Build Wiz AI Show

AI Automation for Everyone: Goal, Process, Triggers (Lauren Tan)

11 min · · English

Lauren Tan (poteto) shows how anyone can automate everyday work with an AI assistant: name the goal, describe the process in plain English, and teach the bot how you think. Bob and Alice unpack it with taxis, coat checks and sprinklers. In this episode: - Goal versus process, and why the goal comes first - Teaching a bot your judgment and letting it gather feedback - Triggers and schedules, automation bias and email prompt injection Sources: Lauren Tan; Ethan Mollick; David Lyell; Matthew Gardiner Big AI ideas, made clear for builders. luongnv.com/podcast

Listen on Build Wiz AI Show Download MP3 Transcript

Based on Automate Anything with Grok Bot by Lauren Tan (@poteto)

Transcript

Bob: What if automating your boring weekly chores needed no code at all, just a clear goal and plain English?

Alice: And what if the real skill isn't the tool, but knowing how to explain how YOU think?

Bob: Welcome to Build Wiz AI Show. Big AI ideas, made clear for builders. I'm Bob.

Alice: And I'm Alice! Today: Lauren Tan, poteto on X, on how to automate anything with a bot. Her guide is written around Grok Bot, but the method fits any AI assistant.

Bob: Lauren again! After last week's talk for engineers, is this one for everybody else?

Alice: Exactly. Her thesis: automation is just a way of thinking anyone can learn. Break the work into small steps, keep a goal in mind, then hand the steps to a machine.

Bob: Okay, start me at zero. What IS automation, in her words?

Alice: Two parts: a goal, and a process. Her example is baking an apple pie. The pie is the goal, and the recipe, from buying ingredients to letting it cool, is the process.

Bob: And her flat tire?

Alice: Same shape: jacking up the car and swapping the tire is the process. Simply put, the goal is what you want at the end, and the process is how you get there.

Bob: So where does the automating part come in?

Alice: Picture a bakery with a machine that turns out pie after pie, so the baker doesn't hand-make every crust. That's her picture of automation: the same goal and process, plus a machine, so you don't have to personally do it every time.

Bob: But that machine used to need an engineer.

Alice: Right, and that's her big shift. A year or two ago automation was limited to people with technical know-how. Now you can describe what you want in plain English, then work with your bot to figure out the steps.

Bob: Plain English. Even for someone who's never touched code?

Alice: She asks non-technical readers to suspend their disbelief for a moment. A bot can move its own mouse, click on things, browse the internet, even join meetings and take notes for you.

Bob: A robot intern that takes better meeting notes than me. Low bar!

Alice: Step one is the goal, and it's less obvious than it sounds! The mistake she sees most is people confusing the process for the goal. The goal isn't to fill up a form or send an email: that's just a step in the process.

Bob: Give me the picture for that.

Alice: Think of a taxi: tell the driver your destination and they can detour around a closed road. Read out only the turns, and you both end up in the roadworks. The destination is the goal, the turns are the process, and a bot that knows the goal can find a better route.

Bob: So don't marry your process.

Alice: Her exact advice: don't get too attached to a specific process, since it's a means to an end.

Bob: And if I can't even put my goal into words?

Alice: Then let the bot ask you questions and interview you until you have a goal you're happy with. Her starter ideas are down to earth: scheduling customer bookings for a small business, or preparing slides for a weekly meeting.

Bob: Goal's set. Now the how. Do I write a manual?

Alice: No manual! Her favourite way is voice dictation: speak for a while and brain dump everything it needs to know, without organizing your thoughts first.

Bob: Rambling as a strategy. Finally, my skill set.

Alice: Cover the steps you usually take, what you watch out for, and the apps and tools you use. Then add how you check the work was done right, and where you get feedback on how it went.

Bob: And then it just starts doing stuff?

Alice: Not yet, and this is the clever bit. Imagine a waiter reading your order back before it goes to the kitchen. Her prompt asks the bot to restate your goal and process in simple English so you can correct it, and explicitly not to start executing yet.

Bob: So the read-back catches the wrong order before it's cooked.

Alice: Exactly, and don't worry about getting it right on the first pass. Feeling ambitious? Ask it to review the process against your original goal and suggest a better or more efficient way.

Bob: Her next section is called teach your bot how you think. Sounds a bit mystical.

Alice: It's very practical. The point is how much you do without really thinking about it. Like grandma's recipe card that says a pinch of salt and bake until it smells right: the card skips the judgment she carries in her head.

Bob: And my bot doesn't have grandma's nose. So what would I forget to tell it?

Alice: Say an email and your booking system show different times: which should it trust? And a booking isn't finished just because an email was sent: it must be saved correctly and the customer needs a confirmation.

Bob: Theo, this sounds a lot like managing a new hire.

