WEBVTT
00:00:00.080 --> 00:00:06.080
I've watched AI state something completely wrong with the same tone of voice it uses when it's right.
00:00:06.240 --> 00:00:10.800
No hesitation, no hedge, just confident and false, delivered like a fact.
00:00:10.960 --> 00:00:15.679
If you've used this stuff for more than a week, you already know that feeling.
00:00:15.839 --> 00:00:18.480
The moment you catch it, after you almost didn't.
00:00:18.640 --> 00:00:26.000
A few episodes in now, you know, I've talked about the faith case, the practical jobs, the bottleneck, and the some setup work.
00:00:26.160 --> 00:00:28.719
Today I want to tell you where I got it wrong.
00:00:28.879 --> 00:00:34.560
Because if I only show you what works, I'm not giving you the whole picture and you deserve the whole picture.
00:00:34.799 --> 00:00:37.600
Well, here's something I haven't said outright in this series yet.
00:00:37.759 --> 00:00:45.359
You know, I've let AI produce something I didn't check closely enough, and it came back wrong in a way that mattered.
00:00:45.520 --> 00:00:51.679
Not once, more than once, unfortunately, before I built a habit that fixed it.
00:00:51.840 --> 00:00:56.880
I want to walk through what that looked like, why it kept happening longer than it should have.
00:00:57.039 --> 00:00:59.600
And, you know, one thing I changed to stop it.
00:00:59.759 --> 00:01:00.719
The pattern went like this.
00:01:00.799 --> 00:01:09.200
I'd ask AI to pull something together, you know, a summary, a set of numbers, a draft, you know, that reference details from an earlier conversation.
00:01:09.359 --> 00:01:16.879
And it'd come back clean, you know, well formatted and confident, reading like it knew exactly what it was talking about.
00:01:17.040 --> 00:01:18.319
And sometimes it didn't.
00:01:18.480 --> 00:01:22.400
A date would be off, a number would be close, but not right.
00:01:22.640 --> 00:01:36.959
A detail from you know something I'd told it, you know, three messages earlier would uh come back slightly changed, close enough that I wouldn't have caught it if I hadn't happened to remember the original myself.
00:01:37.200 --> 00:01:39.359
The output never looked uncertain.
00:01:39.439 --> 00:01:43.760
It never said something like, I'm not fully sure about this part.
00:01:44.079 --> 00:01:50.480
It just stated the wrong thing with the same tone it used for you know the parts that were correct.
00:01:50.640 --> 00:01:56.239
And unless you already knew the right answer, you'd have no way to tell the difference by reading it.
00:01:56.480 --> 00:01:58.159
I caught some of those mistakes.
00:01:58.319 --> 00:02:00.159
I didn't catch all of them early enough.
00:02:00.319 --> 00:02:06.000
And the ones I caught late cost me more than the time I saved by using AI in the first place.
00:02:06.159 --> 00:02:08.159
Here's what that actually looks like in practice.
00:02:08.319 --> 00:02:10.319
So this stays concrete instead of vague.
00:02:10.400 --> 00:02:18.000
You know, say you ask AI to summarize a call and pull out the follow-up date you agreed to with a client.
00:02:18.159 --> 00:02:20.159
You know, it gives you back a clean summary.
00:02:20.319 --> 00:02:23.599
Three bullet points, professional, done in 10 seconds.
00:02:23.919 --> 00:02:32.719
Except the date is wrong by week because somewhere in the conversation you mentioned two different possible dates before landing on the real one.
00:02:32.800 --> 00:02:36.639
And it grabbed the wrong one with total confidence.
00:02:36.879 --> 00:02:44.879
You paste that summary into a calendar invite or a follow-up email without you know rereading it closely because it looked finished.
00:02:45.039 --> 00:02:50.960
Now you've told a client the wrong date, and you're the one who has to clean that up, not AI.
00:02:51.199 --> 00:02:52.319
That's a small example.
00:02:52.479 --> 00:02:59.439
Multiply it across every summary, every draft, every you know, quick pull of information you didn't personally verify.
00:02:59.599 --> 00:03:09.199
And you can see how a habit of trusting clean looking output without checking it compounds into something that eventually costs you somewhere that matters.
00:03:09.439 --> 00:03:14.479
I want to explain why you know this happens because understanding it changes how you use the tool.
00:03:14.719 --> 00:03:22.000
AI isn't looking something up the way you'd search for a fact in a book, it's predicting the most likely words based on patterns.
00:03:22.080 --> 00:03:32.960
And most of the time, that prediction lands close to true because true things tend to be more common in pattern and in the data, you know, that it learned from.
