#aiautomation

pro.glitch's profile picture

You have exactly four months to change how you work forever with AI and Automation Four months. September, October, November, December. That's exactly enough time to go from wherever you are to running AI automations and your first AI employee. But the whole thing hinges on one week — this one. SEPTEMBER — foundations + your first automation. Learn the words fast, then build one small thing. By September 30, for the first time in your life, something you built runs while you sleep. OCTOBER — AI in the middle. Your automations start to think: the email comes in, it reads it, decides, drafts, hands you only the tricky ones. By Halloween, the busywork is off your plate — a machine handles it, in your voice, all day. NOVEMBER — your first real AI employee. Connect it to your tools, write its instructions, give it memory, set the brake so the risky stuff waits for your yes. By November 30, you have an employee working one real job, every single day, getting sharper every time you correct it. DECEMBER — multiply. The second takes a fraction of the effort. And then New Year's Eve arrives: everyone around you is writing resolutions… and you're already running. Here's the catch, and it's the only one: four months is exactly enough, with nothing to spare. So the version of you holding all of this at midnight doesn't start in October — they start this week. The four months are the easy part. This week is the decision. That's the whole difference between the two people you could be at midnight on the 31st: the one who started, and the one who meant to. Do it right — in order, from your first automation, with a group starting the exact same week. Get in this week and start September on foundations. It's all in my profile. #aiautomation #aiagent

Professor Glitch

360,300 views • 2 months ago

pro.glitch's profile picture

How to build an AI front desk for your info@ inbox in Microsoft 365 (no $30 Copilot seats) Built for real in a Microsoft 365 lab: shared mailbox → Power Automate flow with cheap checks first (own mail, out-of-office, duplicates, mail loops) → a Copilot Studio agent that answers only from your SharePoint files → a Teams approval card → the reply goes out from info@ in the same thread. Plus a chase if nothing's answered and a morning digest. Tested on 20 emails: 18/20 routed right on the first run, 8/8 after two instruction fixes, zero wrong replies reached a customer. The listening flow costs nothing; the AI costs a few cents per drafted email. The build, step by step: 1. Microsoft 365 admin center: create a shared mailbox called info@ (no licence needed) and give yourself access. 2. Make a Teams team for the front desk. In its SharePoint, add your own documents (prices, booking and cancellation policy, vendor rules, FAQ, directions) plus two lists: who handles what, and an inbox log. 3. Copilot Studio: create an agent on a fast, standard model (not a reasoning one), point its knowledge at those files, and tell it to answer only from the files, name the file it used, and hand dates, money, complaints, gate codes, invoices and anything it can't find to a person. 4. Power Automate: trigger on a new email in the shared mailbox, run the cheap checks before any AI, log the rest and send it to the agent. 5. Teams: the flow posts a card with the draft, the reason and the source file. Approve and the reply goes out from info@ in the same thread. 6. Nothing gets forgotten: a second flow chases anything unanswered after 24 hours, and a third posts a morning digest. Two setup details that stop most people: a company data policy can block the agent from reading your own files, and the agent needs generative orchestration turned on or it ignores its instructions. Works for venues, clinics, agencies, trades, offices, any small business with an info@ inbox and Microsoft 365. #copilotstudio #aiautomation #powerautomate #claude #microsoft365

Professor Glitch

22,400 views • 9 days ago