WEBVTT
00:00:01.600 --> 00:00:04.000
Everything is fake.
00:00:04.559 --> 00:00:14.320
The text you read on that landing page, that amazing photo you saw on Instagram, the viral video on Twitter, and all of those comments.
00:00:14.560 --> 00:00:21.199
That customer service rep you talked to last week that was surprisingly chipper and actually helpful.
00:00:21.440 --> 00:00:25.519
There's a good chance all of it was fake, or at least AI.
00:00:25.679 --> 00:00:38.079
And I think we're about to cross over into a very dangerous place where the majority of our day-to-day interactions will be engaging with some type of media or medium that's either partially AI generated or completely AI.
00:00:38.640 --> 00:00:43.359
Yes, there's the fast emerging threat of deep fakes and fraud that's not going to go away.
00:00:43.520 --> 00:00:55.119
And while I think there's a whole other episode to be done on the impacts of the everything is fake disease that has on us as individuals, there's probably a more important chip that you have to write right now.
00:00:55.359 --> 00:01:00.320
That's how your company actually uses AI, but remains human.
00:01:00.640 --> 00:01:11.359
See, I'm someone that encourages AI use like all day, every day, but I know most people and most companies don't take the proper care in elevating human expertise while using AI.
00:01:11.519 --> 00:01:22.319
And that's led to this onslaught of work slop in 2026 or the never-ending barrage of generic sounding, uninspiring outputs that your company may be rubber stamping.
00:01:22.560 --> 00:01:25.680
So, how can you fight back against the year of the fake?
00:01:25.840 --> 00:01:29.840
How can your company resonate in a sea full of mundane work slop?
00:01:29.920 --> 00:01:37.200
And how can you make sure the expertise of your brightest people doesn't get drowned out by agentifying all of your business processes?
00:01:37.439 --> 00:01:43.439
Well, we're going to talk about that on today's episode of Everyday AI in our Start Here series.
00:01:49.920 --> 00:01:52.239
So here's the big picture.
00:01:52.799 --> 00:01:56.400
Right now, there is an AI trust crisis.
00:01:56.560 --> 00:01:57.599
It's already here.
00:01:58.159 --> 00:02:11.439
So a recent study from Europool uh sorry, Europol projects that up to 90% of online content may be synthetically generated by the end of this year.
00:02:11.680 --> 00:02:15.680
And that means that trust is essentially gone.
00:02:16.240 --> 00:02:25.680
A new Salesforce state of the AI connected customer survey said that 72% of consumers trust companies less than they did a year ago.
00:02:26.000 --> 00:02:34.319
So if almost everything that we see or read online is going to be AI and people don't trust companies, what's the answer?
00:02:34.560 --> 00:02:39.280
Well, it's elevating your human expertise while still using AI.
00:02:39.599 --> 00:02:42.400
And that's what we're going to talk about on today's show.
00:02:42.800 --> 00:02:46.319
So if you stick around for the next 20 or so minutes, here's what you're going to learn.
00:02:46.400 --> 00:02:53.759
You're going to learn why defaulting to everything is AI generated is now the sound professional discipline and not cynicism.
00:02:54.080 --> 00:03:00.960
You're going to learn how AI works slop is silently destroying trust and revenue before any dashboard shows it.
00:03:01.120 --> 00:03:08.080
And you're going to learn what the most AI native companies do to stay human, authentic, and competitive at scale.
00:03:08.479 --> 00:03:09.759
All right, let's get into it.
00:03:09.840 --> 00:03:12.879
Welcome to Everyday AI and our start here series.
00:03:13.039 --> 00:03:22.479
This, if you're new, well, it's the essential podcast series to both learn the basics of AI, but if you are a regular, to double down your knowledge.
00:03:22.879 --> 00:03:31.120
So uh after doing 700 plus episodes of everyday AI, I didn't have a good answer when people always said, like, hey Jordan, you have a lot of information out there.
00:03:31.280 --> 00:03:31.919
Where do I start?
00:03:32.000 --> 00:03:33.840
Well, you start here with the Start Here series.
00:03:33.919 --> 00:03:36.560
I think it's best if you listen to all of these in order.
00:03:36.719 --> 00:03:40.560
They're shorter episodes, usually about 25 to 30 minutes.
00:03:40.719 --> 00:03:46.319
But if you want to catch up on all of them in order, I suggest you go to startheseries.com.
00:03:46.479 --> 00:03:52.639
That is going to give you free access to our inner circle community and in our start here series uh space.
00:03:52.800 --> 00:03:56.240
You can go and listen to all of the episodes in order, read about them.
00:03:56.319 --> 00:03:59.439
There's a playlist that we keep updated there and everything else.