00:03:33.280 --> 00:03:36.800
But close to true and true aren't the same thing.
00:03:36.960 --> 00:03:40.080
And when it misses, it doesn't know it missed.
00:03:40.240 --> 00:03:43.039
There's no internal alarm going off.
00:03:43.199 --> 00:03:52.400
It hands you the wrong answer with you know the exact same confidence as the right one because from the inside they feel identical to produce.
00:03:52.639 --> 00:03:55.120
That's the part that catches people off guard.
00:03:55.199 --> 00:04:02.400
We're used to a certain kind of person who hedges when they're unsure, you know, who says something like, I think, or, you know, let me double check that.
00:04:02.560 --> 00:04:07.439
AI doesn't naturally do that unless you specifically ask it to flag uncertainty.
00:04:07.599 --> 00:04:11.439
And even then, it isn't always right about when it should be uncertain.
00:04:11.599 --> 00:04:17.839
You know, confident and correct look the same as confident and wrong right up until you check.
00:04:18.160 --> 00:04:23.519
I found this shows up most on you know the details that feel too small to matter until they do.
00:04:23.680 --> 00:04:28.480
Someone's name, you know, spell it away that looks right, but isn't the way they actually spell it?
00:04:28.639 --> 00:04:38.959
A dollar figure rounded in a direction that seemed harmless, a quote attributed to the wrong person because two names came up close together in the conversation you've had it.
00:04:39.120 --> 00:04:42.639
None of these announce themselves as risky.
00:04:42.879 --> 00:04:49.680
They read exactly like the correct version would read, which is precisely why they get missed.
00:04:50.000 --> 00:04:59.040
This is where I want to be honest about the stakes because you know, I don't think most men fully feel this until it happens to them.
00:04:59.199 --> 00:05:03.920
A wrong number in an internal note costs you a little embarrassment and nothing more.
00:05:04.160 --> 00:05:07.439
A wrong number in something that goes to a client costs you credibility.
00:05:07.519 --> 00:05:10.720
And credibility is slow to rebuild once it cracks.
00:05:10.959 --> 00:05:17.839
A wrong detail in something you say, you know, to your own team gets repeated as if it were true.
00:05:17.920 --> 00:05:21.040
And now the mistake isn't just yours anymore.
00:05:21.199 --> 00:05:22.639
It's loose in your business.
00:05:22.800 --> 00:05:29.519
And if a wrong detail gets in front of the wrong person at the wrong time, it can cost you a relationship you spent years building.
00:05:29.680 --> 00:05:34.639
There's a version of this that's purely financial, and men understand that version fastest.
00:05:34.959 --> 00:05:46.800
A pricing error that goes out in a proposal, you know, a wrong figure in something, you know, you send an investor or a partner, a miscalculation that makes it into a document someone signs.
00:05:46.959 --> 00:05:54.480
Those are the mistakes that get fixed quickly because the cost shows up on a screen where everyone can see it.
00:05:54.720 --> 00:05:57.279
The version that worries me more is quieter.
00:05:57.439 --> 00:06:04.800
It's the erosion of trust that happens when someone catches you being wrong two or three times, you know, on things that should have been simple.
00:06:04.959 --> 00:06:07.040
They don't always say something either.
00:06:07.199 --> 00:06:09.680
They just start double checking everything you send them.
00:06:09.759 --> 00:06:13.040
Or worse, they start assuming you're just careless.
00:06:13.120 --> 00:06:20.240
Uh, and the truth is you got moved by a tool that made carelessness easier to fall into without meaning to.
00:06:20.560 --> 00:06:23.199
You know, that kind of cost doesn't show up on a screen.
00:06:23.360 --> 00:06:35.120
It shows up in the in the tone of the next phone call, the hesitation before someone signs something you sent, you know, the quiet decision to loop in someone else to confirm what you already told them.
00:06:35.360 --> 00:06:40.879
None of that is the AI's fault in any way that matters because AI didn't send the message.
00:06:41.120 --> 00:06:41.680
You did.
00:06:41.920 --> 00:06:49.120
Whatever it produces, the responsibility for what goes out under your name is still yours every time, no exceptions.
00:06:49.279 --> 00:06:50.240
That's not a burden.
00:06:50.319 --> 00:06:52.399
You know, AI removes from you.
00:06:52.639 --> 00:06:54.480
It never was going to be.
00:06:54.639 --> 00:06:55.759
Here's what changed.
00:06:55.920 --> 00:06:59.839
I stopped treating a clean looking output as a finished output.
00:07:00.000 --> 00:07:10.399
Specifically, anything with a number, a date, a name, or a direct reference to something, you know, someone told me, I check against the source before it goes anywhere.
