ChatGPT for Construction - What It Actually Does (and Doesn't)

Join ContractorOS: https://www.skool.com/contractor-os/a... Work with me: https://contractoros.build/work-with-us/ Follow on LinkedIn:   / josh-turner-pmp-10608b1b7   ChatGPT for construction changed this month. ChatGPT Work now connects it to your whole project folder, drawings and specs included, and it runs the same .skill files Claude does. This is the full setup, start to finish. Browser vs the desktop app and why the desktop one matters for us, the settings to sort before you put any business information through it, how to get a skill library loaded, the new Sol, Terra and Luna models, and then the part that actually matters: pointing ChatGPT Work at a real project folder and giving it enough context that it answers questions about the job properly instead of guessing. The example running through it is an electrical subcontractor's job folder, built around a real 42 sheet drawing set. WHAT YOU'LL LEARN Getting set up Why the desktop app connects to your actual project folders and the browser only connects to cloud projects The two settings to change before you put any business information through ChatGPT How to load the .skill files you already have, plus the workaround for getting them onto the desktop app Sol, Terra and Luna: which model for which job, and what effort and speed actually cost you Pointing it at a real job Why you point ChatGPT Work at one project, not your whole company drive How to run it against SharePoint or Google Drive without duplicating everything What agents.md does, and why ChatGPT reads it every single time The project indexer: turning a folder of Word, Excel and PDF into context ChatGPT can use Why context beats prompting now, and what that changes about how you set a job up Business context: the standards and preferences that follow you onto every project Connecting it to the rest of your stack What read and write access actually means when you connect a plugin, and why it matters when the other end is a database Connecting a structured database so it can check work against your own standards Where the database schema lives inside a skill, so it queries the right thing every time Letting it run on its own Scheduled tasks: a correspondence log that reads the day's email and updates itself Why you run a task by hand three or four times before you ever put it on a schedule Making it something the team uses Sites: turning an Excel register into an RFI viewer your team will actually keep current Why a front end over a structured database beats a spreadsheet everyone forgets to update Stacking it into an operating system How one workflow's output becomes the next one's input, so the context compounds over the job Inside a workflow map: breaking a phase into workflows, and naming which parts are yours and which are the AI's Working through pre-construction with it, from onboarding the project to the go/no-go CHAPTERS 0:00 - What changed in ChatGPT this week 1:15 - Browser vs the desktop app 4:49 - The settings to change first 7:03 - Skills and plugins in ChatGPT 12:27 - Uploading a .skill file to the desktop app 14:36 - Sol, Terra and Luna: picking the model 17:01 - What ChatGPT Work actually is 19:00 - Think in projects, not companies 22:26 - Running it on SharePoint or Google Drive 25:19 - Setting up the project folder 26:08 - The project indexer and agents.md 30:59 - Why context beats prompting 31:45 - Business context 32:24 - Connecting a plugin: Supabase 33:34 - Read vs write: what you're granting it 35:06 - Putting your database schema in the skill 36:14 - Scheduled tasks 37:20 - Test it manually before you schedule it 38:58 - Sites: a register your team will use 40:23 - A front end over a structured database 42:03 - Stacking workflows into an OS 42:41 - Inside a workflow map: your role vs the AI's TOOLS ChatGPT Work (desktop app), ChatGPT Plugins, ChatGPT Sites, Supabase, GPT-5.6 (Sol, Terra, Luna) #ChatGPTWork #ChatGPT #ConstructionAI #AIforConstruction #ContractorOS