00:03:59.680 --> 00:04:07.199
So if you missed our last episode, we went over the AI labor shift, when it will happen, and what it means for chops.
00:04:07.680 --> 00:04:17.279
So let's talk about why everything is fake and why I think this is actually one of the biggest problems that most people don't know that we're fighting against.
00:04:17.839 --> 00:04:25.680
So again, when I tell you this, I'm not trying to say this is some, you know, oh, like bragging about something.
00:04:25.839 --> 00:04:39.519
Uh in many instances, because I am drowning in AI every day, um, I feel that I pick up on certain trends and topics, maybe a little bit for the before the average, you know, business user.
00:04:39.680 --> 00:04:44.399
And what I've seen over the last six months is this onslaught of work slot, right?
00:04:44.560 --> 00:04:50.000
Um, and I don't know if the average everyday business consumer has realized this.
00:04:50.160 --> 00:04:50.879
Maybe you have.
00:04:51.040 --> 00:04:56.800
If you spend any time on, you know, LinkedIn, Twitter, you've probably already seen this.
00:04:56.879 --> 00:05:01.600
Uh so I think it's already impacted the written word, uh, but I think it's gonna impact everything, right?
00:05:01.680 --> 00:05:08.079
So as vibe coding uh becomes very commonplace, it's gonna impact, you know, all of the apps we use.
00:05:08.160 --> 00:05:11.680
There's gonna be app slop, uh, there's gonna be video slop, right?
00:05:11.759 --> 00:05:18.399
Uh, but essentially, I think the smart business consumer needs to realize that this is just the future.
00:05:18.560 --> 00:05:21.279
And there are things that we can do to fight against it.
00:05:21.439 --> 00:05:22.959
So let's start here.
00:05:24.079 --> 00:05:32.399
Right now, no one really knows when or if something is AI generated, which is maybe a good thing, right?
00:05:32.480 --> 00:05:35.600
If you're thinking, oh, well, I can use AI at scale.
00:05:35.839 --> 00:05:36.639
Yes, you can.
00:05:36.800 --> 00:05:51.360
So when something is done correctly, it is very hard uh to tell the difference between uh, you know, an expert who is using AI correctly at scale versus a team of maybe 20 humans who aren't using AI, uh, right.
00:05:51.439 --> 00:05:56.959
As a former uh, you know, journalist, I was a Pulitzer uh fellow, right?
00:05:57.120 --> 00:05:59.279
I won the ACP Story of the Year.
00:05:59.439 --> 00:06:03.519
Um, so I was a pretty decent journalist way back in the day, pretty good at writing, right?
00:06:03.600 --> 00:06:09.120
I I think writing, we've already far surpassed what the average large language model can do.
00:06:09.279 --> 00:06:10.800
But there's a big gap there.
00:06:10.959 --> 00:06:13.040
Because when I say that, a lot of people won't believe me.
00:06:13.120 --> 00:06:19.439
And they're gonna be like, no, what comes out of you know, Chat GPT or Gemini or Claude or, you know, Copilot, whatever, is kind of garbage.
00:06:19.519 --> 00:06:20.720
No, it's it's not.
00:06:20.879 --> 00:06:26.160
It's better than what award-winning writers like myself can do if you know what you're doing.
00:06:26.240 --> 00:06:27.199
But that's a big if.
00:06:27.279 --> 00:06:30.240
And that is elevating the human expertise, uh, right.
00:06:30.319 --> 00:06:40.079
Because if someone goes in and doesn't put a lot of care, they don't understand the basics of context engineering or understand how large language models work, they're not going to be able to produce something that's economically valuable.
00:06:40.160 --> 00:06:41.519
They're gonna produce work slop.
00:06:41.600 --> 00:06:49.519
So that is the uh generic output that you get from you know, just trying to either take the shortest way out or try to get the quickest output.
00:06:49.680 --> 00:07:01.680
Uh, but right now, humans can't tell the difference between AI content that's produced at by an expert and actual, you know, expert uh created good human-only AI content.
00:07:01.759 --> 00:07:12.480
Uh, in a study from Beringa last year showed that people claiming to confidently be able to spot AI images only scored about 30% accuracy.
00:07:12.639 --> 00:07:18.959
So, yeah, uh, even when people are like, oh, I'm sure that this is or is not an AI image, no.
00:07:19.120 --> 00:07:27.199
Uh also self-reported confidence in AI detection is rising every year while actual accuracy keeps just uh declining.
00:07:27.759 --> 00:07:33.120
And this is not something that your team can just solve with better instincts or better tools.
00:07:33.279 --> 00:07:34.560
So, what do I mean by this?
00:07:34.720 --> 00:07:37.680
The quality of what AI can produce.
00:07:38.160 --> 00:07:42.399
I keep, I've I've been saying this over and over the last couple of months.