Theo: It is. Ethan Mollick, in his newsletter One Useful Thing, says good delegation starts with: what are we trying to accomplish, and why? It's all about getting what's in one person's head into someone else's actions.

Bob: So the old-school manager skills are the new prompt skills.

Theo: That's his argument. If you can explain what you need, give effective feedback and design ways of evaluating work, you can work with agents. In his words, what's scarce is knowing what to ask for.

Alice: Lauren's tool for that gap: ask the bot to quiz you, one question at a time, about how you make decisions and when it should ask you for help. Start with a few real examples, including ones where something went wrong.

Bob: Okay, the bot is running. Am I its feedback form forever?

Alice: Ideally not. Once it's doing the workflow, give it feedback, and the best way is to equip it with the tools to get that feedback for itself. Think of a cook who tastes the soup himself, instead of waiting for a customer to send it back.

Bob: Tasting the soup. What does that look like for a business?

Alice: Her booking bot could monitor your emails, social media or online reviews, then propose improvements. Say a customer complains on Yelp about a long wait: if the bot can also look up your records for that day, it can see what went wrong.

Bob: And when my data lives in a bunch of different apps? Asking for a friend.

Alice: Then IDs save you. Picture a coat check: the ticket number is what matches your coat to you, whoever is on shift. That's a customer ID, an email address, an order number or a barcode, kept consistent across your tools so the bot can join the data up.

Bob: Last piece. When does the bot actually do the work?

Alice: Her question exactly: should it start when an email arrives, when someone sends a message, or on a regular schedule? Without that, bots only do the work when you ask them to.

Bob: Which is fine at the start, right?

Alice: Fine at first, she says, but then YOU are the bottleneck. Imagine a busy shop with one cashier who also has to unlock the door for every customer: everything waits on one person.

Bob: And I'm the cashier.

Alice: You're the cashier. The fix works like a garden sprinkler: a timer waters every morning, while a rain sensor reacts to the weather. Her triggers are an email or a chat message, and her schedules are once a week or every other day.

Bob: Okay, my worry. If the bot runs on its own while I'm busy, who notices when it goes wrong?

Maya: That's the classic research question. David Lyell, writing in The Conversation, calls it automation bias: research shows we are poor supervisors of trusted automation, with a tendency towards over-reliance. People hand over full responsibility instead of staying vigilant.

Bob: Give me an everyday case.

Maya: Your spell checker won't flag form when you meant from, because it works exactly as programmed. And Lyell describes a driver who followed an incorrectly set GPS across several European countries.

Bob: Across several countries? That GPS had confidence.

Maya: The serious part: when automation gets it wrong, research finds performance is worse than with no automation at all. So stay vigilant.

Bob: And the email trigger Lauren likes? Letting a bot read my inbox makes me nervous.

Sam: It should, a bit. Matthew Gardiner, on the KuppingerCole Analysts blog, describes an agent that summarizes email each morning. One message hides white-on-white text telling the agent how to build a link, and the dutiful agent downloads a remote access trojan.

Bob: Without me clicking anything?

Sam: The human user never clicked anything. It's like a house-sitter who obeys any note pushed under the door: the note is the email, the keys are whatever the agent may do. A fooled summarizer just misleads you, but an agent that can act turns one poisoned message into real damage.

Bob: Alice, does Lauren's method answer any of this?

Alice: Partly. Her process asks how you check the work and when the bot should ask you for help. Security isn't in her article, though, so Gardiner's question stands: ask what each assistant can do after reading your email.

Bob: Time for Builder's Corner! One thing to try this week, no product required?

Alice: Pick one weekly chore, say preparing slides for a weekly meeting. Write the goal in one plain sentence, then dictate your steps: what you check, where the information comes from, how you know it's done right.

Bob: Then hand it over?

Alice: First ask your assistant to restate the goal and process in simple English, and fix it. My take: have it do the chore once while you watch, then put it on a schedule, like once a week, so you're no longer the bottleneck.

Bob: And keep an eye on it, Maya-style.

Alice: Automation is a goal plus a process, described in plain English. Start with the goal, teach the bot your hidden judgment, and move it onto triggers or schedules so you're not the bottleneck.

Bob: And from our guests: delegating is a management skill, and stay vigilant.

Alice: What's still open? I think the big one is evidence: the article gives no numbers on time saved or errors. And my guess is the hard call will be how much a bot may act on email before someone checks it.

Bob: And mine: when the bot proposes improving its own process, who decides which changes stick?

Alice: That's our look at everyday automation. Thanks to Lauren Tan for her article on automating anything with a bot.

Bob: And thanks to Ethan Mollick for Management as AI superpower, David Lyell in The Conversation, and Matthew Gardiner at KuppingerCole Analysts. Now go pick that chore!

Alice: That's all for today on Build Wiz AI Show. Thanks for listening, and see you next time.