00:07:10.560 --> 00:07:15.360
Not the whole document every time, just the parts where being wrong actually costs something.
00:07:15.519 --> 00:07:19.519
You know, a price, a deadline, a quote, you know, I'm attributing to someone.
00:07:19.680 --> 00:07:23.360
A detail from an earlier conversation that AI is repeating back to me.
00:07:23.519 --> 00:07:33.360
You know, those get a second look every single time, no matter how confident the draft sounds, because confidence was never, you know, the thing that made it trustworthy in the first place.
00:07:33.759 --> 00:07:38.399
This takes far less time than fixing the damage after something wrong goes out.
00:07:38.639 --> 00:07:42.240
A minute, maybe two, you know, on the parts that actually carry risk.
00:07:42.399 --> 00:07:50.160
Everything else, you know, the tone, the structure, the parts where being slightly off costs nothing, I let you know move faster.
00:07:50.399 --> 00:07:53.279
The check isn't about distrusting the tool broadly.
00:07:53.360 --> 00:07:58.480
It's about knowing exactly where a mistake would actually hurt and looking there every time.
00:07:58.639 --> 00:08:03.120
I think of it the same way I think about an employee who's talented but new.
00:08:03.279 --> 00:08:06.160
You don't double check everything they do forever.
00:08:06.399 --> 00:08:10.160
You watch closely on the parts where a mistake would cost you something real.
00:08:10.319 --> 00:08:14.079
And once you see they're solid there, you loosen up everywhere else.
00:08:14.399 --> 00:08:26.959
AI earns that same graduated trust, fast and unchecked on low-stakes stuff, slow and verified on anything that touches money, a client relationship, or your own word.
00:08:27.120 --> 00:08:34.399
And there's a second way this same problem shows up, and it's worth naming separately because it's easier to miss.
00:08:34.639 --> 00:08:38.080
The first version is trusting AI to get the facts right.
00:08:38.240 --> 00:08:52.879
The second version is trusting it to make a decision that was never its decision to make, a price and exception, you know, whether to give a client extra time on something they shouldn't have gotten extra time on, how to respond to someone who's upset.
00:08:53.039 --> 00:08:55.279
You know, those aren't fact-checking problems.
00:08:55.519 --> 00:09:00.960
AI can hand you a perfectly reasonable sounding answer on any of these.
00:09:01.120 --> 00:09:10.000
And reasonable sounding isn't the same as right for your business, you know, your relationship with that specific person or the standard you actually hold.
00:09:10.240 --> 00:09:14.159
I talked in the last episode about writing down your standing decision.
00:09:14.240 --> 00:09:17.600
So AI stops guessing on the routine ones.
00:09:17.840 --> 00:09:29.679
But you know, some decisions aren't supposed to become routine, no matter how many times they come up, the people calls, the moments where your presence and your judgment are the actual point, not a byproduct of the outcome.
00:09:29.919 --> 00:09:36.480
Handing those to AI because it can produce an answer quickly is a different mistake than a wrong date in a summary.
00:09:36.639 --> 00:09:42.799
But it comes from the same route, moving fast because slowing down felt expensive.
00:09:43.039 --> 00:09:48.720
I want to go one layer under this because the real mistake wasn't really about AI at all.
00:09:48.960 --> 00:09:56.399
That real mistake was moving fast on something that mattered because slowing down felt like it was costing me time I didn't have.
00:09:56.559 --> 00:10:00.240
That's not a new failure that AI introduced into my life.
00:10:00.320 --> 00:10:01.279
That's an old one.
00:10:01.519 --> 00:10:07.759
The same one that, you know, it's gotten men like me in trouble long before any of this technology existed.
00:10:08.000 --> 00:10:11.279
Skip the careful step because you know, you're in a hurry.
00:10:11.519 --> 00:10:13.759
And usually you get away with it.
00:10:13.840 --> 00:10:15.120
Then one day you don't.
00:10:15.279 --> 00:10:18.480
And the cost is bigger than the time you thought you were saving.
00:10:18.639 --> 00:10:24.320
I made this exact trade in other parts of my life too, long before AI ever entered the picture.
00:10:24.559 --> 00:10:30.879
Rushing a conversation with Chelsea because I wanted the disagreement over rather than resolved.
00:10:31.120 --> 00:10:36.799
Rushing a decision at work because waiting felt uncomfortable, even when waiting was a wiser move.
00:10:36.960 --> 00:10:38.480
The pattern is always the same.
00:10:38.639 --> 00:10:41.440
Speed feels like progress in the moment.
