The Space Between: Why AI's Next Gains Sit Outside Your Four Walls

Order accuracy, procurement, and the quiet cost of the industry's handoffs

Last month, we looked at the accuracy problem and how to use AI confidently by matching the level of verification to the level of risk. That conversation was about what happens inside your own business. This month, the canvas gets bigger. What happens when AI starts working between businesses?

Glazing is a chain. Systems companies, fabricators, distributors, installers, and the builders and homeowners at the end of it. Every business in that chain has spent years optimising its own operation, and rightly so. But a striking amount of the industry's cost of error doesn't sit inside any one business. It sits at the handoffs. The order that didn't quite match the survey. The confirmation nobody read closely. The delivery date that moved without anyone downstream hearing about it. The remake that turned a profitable job into a break-even one.

We've learned to treat these as the cost of doing business. They aren't. They're the cost of information moving between businesses through emails, PDFs, portals and phone calls, checked by busy people at the end of long days. And that is precisely the kind of work AI has become very good at.

Start with the most expensive handoff: the order. On the installer side, AI can compare an order against the survey notes before it's submitted, checking colours, handing, sizes, and the details that were flagged on site but never made it onto the order form. On the fabricator side, the mirror image: flagging incoming orders that fall outside standard parameters before they reach the production floor. Same principle at both ends of the transaction. Catch the discrepancy while it's still a keystroke, not a remake.

Then confirmations. Most order confirmations get filed, not read. There's rarely time to check thirty line items against the original order, so the checking happens at the delivery door instead, or worse, on site with the customer watching. An AI agent can read every confirmation as it arrives, compare it line by line against what was ordered, and flag anything that changed. A moved delivery date. A substituted product. A quantity that doesn't match. Ninety-five per cent of confirmations will sail through untouched. The five per cent that don't are exactly the ones costing you money.

Deliveries deserve the same attention. Lead times in this industry move, and the businesses that hear about it last absorb the cost: fitting teams stood down, customers rearranged, scaffold hire extended for a week nobody budgeted for. An agent that monitors expected delivery dates against confirmed ones, and raises a flag the moment something slips, turns a Friday surprise into a Tuesday adjustment. The information almost always existed somewhere. It just didn't reach the person who needed it in time to act.

Procurement is the strategic layer above all of this. Most glazing businesses hold years of supplier data they've never analysed. Spend by supplier. Remake rates. Delivery reliability. Lead time drift over the past eighteen months. Upload that data and ask AI what it sees, and you put numbers behind what has always been gut feel. This isn't about catching suppliers out. It's about better conversations. A supplier review backed by data is a negotiation. One backed by frustration is an argument.

It's worth noting the direction of travel here. This year's wave of agent platforms from the major AI companies is pushing at exactly this space: AI that doesn't just draft and summarise, but monitors, reconciles and acts across connected systems. The technology increasingly assumes that businesses will link their tools together. Which raises the question this column is really about.

What happens when your customers and suppliers are running these checks and you're not? The businesses that make themselves easy to trade with, through clean orders, consistent references, and confirmations that a machine can read as easily as a person, will quietly rise to the top of preferred supplier lists. The cost of being the messy link in the chain is about to go up.

The honest caveats. Many of the systems in this industry still don't talk to each other, and getting data out of some of them remains harder than it should be. Data quality matters as much here as anywhere. And not every trading partner will engage at the same pace. Relationships remain the foundation of this industry; a good account manager will still fix problems no AI agent can.

But most of what fails at the handoffs isn't relationship failure. It's attention failure. Nobody had time to check. AI has time.

There's also a first-mover dynamic worth being candid about. None of this requires the whole chain to modernise at once. An installer can start checking confirmations tomorrow without asking the fabricator's permission. A fabricator can start flagging out-of-parameter orders without waiting for installers to change how they order. Each business that adds a layer of checking makes its own operation more reliable, and quietly raises the standard for everyone it trades with. That's how industry practice actually shifts. Not through a grand agreement, but through one business at a time deciding the errors aren't acceptable any more.

So here's the practical step. Map the handoffs where errors cost you money last year. Orders, confirmations, deliveries, invoices. Pick the most expensive one and put AI to work checking it, with a person reviewing what it flags. Then move to the next.

The next efficiency gain in glazing probably isn't inside your factory or your office. Most of that ground has been walked. It's in the space between you and the businesses you trade with, and the first movers there will be hard to catch.