00:07:42.720 --> 00:07:52.000
The quality and the the just scale that we've seen in the last three months has far surpassed what we got the three years prior.
00:07:52.160 --> 00:07:57.759
And that's what I think has gotten us to this crisis of, well, everything's fake, because you can't tell, right?
00:07:57.920 --> 00:08:08.319
If someone knows what they're doing uh in, you know, Nano Banana Pro or you know, Seed Dance for video generation, Vio for uh right, all of these platforms.
00:08:08.480 --> 00:08:23.120
If you're using the best platform for image editing, video editing, text generation, web, right, for software engineering, agentic coding, the output ultimately, if you have an expert driving it, is pretty much indistinguishable, which is crazy to say that.
00:08:23.279 --> 00:08:24.639
Even video, right?
00:08:24.720 --> 00:08:25.360
We never thought.
00:08:25.439 --> 00:08:34.000
We thought, you know, back when we in the early Dolly one, Dolly two days, you would have said, oh, you know, AI video is 20 years away, right?
00:08:34.159 --> 00:08:38.240
But here we are, you know, five years after that point in time.
00:08:38.320 --> 00:08:46.240
And well, no, it's already, you know, gotten to the point of human level again if you have an expert human driving it.
00:08:47.679 --> 00:08:53.200
So when everything is fake, you can't look over fraud.
00:08:53.360 --> 00:08:54.559
That's only one concern.
00:08:54.639 --> 00:08:56.159
But I think you have to understand this.
00:08:56.320 --> 00:08:59.440
So I am speaking to this from a couple of different ways.
00:08:59.679 --> 00:09:08.000
Um, for consumers, I think today's show may be helpful that you need to change your mindset and assume literally everything is fake.
00:09:08.080 --> 00:09:09.200
Everything is AI, right?
00:09:09.360 --> 00:09:12.000
I've gotten a lot more comments recently, right?
00:09:12.080 --> 00:09:16.159
That uh, you know, this podcast or myself, uh AI, no.
00:09:16.320 --> 00:09:21.600
Uh I've literally had a cold for off and on for like three months, yeah, right.
00:09:21.679 --> 00:09:23.120
But I've gotten all these comments recently.
00:09:23.279 --> 00:09:25.519
It's like, oh no, this is this is no.
00:09:25.600 --> 00:09:37.600
It's uh I don't think AI can uh quite ramble on like I do, uh, not yet, or uh yet simulate my stuffy nose or horsey throat, right?
00:09:37.679 --> 00:09:42.000
So no, uh if if you're uh listening on the podcast, this is not AI.
00:09:42.080 --> 00:09:45.120
I'm a real human doing this live and unedited, unscripted.
00:09:45.200 --> 00:09:48.080
My voice has just been gone off and on for three months.
00:09:48.159 --> 00:09:51.120
But I think it's a good thing for consumers to have, right?
00:09:51.200 --> 00:09:55.120
When you are having that conversation with uh someone over the phone, right?
00:09:55.200 --> 00:10:02.639
When you're watching something or, you know, probably the biggest thing that no one's talking about, just the right agentic AI, right?
00:10:02.720 --> 00:10:05.600
With all these, you know, the open claw variants out there.
00:10:05.759 --> 00:10:08.639
Probably anything you read online, even discourse, right?
00:10:08.799 --> 00:10:19.120
People debating things on, you know, Twitter, LinkedIn, Reddit, Quora, it's probably mostly going to be bot driven, I would say, uh, within a couple of months.
00:10:19.200 --> 00:10:20.799
And people aren't going to understand that.
00:10:21.039 --> 00:10:27.600
So, yes, I think this is important for consumers to understand that pretty much anything you interact with will probably be fake.
00:10:27.759 --> 00:10:31.759
But I think it's also as business leaders, I think it's important to understand, right?
00:10:31.840 --> 00:10:35.440
That that email you got from a client, is it real or is it not?
00:10:35.600 --> 00:10:36.000
Right.
00:10:36.159 --> 00:10:39.120
Um, you know, pretty, pretty soon, right?
00:10:39.279 --> 00:10:40.639
Uh, I'm I'm getting this all the time.
00:10:40.799 --> 00:10:43.120
I'm getting emails from AI agents.
00:10:43.279 --> 00:10:45.120
Sometimes they disclose it, sometimes they don't.
00:10:45.200 --> 00:10:51.519
I've gotten video pitches from, you know, clearly things that are AI and people think that I maybe I can't tell the difference.
00:10:51.600 --> 00:10:52.080
Uh, right.
00:10:52.240 --> 00:10:59.759
So, yes, this does lead to problems on the consumer side and on the business decision maker side, but also fraud, right?
00:10:59.840 --> 00:11:03.120
Company wide, I think you have to be uh paying attention to this.
