Industry landing pages, and cutting the copy back to what the product actually does

What I built

Today I built industry-specific landing pages on the Dialessa website for real estate, trades and finance brokers. Dialessa is an AI voice sales agent for Australian businesses. A tradie, a property agent and a broker each want to know whether it fits the way they work. A generic page leaves them to work that out on their own.

I didn't write each page from scratch. First I added a shared content structure so every industry page draws on the same building blocks. Each page fills those blocks with its own industry content, while the layout and the order of sections stay the same. Adding another industry later should mean writing content, not building a new page. It also makes the pages easier to review side by side, because they all have the same shape.

I also added tests for the new industry pages, so I find out early if a change to the shared structure breaks any of them.

What was hard: the copy overreached

The build itself was the easy part. The words were harder.

Once the pages were written, I had separate automated reviews check each industry page. Each review looked at its page on its own instead of all three together. They all found the same problem. The copy overreached. It said things that sounded good for each industry but went further than what the product actually does.

This is easy to slip into with industry pages. When you write for a specific audience, you want to show you understand their problems. From there it's a short step to suggesting the product solves every one of them. Each sentence on its own sounds reasonable. Together they make promises the product doesn't keep.

So I rewrote the copy to say only what the product actually does. Some of the more confident lines came out. What's left is plainer, but I can stand behind all of it.

Shipping it

After the rewrite I deployed the updated website to production. Here's the day by the numbers:

• 1 working session with coding agents
• 2 prompts typed
• 2 commits
• 1 deploy
• about 1 hour of active work
• 1933 lines of code changed

Most of those lines are page content, the shared structure and tests, not complicated logic.

What I learned

• Review copy the way you review code. I run checks on behaviour without thinking about it. Running checks on claims turned out to be just as useful. The reviews caught things I'd skimmed past.
• Review each page separately. Each review looked at a single page, so the overreach in each industry's copy couldn't hide behind the others.
• Shared building blocks are worth setting up first. Putting the structure in place before the content kept the pages consistent and made the tests simple to write.
• Speaking to an industry is not the same as promising it everything. A page can be specific to a tradie or a broker and still be honest about what the voice agent does.

The pages are live, and the claims on them are ones the product can back up.