From Sprawl to System: Rethinking the Glazing AI Tool Stack

Why the next stage of AI adoption is about how tools fit together, and who gets to use them

Last month, we looked at the space between businesses. The handoffs where orders, confirmations and deliveries quietly leak money, and how AI can stand watch over them. Running through that column was an assumption worth examining: that your business has the right tools in place, and that the people who need them can actually use them. This month, the tool stack itself.

There's a pattern in how glazing businesses have adopted AI over the past two years, and it's visible across almost every company I speak to. It starts with one subscription, usually the owner's ChatGPT account. Then Copilot arrives for the office because it comes with Microsoft 365. Then a transcription tool for meetings. Then something for images, something for the website chat, maybe an add-on inside the quoting system. None of these were bad decisions. Each one solved a real problem. But nobody designed the result, and the result is what I'd call subscription sprawl: half a dozen logins, half a dozen monthly charges, data scattered across all of them, and nobody entirely sure what the business is paying for.

Sprawl creates two problems, and the first is quieter than it looks. Most AI tools are priced per user, per month. That sounds modest at ten or twenty pounds a seat until you multiply it across every tool and every person who could benefit. So businesses do the rational thing: they ration. The owner gets the accounts. Maybe the office manager. The surveyors, the fitters, the order processors — the people this column has spent two years arguing have the most to gain from AI — are usually last on the list, if they make the list at all.

And that rationing undermines the whole project. The consistent lesson from the businesses getting real value out of AI is that the return comes from breadth, not depth. Not one power user doing something clever, but a whole team saving twenty minutes here and catching one error there, every day. If only two people in a fifteen-person business have access, AI isn't infrastructure. It's a hobby.

The second problem is one we covered back when this column looked at glazing intelligence: generic tools don't speak this industry. Every session starts with context. Every output needs translating into the language of profiles, cills, handing and lead times. A general assistant can be taught, but the teaching cost is paid over and over, by every person, in every chat.

Here's what's changing, and you can see it first in the AI world itself. The most capable AI products are converging on the same pattern: many capabilities, one place. ElevenLabs started as a voice tool and is now a content platform where you choose between different AI models without leaving it. ChatGPT, Claude and Copilot are all becoming places other tools connect into, rather than apps you visit for one job. The people building AI have concluded that the enemy of value is fragmentation — hopping between disconnected tools, re-explaining yourself at every door.

There's a deeper reason consolidation wins, and it's the most important point in this column. An AI is only ever as good as what it can see. Give a capable model a blank chat window and you get competent, generic answers. Give the same model your surveys, your quotes, your order history and your product range, and it stops being generic. It knows the job before you ask. Better-informed AI simply performs better; every piece of business information it can't see is a question it will answer worse.

Our industry, to its credit, already understands the first half of this story. Business Pilot proved the value of running the whole operation through one system, and arguably does it better for glazing than the big construction platforms manage for theirs. That was the consolidation of workflow: one version of the truth about what's happening in the business.

What's emerging now is the next layer: the consolidation of intelligence. And for once, glazing isn't waiting for other industries to go first. Platforms like The Glazing Hub are built on exactly this principle — multiple AI-driven tools living in one ecosystem, working from the same shared data, so the AI checking a document already knows what the survey said, and a specification builds itself from notes that already exist. Each tool makes every other tool smarter. In the interests of full transparency, thinkivity is involved there, so read this knowing that. But the principle stands whoever builds it: an AI with a deep understanding of your business beats a smarter model with no memory of it.

Pricing matters too. As AI spreads across more roles and more tasks, per-seat charges would rack up the outgoings fast. That's why AI tools are increasingly priced by usage rather than by seat. It makes them more accessible, and more scalable: you can put your whole team on them without adding another line of cost for every name.

This isn't only an installer conversation. Fabricators and distributors carry the same sprawl in different clothing, and clean, connected data multiplies in value when it crosses the boundary between businesses.

Whoever the vendor, the questions are the same. Does it speak this industry? Can everyone use it without a per-head penalty? Does the data flow — survey feeding specification, specification feeding quote — or does each step start from scratch?

The caveats still apply. A platform doesn't fix a bad process; it industrialises it. Ask how you get your data out, not just in. And the verification habits from the accuracy column matter more, not less.

This month's practical step is an audit. List every AI subscription, what it costs, who has access, and where the data ends up. Few owners have seen that picture in one place; most are surprised by it.

For two years, the question was "which AI tool is best?" That question is expiring. The better one is: what does our AI look like as a system? And that points to the real positive: AI isn't only changing what your business can do. The model it arrives with is changing who can do it.