired and Growth Minded. We're a curious company and because our product is built on a foundation of high quality, finely tuned AI, we closely monitor the evolution of AI and Machine Learning (ML). We are constantly experimenting to understand how we can leverage each new development to accomplish our mission: Helping EVERYONE in the construction industry to understand and optimize their contracts. Our customers don’t care if we’re an AI company and our mission is not to be the most advanced AI company. We don’t even call ourselves a tech company, meaning we are not married to any sort of technology.?

We just want the best technology to deliver on our mission, and AI is now a requirement to deliver the best solution for our customers.?

Our core products couldn’t exist without AI. It empowers our software to constantly learn from new contracts, compile knowledge from real construction professionals and attorneys, and as a result:

• Identify risks with high accuracy by leveraging state-of-the-art machine learning
• Automatically generate compliance checklists that Project Teams can use to manage key terms and conditions
• Provide internal and market-driven benchmarks that are continuously learning from our database of contracts and industry partnerships
• Share insights from our proprietary knowledge graph that expands organically as our customers review each and every contract with Document Crunch

We're still learning what AI models are capable of and we’re not using them blindly. Our customers cannot be the product testers, so it's our responsibility to provide appropriate guardrails, calibrated data, and a great product experience in AI contract analysis.

OpenAI choosing not to disclose the data set used to power ChatGPT-4 is controversial, but not unfounded. We believe there’s always a place for proprietary datasets and models. OpenAI is providing a free product that we can choose to use, or not. We choose to use it and proceed with necessary caution to ensure there's a human subject matter expert in the loop to validate and ensure the quality of any AI models our software incorporates. There are many things these black box models do very well, like drafting and summarization, that are very helpful accelerators for us and our customers, but we use them to power our product, not as the product itself. You should be wary of any software companies that do.

For example, we often get asked if talking to ChatGPT or a similar ability to talk to an AI will be added soon to Document Crunch. The answer is yes and no.?We know one of our biggest value levers stems from getting our customers the right information at the right time, and not really from some open ended search.??

The Future of AI in the Construction Industry

We believe purpose-built AI models, like ChatGPT, will soon be as pervasive and necessary to software products as spreadsheets and databases. Learning how to leverage AI capabilities will be required to compete for everyone, from tech companies and the contractors they serve to the legal professionals who minimize everyone’s risk. Thanks to Hollywood’s portrayal of artificial intelligence taking over the world, this may seem scary. In reality, it’s going to make everyone’s job more efficient, and we think, more enjoyable.

The goal of most technology is to make human workflows more valuable, not replace them.?

We see so many opportunities for this in construction AI. From generative design to predictive cost analysis to contract review, there's a world of opportunity in pre-construction and project administration. On the construction site, there's so much that can be accelerated through AI imaging, change order tracking and materials management. Post-construction, we all know there are many opportunities for improved facilities management, which is where 80% of the cost of ownership resides.

Years ago, we saw automated machinery scare the industry into thinking robotics would replace human labor - and here we are still lamenting over labor shortages. We need humans, and we need technology to continue to elevate their skills and expertise. We are excited to be a part of how AI will do just that in construction for contractors, lawyers and software companies alike.

Every month, we’ll publish what we’re learning and discussing around artificial intelligence on our blog. Follow along so you’re ahead of the AI discussion at your company by subscribing below.