00:10:41.519 --> 00:10:53.039
And you know, the bill for skipping the careful steps usually arrives later than the decision to skip it, which makes it easy to convince yourself you got away with something you didn't get away with.
00:10:53.200 --> 00:10:54.879
You just haven't got charged yet.
00:10:55.039 --> 00:11:02.639
AI didn't create that instinct in me, it just made it faster to act on, which meant the consequence showed up faster too.
00:11:02.799 --> 00:11:05.519
The tool amplifies whatever habit you already have.
00:11:05.759 --> 00:11:08.639
If your habit is care, it multiplies your care.
00:11:08.799 --> 00:11:13.440
If your habit is rushing past the parts that need attention, it multiplies that instead.
00:11:13.600 --> 00:11:18.960
And it does it quietly because the output looks the same either way until someone finds the mistake.
00:11:19.120 --> 00:11:22.000
I don't think this is only a productivity lesson.
00:11:22.240 --> 00:11:23.759
And I want to say it plainly.
00:11:24.000 --> 00:11:29.919
A man who can't admit when he's got something wrong will keep getting it wrong, with AI or without it.
00:11:30.080 --> 00:11:34.240
Scripture has plenty to say about a man who's quick to speak and slow to check himself.
00:11:34.320 --> 00:11:39.519
And being careless with the truth in something small is still being careless with the truth.
00:11:39.679 --> 00:11:44.559
It doesn't stop mattering just because you know, machine typed the sentence instead of you.
00:11:44.720 --> 00:11:57.279
I think about the men I interviewed over the years who built real things, just large businesses and have done just amazing things in business and still lost what mattered most at home.
00:11:57.519 --> 00:12:00.639
The pattern was rarely one dramatic failure.
00:12:00.799 --> 00:12:14.240
It was a thousand small moments of moving fast where care was required, each one small enough to excuse on its own, until the excuses added up to a life that looked successful and felt hollow.
00:12:14.480 --> 00:12:19.600
This mistake with AI is small by comparison, but the shape of it is the same shape.
00:12:19.679 --> 00:12:28.159
And learning to catch it here in something as low stakes as a summary or a draft is practice for catching it in the places that matter far more.
00:12:28.639 --> 00:12:37.759
Being teachable here looks like this, you know, catching the mistake, naming it plainly instead of quietly fixing it and moving on like it never happened.
00:12:37.919 --> 00:12:41.519
The building the habit that keeps it from happening again.
00:12:41.759 --> 00:12:43.759
That's not a big dramatic moment.
00:12:44.000 --> 00:12:51.600
It's a small, boring discipline, you know, checking the parts that matter before you hit sin over and over until it's just how you work.
00:12:51.759 --> 00:12:58.720
I said in an earlier episode, you know, that the daily heroic act matters more than the heroic moment.
00:12:58.960 --> 00:13:01.039
This is one of those daily acts.
00:13:01.200 --> 00:13:02.720
Nobody sees you do it.
00:13:02.960 --> 00:13:06.399
It just quietly keeps your word worth something.
00:13:06.720 --> 00:13:14.240
Setting up AI right, you know, the work from the last episode and you know, checking its work before you trust it.
00:13:14.480 --> 00:13:19.600
The work from this one, you know, are you know two different disciplines and you need both.
00:13:19.759 --> 00:13:22.960
Good setup means the answers start closer to right.
00:13:23.039 --> 00:13:26.960
You know, checking means you catch it on the days they're not.
00:13:27.200 --> 00:13:29.519
And there'll be days they'll they're not.
00:13:29.679 --> 00:13:36.320
You know, no matter how good your setup is, skip either one and you're exposed somewhere you can't see yet.
00:13:36.559 --> 00:13:44.080
Do both, and you get most of the speed AI offers with almost none of the risk that comes with trusting it blindly.
00:13:44.240 --> 00:13:58.879
And if you haven't built the setup work from the last episode yet, that's worth doing first because a well-trained AI still, you know, needs a man watching its work, but a poorly trained one needs it far more.
00:13:59.200 --> 00:14:07.600
If AI has ever handed you something confidently wrong and you caught it late or worse, didn't catch it at all, you know, come tell me about it.
00:14:07.759 --> 00:14:11.519
Find me on Instagram at the WhitneySool and send me a message.
00:14:11.600 --> 00:14:14.159
Tell me what happened and where it cost you.
00:14:14.320 --> 00:14:21.440
I read them and I'll help you, you know, build the check that would have caught it, even you know, if we never work together.
00:14:21.759 --> 00:14:22.799
Thanks for being here.
00:14:22.960 --> 00:14:26.000
Go build the habit before you need it, not after.
