Field Materials AI founder Eldar Sadikov on AI agents that automate procurement for commercial subcontractors, and why construction tech is in a golden era.
AI is changing construction by taking over the paperwork behind buying materials: reading supplier quotes, delivery tickets and invoices, and entering them into the accounting system so people only have to check the work. That is what Field Materials AI does for commercial and civil subcontractors, and founder Eldar Sadikov explained it on Episode 33 of Built This Week. He believes construction technology is entering a golden era, with AI and the data center boom pushing a slow-moving industry to automate.
To open the episode, Sam built a scenario tool based on features we discussed with Eldar before recording. You upload a bill of materials, adjust the price of items like concrete, and see the risk of a cost overrun. Preset shocks, such as a tariff shock or a port delay, show which cost codes are exposed and suggest a procurement response, like locking structural steel prices with a forward contract.
Eldar said it reminded him of Field Materials' pricing intelligence module, which shows what a company has paid for materials across jobs over the year. The platform also handles bills of materials. What Sam's version adds is the ability to throw in variables and see the effect on total job cost, and Eldar said parts of it could be good enhancements to the platform.
Watch Eldar Sadikov on construction procurement in Episode 33 (starts at 5:05) · Episode 33 notes and transcript
Eldar has worked in AI for a long time. He went to Stanford for grad school and founded his first company, an AI company for retail, in 2011. He started Field Materials AI in January 2022, the same year ChatGPT came out.
The idea was simple. Commercial builders buy huge amounts of material, and the process is manual. Someone gets quotes from suppliers and types up purchase orders. When materials arrive at the job site or warehouse, someone checks what was received, what is still outstanding and what is back ordered. Then invoices come in and someone has to verify them and enter them into the accounting software. That work falls on purchasing staff, accounts payable staff and project managers.
In early 2023 the team realized much of it could be automated, because AI could now read the documents themselves. Combine that with tight integrations into the customer's ERP, and the data entry and many workflows could run on their own. Eldar says the product launched in 2023 and took off, growing about 7x one year and 3x the next.
Eldar says about 80% of the customer base is specialized trade subcontractors on commercial jobs: electrical, mechanical (HVAC and plumbing), concrete and drywall. Not home renovation, but the firms that build hospitals, data centers and large office buildings. They buy thousands of different items per project, and data center work comes with very aggressive timelines.
Jordan guessed the industry still ran on pen and paper. Eldar says it mostly runs on email, spreadsheets and manual data entry into accounting software, often on premise, where someone types every line item with quantities and units. Field Materials automates the data entry and the verification steps while keeping a human in the loop.
Nobody likes data entry if you talk to people, or nobody likes going and verifying is this correct or not.
— Eldar Sadikov, Episode 33
Eldar named three levers.
Jordan pointed out that these are not high-tech companies, so building their own vibe coded tools is unlikely to work for multi-year projects of this size. That is the gap a dedicated product fills.
Jordan asked whether new frontier releases, like the latest Claude Opus and Claude Code models, had made a difference. Eldar showed the agents on screen. Field Materials builds a dedicated agent for each document type, such as switchgear quotes, HVAC quotes, trucking tickets and asphalt tickets. Each one reads its document, does the data entry and handles verification and matching. There are now 20-plus of them.
Because we're using these foundational models, when these foundational models get better every single quarter, every single month, our fundamental ability to read, interpret, and verify the documents is also going up.
— Eldar Sadikov, Episode 33
On top of that, the team fine tunes its models on the thousands of documents it has. Next, Eldar says, are agents that go beyond reading documents and carry out steps inside the ERP and the procurement process.
Eldar says construction has lagged on technology for a long time. Its IT spend is often quoted at a couple of percent, far below healthcare, retail or banking, and labor productivity has not risen much in decades. Over the past five to seven years, he says, more construction tech companies have appeared. COVID was one push. The bigger one is the data center boom.
Companies like Google, Meta and Verizon arrive with big jobs and aggressive schedules, and construction firms realize pen and paper and manual data entry will not keep up. So they look for automation, physical AI and robots on job sites, and more prefabrication off site. That pressure is why Eldar believes construction tech is entering a golden era. For more examples, see our AI in construction page and Episode 16, where we also talked about AI agents in construction.
In the news segment Sam brought up two stories: software stocks rebounding after Anthropic announced enterprise plugins and partnerships, and SAP users questioning the value of the company's AI tools. Eldar called the plugins a logical step. The first wave of AI was advice and summaries in a chatbot. The next wave gives AI hands to act on your data and tools. He noted Gemini already had an edge here through Gmail and Calendar.
On SAP, his point was about how AI products get built. You start from the use case, what you can do for the user that the current software cannot, and build the AI to deliver that. Starting from existing software and adding the minimum AI feature produces a gimmick.
The key is not to just check the box and say, yes, we have an AI tool in our toolbox and now you can use AI with this old enterprise software. The key is to make it useful.
— Eldar Sadikov, Episode 33
Jordan agreed. He thinks incumbents have been slow to add AI, so their customers are pulling data out and building AI tools on top. Large companies cannot leave Salesforce quickly, so he expects the squeeze to show up in new customer acquisition. He also pointed to news about Anthropic translating COBOL into modern languages as a sign of how fast competition can arrive. Eldar added that incumbents' AI launches are partly a retention play, much like IBM holding enterprise customers while AWS grew up from startups, and that good AI software takes many iterations to get right.
Eldar shows the Field Materials agents on screen and Sam demos the tariff scenario tool live. Watch Episode 33 of Built This Week, where the full transcript is also on the page, or read more about Eldar Sadikov.
