The first time we walked through a steel service center, the GM was running his $30 million operation on a combination of MetalTrax, QuickBooks, three shared Excel spreadsheets, and a whiteboard in the warehouse that everyone called "the brain." The brain tracked which orders were being processed, which trucks were loading, and which customers were upset that day. When the warehouse manager went on vacation, nobody could read his handwriting.
That was not an outlier. That was normal. We visited 14 service centers over three months, from a 10-person shop in Houston to a 200-person operation outside Chicago. The technology ranged from a DOS-based system installed in 1998 to a cloud ERP that the sales team had quietly stopped using because it took 11 clicks to generate a quote.
The pattern was the same everywhere: smart people running complex businesses on tools that did not respect their complexity.
What Bothered Us
Steel service centers are among the most operationally complex small businesses in America. A 30-person center does manufacturing (processing), distribution (logistics), financial services (trade credit), retail (will-call), and commodity trading (pricing). Name another business type that juggles this many functions under one roof with 30 people.
And yet the software options were either horizontal tools that ignored the industry (NetSuite, SAP Business One) or legacy vertical tools that ignored the last 20 years of software development (on-premise, desktop-only, designed before smartphones existed). There was nothing in between. Nothing that combined deep steel industry knowledge with modern software architecture.
That gap bothered us. Not in an abstract, market-opportunity way. In a "how is this possible" way. Toast built a $15 billion company understanding that restaurants are not retail stores. Procore reached $12 billion knowing that construction sites are not factories. Nobody had done this for steel.
What We Got Wrong Early
We assumed service centers wanted more features. They did not. They wanted fewer screens.
Our first prototype had a dedicated module for everything: a CRM module, a quoting module, an inventory module, an order module, a processing module, a shipping module, a finance module. Each one was well-designed in isolation. Together, they recreated the exact problem we were trying to solve: too many places to look, too many clicks to get things done.
A sales rep at a service center does not think in modules. She thinks in workflows. A customer calls, she needs to see their account (CRM), check what is in stock (inventory), generate a price (quoting), and confirm delivery (logistics) in one fluid motion. Our first design made her jump between four screens. The industry tools she was replacing made her jump between four applications. Same problem, prettier interface.
We scrapped three months of work and started over with workflows instead of modules. Every screen answers a question: What does this customer need? What do we have? What will it cost? When can we deliver? The data comes from what other software would call separate "modules," but the user never sees those boundaries.
What Changed Our Thinking About AI
We did not start as an AI company. We started as a steel software company that happened to launch at the same time the AI wave hit. The early temptation was to bolt AI onto everything because that is what investors and the market expected.
Then we spent a day with a sales manager in Detroit who showed us his quoting process. He had 47 open quote requests. Each one required checking inventory, looking up the customer's pricing history, calculating processing charges, estimating freight, and formatting the document. He was working through them at about four per hour. Simple math: he could not respond to all 47 in a single day.
His best customers were not getting faster service than his newest prospects. Everyone waited in the same queue. And while they waited, competitors responded first.
That conversation reframed how we thought about AI. Not as a feature to market, but as a way to give that sales manager his time back. The AI reads the RFQ, pulls the relevant data, generates a draft quote in the correct format, and presents it for review. The sales manager's job shifts from data assembly to decision-making. He reviews the AI's work, makes adjustments based on his knowledge of the customer and the competitive situation, and sends the quote. Four per hour becomes twenty per hour.
That was the moment we knew this was going to work. Not because the technology was impressive, but because the problem was so clearly painful and the solution so obviously better.
What We Believe
Steel service centers are essential infrastructure. They sit between mills and the end users who build buildings, manufacture equipment, construct bridges, and assemble vehicles. When service centers operate well, the supply chain moves. When they do not, projects stall, costs rise, and relationships break.
The people who run service centers are among the most resourceful operators in any industry. They have built successful businesses despite their software, not because of it. They deserve tools that match their capability.
We are building those tools. Not a generic platform with a steel skin. A system designed from the ground up by people who understand that a 48-inch by 120-inch sheet of A36 plate is not a box of screws.