00:00:00.080 --> 00:00:06.080
I've watched AI state something completely wrong with the same tone of voice it uses when it's right.
00:00:06.240 --> 00:00:10.800
No hesitation, no hedge, just confident and false, delivered like a fact.
00:00:10.960 --> 00:00:15.679
If you've used this stuff for more than a week, you already know that feeling.
00:00:15.839 --> 00:00:18.480
The moment you catch it, after you almost didn't.
00:00:18.640 --> 00:00:26.000
A few episodes in now, you know, I've talked about the faith case, the practical jobs, the bottleneck, and the some setup work.
00:00:26.160 --> 00:00:28.719
Today I want to tell you where I got it wrong.
00:00:28.879 --> 00:00:34.560
Because if I only show you what works, I'm not giving you the whole picture and you deserve the whole picture.
00:00:34.799 --> 00:00:37.600
Well, here's something I haven't said outright in this series yet.
00:00:37.759 --> 00:00:45.359
You know, I've let AI produce something I didn't check closely enough, and it came back wrong in a way that mattered.
00:00:45.520 --> 00:00:51.679
Not once, more than once, unfortunately, before I built a habit that fixed it.
00:00:51.840 --> 00:00:56.880
I want to walk through what that looked like, why it kept happening longer than it should have.
00:00:57.039 --> 00:00:59.600
And, you know, one thing I changed to stop it.
00:00:59.759 --> 00:01:00.719
The pattern went like this.
00:01:00.799 --> 00:01:09.200
I'd ask AI to pull something together, you know, a summary, a set of numbers, a draft, you know, that reference details from an earlier conversation.
00:01:09.359 --> 00:01:16.879
And it'd come back clean, you know, well formatted and confident, reading like it knew exactly what it was talking about.
00:01:17.040 --> 00:01:18.319
And sometimes it didn't.
00:01:18.480 --> 00:01:22.400
A date would be off, a number would be close, but not right.
00:01:22.640 --> 00:01:36.959
A detail from you know something I'd told it, you know, three messages earlier would uh come back slightly changed, close enough that I wouldn't have caught it if I hadn't happened to remember the original myself.
00:01:37.200 --> 00:01:39.359
The output never looked uncertain.
00:01:39.439 --> 00:01:43.760
It never said something like, I'm not fully sure about this part.
00:01:44.079 --> 00:01:50.480
It just stated the wrong thing with the same tone it used for you know the parts that were correct.
00:01:50.640 --> 00:01:56.239
And unless you already knew the right answer, you'd have no way to tell the difference by reading it.
00:01:56.480 --> 00:01:58.159
I caught some of those mistakes.
00:01:58.319 --> 00:02:00.159
I didn't catch all of them early enough.
00:02:00.319 --> 00:02:06.000
And the ones I caught late cost me more than the time I saved by using AI in the first place.
00:02:06.159 --> 00:02:08.159
Here's what that actually looks like in practice.
00:02:08.319 --> 00:02:10.319
So this stays concrete instead of vague.
00:02:10.400 --> 00:02:18.000
You know, say you ask AI to summarize a call and pull out the follow-up date you agreed to with a client.
00:02:18.159 --> 00:02:20.159
You know, it gives you back a clean summary.
00:02:20.319 --> 00:02:23.599
Three bullet points, professional, done in 10 seconds.
00:02:23.919 --> 00:02:32.719
Except the date is wrong by week because somewhere in the conversation you mentioned two different possible dates before landing on the real one.
00:02:32.800 --> 00:02:36.639
And it grabbed the wrong one with total confidence.
00:02:36.879 --> 00:02:44.879
You paste that summary into a calendar invite or a follow-up email without you know rereading it closely because it looked finished.
00:02:45.039 --> 00:02:50.960
Now you've told a client the wrong date, and you're the one who has to clean that up, not AI.
00:02:51.199 --> 00:02:52.319
That's a small example.
00:02:52.479 --> 00:02:59.439
Multiply it across every summary, every draft, every you know, quick pull of information you didn't personally verify.
00:02:59.599 --> 00:03:09.199
And you can see how a habit of trusting clean looking output without checking it compounds into something that eventually costs you somewhere that matters.
00:03:09.439 --> 00:03:14.479
I want to explain why you know this happens because understanding it changes how you use the tool.
00:03:14.719 --> 00:03:22.000
AI isn't looking something up the way you'd search for a fact in a book, it's predicting the most likely words based on patterns.
00:03:22.080 --> 00:03:32.960
And most of the time, that prediction lands close to true because true things tend to be more common in pattern and in the data, you know, that it learned from.
