Write something
Pinned
Jun 3 •
Pick Your First AI Automation
Most people start too big with AI. They try to automate a whole business process, connect tools, rebuild their stack, or pick the “perfect” AI tool. That is usually the wrong first move. The better first move is smaller: Pick one annoying recurring task and turn it into one useful AI output. No integrations. No setup. No new software decision. Just one repeat task, one clear input, and one output you can review. Good first examples: - Weekly notes into a cleaner status update - A long email thread into a short summary and draft reply - Meeting notes into decisions and next actions - Vendor quotes into a simple comparison - A rough idea into an outline and checklist The rule is simple: If AI can prepare a better first draft and a human can still approve the final result, it is probably a good first workflow. If it touches money, legal decisions, HR, customer promises, or sensitive data, keep a human approval step in the middle. Start here: 1. Pick one recurring task. 2. Write what you usually give AI. 3. Write what output you want back. 4. Run it once with safe input. 5. Save the prompt if it works. You are done when you have one small AI output you would actually use. If your mind goes blank, use weekly reporting as the first test: Input: rough notes, wins, blockers, next steps. Output: a clean update you could send to a partner, client, or team. If you want feedback, post your first task in the “Post Your First AI Task” thread.
0
0
Pinned
Jun 2 •
Post Your First AI Task
Use this thread to share one small AI result without posting private business information. Start with one repeated task. Try the workflow. Then share: • Task: What repeated task did you try? • Before: What made the old way slow or annoying? • AI output: What did AI prepare? • Human check: What did you verify, change, or reject? • Result: Did it save time, improve clarity, or show that the workflow was not ready? • Next step: Keep, fix, pause, or remove? Example: • Task: Turn meeting notes into follow-up actions. • Before: Actions were scattered across a page of notes. • AI output: A draft list of owners, due dates, and open questions. • Human check: Two due dates were not agreed, so they were changed to “confirm date.” • Result: The follow-up draft was faster to review, but it still needed a human before sending. • Next step: Keep the workflow and improve the instruction about unconfirmed dates. AI can prepare the work. A human still owns the decision. Do not include customer names, personal data, passwords, API keys, contracts, private financial details, or confidential company material.
Pinned
Jun 2 •
Start Here: Your First 10 Minutes
New here? Start with one small win. Do this: 1. Open the Classroom. 2. Start with the course called Start Here. 3. Pick one task you want AI to help with. 4. Post that task in the community. Good first tasks: - Draft an email reply. - Turn notes into a checklist. - Summarize a meeting. - Make a list of next steps. Do not try to learn everything today. Pick one task. Get one win.
0
0
Jun 8 •
Pick the AI task with the smallest blast radius
Most people pick their first AI task the wrong way. They pick the thing that looks most impressive. “Let AI answer customer emails.” “Let AI run follow-up.” “Let AI make decisions from my notes.” That can work later. But it is usually the wrong first move. Your first AI task should be boring, repeated, and easy to check. Use this simple rule: Pick by reversibility, not excitement. Ask: If AI gets this wrong, can I catch it before it reaches a customer, employee, vendor, or bank account? If yes, it may be a good first test. If no, shrink the task. Here is the best example. Bad first task: Let AI answer customer emails automatically. Why it is risky: - the customer sees it first - it can make promises - it can miss context - it can get the tone wrong - the mistake is hard to pull back Better first task: Let AI draft customer email replies for a human to review. Same workflow. Much lower risk. The AI helps with the blank page. The human still owns the promise. Use this test card: Workflow: Input I already have: Output I want: Who reviews it before use: What could go wrong if AI is wrong: How I will test it safely: Keep / adjust / discard decision: Example: Workflow: Draft replies to customer support emails. Input I already have: The customer email, our FAQ, our tone rules, our refund policy. Output I want: A suggested reply and a list of anything AI is unsure about. Who reviews it before use: Me. What could go wrong: It promises a refund we do not give, or sounds too cold. How I will test it: Run it on five old emails I already answered. Compare the drafts to what I actually sent. Send nothing. That is a safe first test. You do not need a new tool yet. You do not need a giant automation build. You need one small workflow that is safe to test. Your action: Pick one repeated task this week. Run it through the card. If you want help, make a new post with your Workflow Test Card.
0
0
Jun 5 •
Workflow of the Week: Email Thread Reply Draft
Workflow of the Week: Turn a Long Email Thread Into a Reply Draft Long email threads are where AI can help quickly, but only if you keep control of the final message. Use this when a customer, vendor, or team thread has too much history and you need: - a short summary - the real next decision - a safer reply draft - a list of what to verify before sending Do not paste private customer details, passwords, payment info, contracts, HR issues, legal advice, or confidential screenshots into a public AI tool. Start by cleaning the thread: - customer names become Customer A - emails become [email removed] - phone numbers become [phone removed] - addresses become [address removed] - invoice/account/payment details get removed Then use this prompt: Help me turn this long email thread into a safe reply draft. Context: - My role: [your role] - Relationship: [customer/vendor/team/internal] - Goal of the reply: [what needs to happen next] - Tone: clear, calm, professional, no overpromising Email thread, with private details removed: [PASTE CLEANED THREAD] Return: 1. Five-bullet summary of what happened 2. The main decision, question, or next step 3. Missing information I should verify before sending 4. A draft reply I can edit 5. Any risky promises, assumptions, or private details I should remove Rules: - Do not invent dates, prices, approvals, policies, or commitments. - If something is unclear, ask a question instead of guessing. - Keep the reply short. - Make it clear what I need to verify before sending. Example: If a vendor says delivery is “probably next week” and a customer asks if the schedule is confirmed, the safe reply is not: “We will have this delivered next week.” The safer reply is: “Thanks for checking in. I am confirming the latest delivery timing now. I do not want to give you a date until I have the updated vendor confirmation. I will follow up by tomorrow morning with the confirmed next step.” Human approval rule: Before sending, check whether AI invented a date, price, refund, discount, policy exception, approval, or promise.
0
0
1-30 of 41