Measuring the Return: How Do You Know If AI Is Actually Working?

Most glazing businesses feel the benefit before they can prove it. Here’s how to change that.

Last month, we looked at AI agents moving from writing things to doing things such as background tasks, automated follow-ups, jobs that happen while you’re talking to customers. The honest message was that these tools are real, they’re available now, and the businesses getting value from them are starting small and building from there.

This month, the question that should come next but usually doesn’t. How do you know if any of it is working?

Most glazing businesses using AI now are going on feel. It saves time. Things move faster. There’s less to chase. All of that might be true, and the feeling is usually a reliable early signal. But feeling isn’t enough to make a business decision, and it isn’t enough to justify expanding what you’re doing or explain to a sceptical colleague why it’s worth investing more time.

The reason businesses don’t measure is usually not laziness. It’s that AI tends to improve things in small increments across several tasks at once, which makes it hard to isolate. The quote that took forty minutes now takes twenty-five. The follow-up list that used to sit for a week gets done the same day. Nothing is dramatic enough to stand out on its own. It adds up, but where?

The first step is to pick one thing and measure it before and after. Not the whole business. One task. Choose the thing you use AI for most consistently. Drafting responses to enquiries, preparing quote follow-ups, summarising site notes, whatever it is. Before you use AI: time yourself on that task, once or twice, to get a rough baseline. After you’re using AI consistently: time it again. The gap is your starting point.

It doesn’t have to be precise. If a task that used to take thirty minutes now takes twelve, the argument for AI is already made. If a task that used to take thirty minutes now takes twenty-eight, either the tool isn’t the right fit for the task, or you haven’t got the workflow right yet.

Volume is sometimes more useful than time. If you’re using AI to help with quote follow-ups, the metric isn’t how long each follow-up takes, it’s how many you’re sending per week compared to before. That’s the number that connects to revenue. If you were following up on four leads a week and you’re now following up on ten, with the same team, AI is delivering.

The trap to avoid is measuring AI in isolation. Time saved by AI doesn’t disappear, it gets absorbed into something else. If you free up two hours a week and those two hours go straight into surveying, customer calls, or business development, the AI benefit is real, but the time measurement won’t show it. Ask instead: what did that time go into, and was it higher-value work than what you were doing before?

For business owners with a team, there’s a simpler version. At the end of each month, ask the question in a ten-minute conversation: where has AI saved us time this month, and what did we do with it? The answers won’t always be precise, but they’ll be honest. Over three months, a pattern will emerge.

The reason this matters isn’t just internal. It’s competitive. The businesses that can say “we handle thirty percent more enquiries with the same team” or “our response time went from a day to two hours” are building something they can describe clearly, train new staff around, and build on. The businesses that just feel like AI is helping are harder to scale and harder to defend.

Measurement doesn’t have to be complex. A simple note in your CRM, a column in a spreadsheet, a five-minute Friday review. The bar is low, and even basic tracking compounds over time.

AI is at the point where the question has shifted from “should we try this?” to “are we making the most of it?” Measurement is how you answer the second question honestly.

If you’ve been using AI for three months or more and you can’t describe the difference it’s made in concrete terms, that’s the project for this month. Not because the benefit isn’t there, but because naming it is how you build on it.