AI writing has a tell, and your LPs, brokers, and tenants have already learned to spot it. That's a problem when your firm's brand depends on sounding like a specific set of people rather than a language model with a real estate vocabulary.

This is a build session, not a lecture. Brad builds a Claude skill file live, working from a real firm's writing, then walks through the same process with attendees so everyone leaves the session with their own skill file already tested, not a set of notes to implement later.

You'll leave with a working voice skill file running against your firm's actual content types: LinkedIn posts, LP updates, deal memos, whatever you write most.

You'll learn how to

• Build a Claude skill file live during the session, not from a template you fill in later

• Break your own writing down into codable elements: vocabulary, rhythm, tone

• Write instructions specific enough that Claude actually follows them

• Test your skill file in real time against your firm's actual content, LinkedIn posts, LP updates, deal memos

• Catch and fix the moments where output still reads as generic AI, on the spot

• Leave with a maintenance process so the skill file gets sharper with use instead of going stale

Questions about the workshop?
View FAQs on format, access, recordings and more: FAQs

Why AI Writing Sounds Like AI

The specific patterns readers and LPs have learned to recognize:

• Common tells: hedge words, uniform sentence rhythm, generic openers

• Why this matters more for firms whose brand depends on a distinct voice

• The gap between writing that's fine and writing that's yours

Live Build: Deconstructing a Voice

Brad builds a skill file from scratch, working from real writing samples:

• Pulling real examples and separating vocabulary, rhythm, and tone into distinct elements

• Deciding what's core voice versus personal quirk that doesn't need to travel

• Attendees do the same with their own writing in parallel

Live Build: Writing the Skill File

Structuring the file for real, durable impact on output:

• Writing instructions Claude actually follows instead of politely ignoring

• Balancing consistency with flexibility, so you're not just trading one robotic voice for another

• Attendees build their own file alongside Brad's, step by step

Testing Against Real Content

Getting it built is step one. Getting it to hold up is step two:

• Running the skill file live against LinkedIn posts, LP updates, and deal memos

• Diagnosing where it breaks down and adjusting on the spot

• Setting up a feedback loop so it improves every time it's used