00:03:33.280 --> 00:03:36.800
But close to true and true aren't the same thing.
00:03:36.960 --> 00:03:40.080
And when it misses, it doesn't know it missed.
00:03:40.240 --> 00:03:43.039
There's no internal alarm going off.
00:03:43.199 --> 00:03:52.400
It hands you the wrong answer with you know the exact same confidence as the right one because from the inside they feel identical to produce.
00:03:52.639 --> 00:03:55.120
That's the part that catches people off guard.
00:03:55.199 --> 00:04:02.400
We're used to a certain kind of person who hedges when they're unsure, you know, who says something like, I think, or, you know, let me double check that.
00:04:02.560 --> 00:04:07.439
AI doesn't naturally do that unless you specifically ask it to flag uncertainty.
00:04:07.599 --> 00:04:11.439
And even then, it isn't always right about when it should be uncertain.
00:04:11.599 --> 00:04:17.839
You know, confident and correct look the same as confident and wrong right up until you check.
00:04:18.160 --> 00:04:23.519
I found this shows up most on you know the details that feel too small to matter until they do.
00:04:23.680 --> 00:04:28.480
Someone's name, you know, spell it away that looks right, but isn't the way they actually spell it?
00:04:28.639 --> 00:04:38.959
A dollar figure rounded in a direction that seemed harmless, a quote attributed to the wrong person because two names came up close together in the conversation you've had it.
00:04:39.120 --> 00:04:42.639
None of these announce themselves as risky.
00:04:42.879 --> 00:04:49.680
They read exactly like the correct version would read, which is precisely why they get missed.
00:04:50.000 --> 00:04:59.040
This is where I want to be honest about the stakes because you know, I don't think most men fully feel this until it happens to them.
00:04:59.199 --> 00:05:03.920
A wrong number in an internal note costs you a little embarrassment and nothing more.
00:05:04.160 --> 00:05:07.439
A wrong number in something that goes to a client costs you credibility.
00:05:07.519 --> 00:05:10.720
And credibility is slow to rebuild once it cracks.
00:05:10.959 --> 00:05:17.839
A wrong detail in something you say, you know, to your own team gets repeated as if it were true.
00:05:17.920 --> 00:05:21.040
And now the mistake isn't just yours anymore.
00:05:21.199 --> 00:05:22.639
It's loose in your business.
00:05:22.800 --> 00:05:29.519
And if a wrong detail gets in front of the wrong person at the wrong time, it can cost you a relationship you spent years building.
00:05:29.680 --> 00:05:34.639
There's a version of this that's purely financial, and men understand that version fastest.
00:05:34.959 --> 00:05:46.800
A pricing error that goes out in a proposal, you know, a wrong figure in something, you know, you send an investor or a partner, a miscalculation that makes it into a document someone signs.
00:05:46.959 --> 00:05:54.480
Those are the mistakes that get fixed quickly because the cost shows up on a screen where everyone can see it.
00:05:54.720 --> 00:05:57.279
The version that worries me more is quieter.
00:05:57.439 --> 00:06:04.800
It's the erosion of trust that happens when someone catches you being wrong two or three times, you know, on things that should have been simple.
00:06:04.959 --> 00:06:07.040
They don't always say something either.
00:06:07.199 --> 00:06:09.680
They just start double checking everything you send them.
00:06:09.759 --> 00:06:13.040
Or worse, they start assuming you're just careless.
00:06:13.120 --> 00:06:20.240
Uh, and the truth is you got moved by a tool that made carelessness easier to fall into without meaning to.
00:06:20.560 --> 00:06:23.199
You know, that kind of cost doesn't show up on a screen.
00:06:23.360 --> 00:06:35.120
It shows up in the in the tone of the next phone call, the hesitation before someone signs something you sent, you know, the quiet decision to loop in someone else to confirm what you already told them.
00:06:35.360 --> 00:06:40.879
None of that is the AI's fault in any way that matters because AI didn't send the message.
00:06:41.120 --> 00:06:41.680
You did.
00:06:41.920 --> 00:06:49.120
Whatever it produces, the responsibility for what goes out under your name is still yours every time, no exceptions.
00:06:49.279 --> 00:06:50.240
That's not a burden.
00:06:50.319 --> 00:06:52.399
You know, AI removes from you.
00:06:52.639 --> 00:06:54.480
It never was going to be.
00:06:54.639 --> 00:06:55.759
Here's what changed.
00:06:55.920 --> 00:06:59.839
I stopped treating a clean looking output as a finished output.
00:07:00.000 --> 00:07:10.399
Specifically, anything with a number, a date, a name, or a direct reference to something, you know, someone told me, I check against the source before it goes anywhere.
