How Haskell cut submittal reviews by an hour with AI

In a construction environment on warp speed, global design-build leader Haskell dared to test Procore’s agentic AI on its own headquarters buildout. 

The trial worked to make massive datasets manageable and retrievable and, in doing so, helped reclaim hours of project time every day that would otherwise have been lost. 

The challenge: Adrift in oceans of data 

With digitization, construction lost mountains of paper but gained oceans of information. 

Teams can waste hours a day hunting for technical data trapped inside submittals, contracts, and giant spec books.

Traditionally, the trade partner who needs clearer direction tracks down an assistant project manager (APM). Time ticks away and frustration builds as they search documents manually or wade through extraneous details called up by generic software. Control-F only gets them so far. 

Haskell demanded better. While renovating a Jacksonville, FL, landmark into its state-of-the-art headquarters, Haskell leveraged Procore’s construction-specific AI. The goal was to help field workers, including trade partners, comb the ocean of project data to quickly and accurately retrieve what they need. 

The Solution: An agile AI assistant for all

Haskell didn’t overcomplicate deploying AI. At its HQ, the company simply posted QR codes around the jobsite that let people access its AI agents. Curiosity did the rest, as team members and trade partners scanned to see what it could do.

“They're clicking it, they're interacting with it, and they're asking, ‘What is this Procore AI? Let me ask it a question. Let me see what it comes back with…' I see a lot of people that have hesitation at the beginning, but once they start using it, you see that light bulb moment.” 

– Hamzah Shanbari, Director of Innovation, Haskell

Flexibility to use the AI as each worker saw fit revolutionized the workflow at Haskell’s HQ. Now, all people need to do is:

1. Look around: With QR flyers throughout the jobsite, anyone who scans, including trade partners, can go straight to the AI agent.

2. Just ask: Workers can ask questions like, "What are the specific submittal and warranty requirements for the roofing membrane on this project? Pull up the exact section in the specs and show me the source drawing if it’s available."

3. Briefly wait: The AI agent quickly combs through the project’s data spectrum, including specs, submittals, and drawings, to generate an answer.

4. Verify outputs: As the agent delivers its response, it simultaneously surfaces source material to enable human verification.

The results: Streamlined communications. Improved document visibility. And perhaps most importantly, potential time savings across workflows.

AI in action: Streamlining submittal audits

Using Procore AI, Haskell custom-templatized an AI agent to shorten submittal reviews. 

With it, APMs use the agent to compare the trade partner's submittal against design specs to surface potential variances for team review.

The Haskell team estimates that every use saves one hour in an APM’s day. At the same time, it helps to deliver organized data instead of a messy jumble. This helps Haskell’s teams to do their best work instead of being tied up in manual data retrieval and comparison. 

“The human is the thought leader. The AI is only the thought partner. And if you're guiding it into what you want it to do, you are going to accomplish things faster and more efficiently with higher quality."

– Hamzah Shanbari, Director of Innovation, Haskell

The Impact: Reclaiming Hours and Expediting the Schedule

Haskell’s trial saved measurable time across the project. Here's a closer look:

Haskell’s rigor showed that Procore’s agentic AI can hold up under a modern jobsite’s demands.