(This column originally appeared in Forbes)
Let’s say you’re a contractor or running a business in the construction industry. And you’ve got questions like these:
· “I’d like to use a design from a prior job and get recommendations for applying it to this new project.”
· “There’s a piece of equipment we’re using and I need specific instructions for using it.”
Or like these:
· “We’re installing a helical pile. Do we need to load test? What have we done before? What does the manufacturer recommend?”
· “How do the submitted electrical bids for this potential project compare?”
Or these:
· “We need a storm water pollution plan that addresses the requirements of this project and is in compliance with regulations in Florida.”
· “What slope is needed for drainage on this site?”
Where do you go for this information? You can search the internet, where results can be dubious and potentially inaccurate. But wait…don’t you already have a lot of this information in your own files? If you’ve been in business for a while you probably do: emails, bids, estimates, documents, contracts, design plans, manuals. But there’s a problem: there are thousands of these files, which makes it very time consuming to search them, let alone know if you’re coming up with the best or most current data.
That’s one of the biggest issues I see with generative AI. There’s no one place to get your information and then ask more questions about it in order to arrive at the best answer.. But a company needs to access the data in all of their databases and file systems in order to make the best decisions. The good news is that a number of startups in a number of industries are tackling that issue now.
One of them is Constructable, a year old company that recently graduated from startup accelerator Y Combinator. Constructable is focusing on the construction industry and has begun to consolidate the information from files found in multiple places into its proprietary large language model that can then be trained in order to provide specific information to a company from their past data as well as external sources…just by chatting in one place.
“We’re really focused on commercial general contractors, and especially project managers,” says Emma Nazim, one of the company’s three co-founders. “We’ve talked to a number of people who are just spending hours a week trying to answer questions that they need to keep the project moving forward, and also to make sure that they’re preventing mistakes.”
Making mistakes is obviously the biggest concern for any project manager, particularly when building things. Constructable is designed to provide quick — and most importantly accurate — answers to their questions.
“If you ask a question and we find two data sources that have a different answer, we’re able to surface that,” she says. “We’re getting better at figuring out the best answer.”
Large Language Models that consolidate data from multiple sources are not just a construction industry need. I know many business owners and managers in other industries that desire the same thing from AI.
For example, I have a manufacturing client that collects data in their order entry and inventory systems as well from sensors on some of their equipment on the shop floor and video cameras on their forklifts. I know many businesses in the service industry that are storing information in separate systems tracking time, materials, expenses, job progress, client communications and contracts. I have distribution clients that are managing their warehouses through various software to track inventory, security, temperatures, safety and product movements.
And I personally know a dozen construction companies that are using multiple systems used for project management, accounting and job costing as well as estimating and design. Some are even testing drones for site surveys and using augmented reality headsets for training and remote management. Regardless of the industry, all of these companies have files and information in their payroll and customer relationship systems. All of them use either Office 365 or Google Workplace.
There’s data everywhere. How do you chat with it all in one place? That’s the challenge being addressed by Constructable and other startups like it. And it starts with files.
Constructable’s cloud-based modelling is first pulling together information from the documents, spreadsheets and emails stored in Office 365, Workplace and Procore — a popular construction industry management application — into one place for its customers to converse, query and get the information they need to estimate, run jobs, communicate with customers and manage their business.
“We’re really heavily focused right now on the documents,” says Nazim. “We’re pulling information from bids and RFI/RFPs (Request for Information/Proposal) and then we’ll continue to expand that as we build out the product and actually help with answering questions that comes from more structured data.”
Nazim also says that a model is being built to access the files in the design and engineering platform AutoDesk. Future versions will expand to not only include more data from popular construction industry applications but also images and videos from job sites.
The aim is to provide quick and accurate information from these places in conversational form.
“We tell our customers think about our AI technology as having a smart and fast reading assistant,” Nazim says. “How would you talk to your coworker? Don’t think about it in the context of what would you type in the search bar, but how would you actually ask a person what you’re looking for? “