00:07:10.560 --> 00:07:15.360
Not the whole document every time, just the parts where being wrong actually costs something.
00:07:15.519 --> 00:07:19.519
You know, a price, a deadline, a quote, you know, I'm attributing to someone.
00:07:19.680 --> 00:07:23.360
A detail from an earlier conversation that AI is repeating back to me.
00:07:23.519 --> 00:07:33.360
You know, those get a second look every single time, no matter how confident the draft sounds, because confidence was never, you know, the thing that made it trustworthy in the first place.
00:07:33.759 --> 00:07:38.399
This takes far less time than fixing the damage after something wrong goes out.
00:07:38.639 --> 00:07:42.240
A minute, maybe two, you know, on the parts that actually carry risk.
00:07:42.399 --> 00:07:50.160
Everything else, you know, the tone, the structure, the parts where being slightly off costs nothing, I let you know move faster.
00:07:50.399 --> 00:07:53.279
The check isn't about distrusting the tool broadly.
00:07:53.360 --> 00:07:58.480
It's about knowing exactly where a mistake would actually hurt and looking there every time.
00:07:58.639 --> 00:08:03.120
I think of it the same way I think about an employee who's talented but new.
00:08:03.279 --> 00:08:06.160
You don't double check everything they do forever.
00:08:06.399 --> 00:08:10.160
You watch closely on the parts where a mistake would cost you something real.
00:08:10.319 --> 00:08:14.079
And once you see they're solid there, you loosen up everywhere else.
00:08:14.399 --> 00:08:26.959
AI earns that same graduated trust, fast and unchecked on low-stakes stuff, slow and verified on anything that touches money, a client relationship, or your own word.
00:08:27.120 --> 00:08:34.399
And there's a second way this same problem shows up, and it's worth naming separately because it's easier to miss.
00:08:34.639 --> 00:08:38.080
The first version is trusting AI to get the facts right.
00:08:38.240 --> 00:08:52.879
The second version is trusting it to make a decision that was never its decision to make, a price and exception, you know, whether to give a client extra time on something they shouldn't have gotten extra time on, how to respond to someone who's upset.
00:08:53.039 --> 00:08:55.279
You know, those aren't fact-checking problems.
00:08:55.519 --> 00:09:00.960
AI can hand you a perfectly reasonable sounding answer on any of these.
00:09:01.120 --> 00:09:10.000
And reasonable sounding isn't the same as right for your business, you know, your relationship with that specific person or the standard you actually hold.
00:09:10.240 --> 00:09:14.159
I talked in the last episode about writing down your standing decision.
00:09:14.240 --> 00:09:17.600
So AI stops guessing on the routine ones.
00:09:17.840 --> 00:09:29.679
But you know, some decisions aren't supposed to become routine, no matter how many times they come up, the people calls, the moments where your presence and your judgment are the actual point, not a byproduct of the outcome.
00:09:29.919 --> 00:09:36.480
Handing those to AI because it can produce an answer quickly is a different mistake than a wrong date in a summary.
00:09:36.639 --> 00:09:42.799
But it comes from the same route, moving fast because slowing down felt expensive.
00:09:43.039 --> 00:09:48.720
I want to go one layer under this because the real mistake wasn't really about AI at all.
00:09:48.960 --> 00:09:56.399
That real mistake was moving fast on something that mattered because slowing down felt like it was costing me time I didn't have.
00:09:56.559 --> 00:10:00.240
That's not a new failure that AI introduced into my life.
00:10:00.320 --> 00:10:01.279
That's an old one.
00:10:01.519 --> 00:10:07.759
The same one that, you know, it's gotten men like me in trouble long before any of this technology existed.
00:10:08.000 --> 00:10:11.279
Skip the careful step because you know, you're in a hurry.
00:10:11.519 --> 00:10:13.759
And usually you get away with it.
00:10:13.840 --> 00:10:15.120
Then one day you don't.
00:10:15.279 --> 00:10:18.480
And the cost is bigger than the time you thought you were saving.
00:10:18.639 --> 00:10:24.320
I made this exact trade in other parts of my life too, long before AI ever entered the picture.
00:10:24.559 --> 00:10:30.879
Rushing a conversation with Chelsea because I wanted the disagreement over rather than resolved.
00:10:31.120 --> 00:10:36.799
Rushing a decision at work because waiting felt uncomfortable, even when waiting was a wiser move.
00:10:36.960 --> 00:10:38.480
The pattern is always the same.
00:10:38.639 --> 00:10:41.440
Speed feels like progress in the moment.
