Latest / Intentional AI Daily / The Low Risk Way To Learn AI Agents
Transcript
- 0:00SPEAKER_01: What's happening everyone?
- 0:01SPEAKER_01: Jason Wright here.
- 0:02SPEAKER_01: Joining me as always is Georgia.
- 0:04SPEAKER_01: I want to give a rule today that I wish somebody had given me.
- 0:07SPEAKER_00: Let's have it.
- 0:08SPEAKER_01: Your first AI agent should be built for yourself, not for a customer.
- 0:12SPEAKER_00: Why?
- 0:12SPEAKER_00: Customer facing is where the money is.
- 0:15SPEAKER_01: Customer facing is also where being wrong costs you a client.
- 0:18SPEAKER_01: Internal, you get to be wrong in private.
- 0:21SPEAKER_00: How wrong are we talking?
- 0:23SPEAKER_01: The first one I built hallucinated a meeting that didn't exist and put it on my calendar.
- 0:28SPEAKER_01: If that had been a client's calendar, that's a phone call I don't want to make.
- 0:33SPEAKER_00: Okay.
- 0:33SPEAKER_00: That's a good argument.
- 0:34SPEAKER_00: But doesn't that just delay the learning?
- 0:37SPEAKER_01: It front loads it.
- 0:38SPEAKER_01: You find out what breaks while the only person affected is you.
- 0:41SPEAKER_00: What should somebody build first then?
- 0:44SPEAKER_01: Whatever you do every week that you resent.
- 0:46SPEAKER_01: For me it was pulling numbers out of three dashboards on Monday mornings.
- 0:51SPEAKER_00: That's not glamorous.
- 0:52SPEAKER_01: Nothing good is.
- 0:53SPEAKER_01: Glamorous is a demo.
- 0:55SPEAKER_01: Useful is the thing that saves you forty minutes every Monday for a year.
- 1:00SPEAKER_00: What did you learn building it that you wouldn't have learned otherwise?
- 1:03SPEAKER_01: How much of the work is the plumbing?
- 1:05SPEAKER_01: Getting data in the right shape, handling the times it fails.
- 1:09SPEAKER_01: The AI part was maybe fifteen percent.
- 1:12SPEAKER_00: How do you handle the failures?
- 1:14SPEAKER_00: Because it will fail.
- 1:15SPEAKER_01: You decide up front what it does when it doesn't know.
- 1:18SPEAKER_01: Mine stops and asks me instead of guessing.
- 1:21SPEAKER_00: That sounds like it defeats the purpose.
- 1:24SPEAKER_01: It saves ninety percent of the time and flags the ten percent that needs a human.
- 1:28SPEAKER_01: That's a good trade.
- 1:30SPEAKER_01: Guessing on the 10% is what Rex trust.
- 1:33SPEAKER_00: How long before you knew it was actually working?
- 1:36SPEAKER_01: About three weeks.
- 1:37SPEAKER_01: The first week I checked everything, second week I spot checked, third week I stopped thinking about it, which is the real signal.
- 1:44SPEAKER_00: That's a longer runway than most people expect.
- 1:47SPEAKER_01: Everybody expects it to work on day one.
- 1:49SPEAKER_01: Nothing works on day one, and that's fine when the only stakeholder is you.
- 1:54SPEAKER_00: So when do you go customer facing?
- 1:56SPEAKER_01: After you've run something internally for a couple months and you know exactly how it fails, then you can price it honestly.
- 2:03SPEAKER_00: Because you know what the support looks like.
- 2:05SPEAKER_01: That's the number that kills people.
- 2:06SPEAKER_01: They price the build and forget the babysitting.
- 2:09SPEAKER_00: Good thing to know before you quote it.
- 2:11SPEAKER_01: One quick thing before you go, if your calendar isn't as full as you'd like it to be, the problem usually isn't your marketing, it's your offer.
- 2:18SPEAKER_01: I put together a free one page checklist for you.
- 2:21SPEAKER_01: It covers twelve things your offer needs before you spend another dime on ads or another six months on content.
- 2:27SPEAKER_01: Head over to highticketchecklist.com and grab it.
- 2:30SPEAKER_01: It takes about two minutes to score yourself, and you'll know exactly where the leaks are.
- 2:36SPEAKER_01: Thanks for hanging out with us.
- 2:37SPEAKER_00: Thanks, everyone.
- 2:38SPEAKER_01: See you in the next one.
- 2:40SPEAKER_01: Stay curious and stay intentional with AI in your business.