00:10:41.519 --> 00:10:53.039
And you know, the bill for skipping the careful steps usually arrives later than the decision to skip it, which makes it easy to convince yourself you got away with something you didn't get away with.
00:10:53.200 --> 00:10:54.879
You just haven't got charged yet.
00:10:55.039 --> 00:11:02.639
AI didn't create that instinct in me, it just made it faster to act on, which meant the consequence showed up faster too.
00:11:02.799 --> 00:11:05.519
The tool amplifies whatever habit you already have.
00:11:05.759 --> 00:11:08.639
If your habit is care, it multiplies your care.
00:11:08.799 --> 00:11:13.440
If your habit is rushing past the parts that need attention, it multiplies that instead.
00:11:13.600 --> 00:11:18.960
And it does it quietly because the output looks the same either way until someone finds the mistake.
00:11:19.120 --> 00:11:22.000
I don't think this is only a productivity lesson.
00:11:22.240 --> 00:11:23.759
And I want to say it plainly.
00:11:24.000 --> 00:11:29.919
A man who can't admit when he's got something wrong will keep getting it wrong, with AI or without it.
00:11:30.080 --> 00:11:34.240
Scripture has plenty to say about a man who's quick to speak and slow to check himself.
00:11:34.320 --> 00:11:39.519
And being careless with the truth in something small is still being careless with the truth.
00:11:39.679 --> 00:11:44.559
It doesn't stop mattering just because you know, machine typed the sentence instead of you.
00:11:44.720 --> 00:11:57.279
I think about the men I interviewed over the years who built real things, just large businesses and have done just amazing things in business and still lost what mattered most at home.
00:11:57.519 --> 00:12:00.639
The pattern was rarely one dramatic failure.
00:12:00.799 --> 00:12:14.240
It was a thousand small moments of moving fast where care was required, each one small enough to excuse on its own, until the excuses added up to a life that looked successful and felt hollow.
00:12:14.480 --> 00:12:19.600
This mistake with AI is small by comparison, but the shape of it is the same shape.
00:12:19.679 --> 00:12:28.159
And learning to catch it here in something as low stakes as a summary or a draft is practice for catching it in the places that matter far more.
00:12:28.639 --> 00:12:37.759
Being teachable here looks like this, you know, catching the mistake, naming it plainly instead of quietly fixing it and moving on like it never happened.
00:12:37.919 --> 00:12:41.519
The building the habit that keeps it from happening again.
00:12:41.759 --> 00:12:43.759
That's not a big dramatic moment.
00:12:44.000 --> 00:12:51.600
It's a small, boring discipline, you know, checking the parts that matter before you hit sin over and over until it's just how you work.
00:12:51.759 --> 00:12:58.720
I said in an earlier episode, you know, that the daily heroic act matters more than the heroic moment.
00:12:58.960 --> 00:13:01.039
This is one of those daily acts.
00:13:01.200 --> 00:13:02.720
Nobody sees you do it.
00:13:02.960 --> 00:13:06.399
It just quietly keeps your word worth something.
00:13:06.720 --> 00:13:14.240
Setting up AI right, you know, the work from the last episode and you know, checking its work before you trust it.
00:13:14.480 --> 00:13:19.600
The work from this one, you know, are you know two different disciplines and you need both.
00:13:19.759 --> 00:13:22.960
Good setup means the answers start closer to right.
00:13:23.039 --> 00:13:26.960
You know, checking means you catch it on the days they're not.
00:13:27.200 --> 00:13:29.519
And there'll be days they'll they're not.
00:13:29.679 --> 00:13:36.320
You know, no matter how good your setup is, skip either one and you're exposed somewhere you can't see yet.
00:13:36.559 --> 00:13:44.080
Do both, and you get most of the speed AI offers with almost none of the risk that comes with trusting it blindly.
00:13:44.240 --> 00:13:58.879
And if you haven't built the setup work from the last episode yet, that's worth doing first because a well-trained AI still, you know, needs a man watching its work, but a poorly trained one needs it far more.
00:13:59.200 --> 00:14:07.600
If AI has ever handed you something confidently wrong and you caught it late or worse, didn't catch it at all, you know, come tell me about it.
00:14:07.759 --> 00:14:11.519
Find me on Instagram at the WhitneySool and send me a message.
00:14:11.600 --> 00:14:14.159
Tell me what happened and where it cost you.
00:14:14.320 --> 00:14:21.440
I read them and I'll help you, you know, build the check that would have caught it, even you know, if we never work together.
00:14:21.759 --> 00:14:22.799
Thanks for being here.
00:14:22.960 --> 00:14:26.000
Go build the habit before you need it, not after.