00:11:03.360 --> 00:11:14.320
So uh in the experienced future of fraud forecast that came out this year, 72% of business leaders identified AI-enabled fraud as a top operational challenge.
00:11:14.639 --> 00:11:23.120
Uh, I think maybe two years ago, I think most people understood that AI was was a huge threat, right?
00:11:23.200 --> 00:11:30.159
Because what you could do with with deep fakes, uh things like that, voice cloning, um, obviously a huge concern.
00:11:30.240 --> 00:11:36.000
But now the barrier to do this, the technical barrier in the technical know-how is essentially zero.
00:11:36.159 --> 00:11:42.879
Uh, you know, anyone with that can read within 30 minutes could probably figure out how to do this at scale.
00:11:43.039 --> 00:11:47.120
Voice cloning, uh, you know, AI videos that look very real.
00:11:47.360 --> 00:11:49.120
Um, it's it's very easy, right?
00:11:49.200 --> 00:11:50.639
These fraud tools are now free.
00:11:50.720 --> 00:11:57.360
They don't require any technical skill, then they allow complete autonomy and anonymity as well.
00:11:57.679 --> 00:12:04.639
So anything that you is important for your day-to-day business operations.
00:12:04.720 --> 00:12:12.480
So vendor proposals, you know, job candidates, uh, executive communications, they can no longer just be verified by site alone.
00:12:12.639 --> 00:12:20.720
I think it's important that we come to that realization that, you know, these routines that we've gone through throughout the years, oh, checking email, right?
00:12:20.879 --> 00:12:21.440
Hiring.
00:12:21.679 --> 00:12:24.240
Oh, yes, looking at these resumes, right?
00:12:24.399 --> 00:12:26.559
Probably we've already seen this over the past two years.
00:12:26.720 --> 00:12:32.080
Now you assume every single resume is written by, you know, ChatGPT or Claude or Gemini or something, right?
00:12:32.159 --> 00:12:38.159
But the same thing with videos uh as well, audio, you know, interviews, similar.
00:12:39.759 --> 00:12:43.440
There's a flip side to that, to this deep fake issue, right?
00:12:43.519 --> 00:12:49.600
And this is again, maybe good conversation for another uh episode where we can go deeper on this.
00:12:49.759 --> 00:12:52.879
But even proof of things that actually happen.
00:12:53.200 --> 00:12:55.440
I think that's gonna be hard to prove, right?
00:12:55.519 --> 00:12:57.360
That's called the liar's dividend.
00:12:57.440 --> 00:13:01.360
That's when real data and real expertise gets lumped in with AI spam, right?
00:13:01.440 --> 00:13:08.799
Kind of how, you know, people have been saying, oh, Jordan, you're, you know, because my voice has been going in and out and my audio is a little weird because of that.
00:13:08.960 --> 00:13:11.279
People are saying, oh, no, you're AI, right?
00:13:11.360 --> 00:13:12.639
No, that's the liar's dividend.
00:13:12.799 --> 00:13:15.600
But I think it's important to realize and understand as well.
00:13:15.840 --> 00:13:18.879
And and because of AI content detection, right?
00:13:18.960 --> 00:13:20.480
Which is not real, right?
00:13:20.559 --> 00:13:35.120
There's obviously things that are a little bit better and a little more reliable on the video and the image side, uh, you know, audio as well in terms of invisible uh watermarks in some type of media, but in written media, it's there's no such thing, right?
00:13:35.279 --> 00:13:44.960
Uh so when you have AI content detection, you know, coming out and say, hey, you know, your company's proposal, you know, couldn't be accepted for this RFP because it was AI generated, right?
00:13:45.120 --> 00:13:46.000
Well, no.
00:13:46.159 --> 00:13:55.759
Uh you know, and that's a problem because I think that your company's real expertise can actually be dismissed as AI generated, uh, just because there's zero recourse or proof.
00:13:56.879 --> 00:13:59.200
And that's led to this trust crisis.
00:13:59.279 --> 00:14:03.519
That if you don't know it's happening right now, it is, right?
00:14:03.679 --> 00:14:16.879
So, like I said, that Salesforce study that said that 72% of people trust companies less this year than they did last year, that is a drastic drop.
00:14:17.519 --> 00:14:18.000
All right.
00:14:18.240 --> 00:14:25.279
And that has led to, I think, or is the result of maybe work slop.
00:14:25.600 --> 00:14:25.840
Right.
00:14:26.159 --> 00:14:30.159
Um, I think consumers are trusting companies less.
00:14:31.200 --> 00:14:40.960
Number one, if you are putting out low effort content at scale, which I think so many, uh, so many people are.
00:14:41.120 --> 00:14:43.360
And that has led to uh work slot.
00:14:43.440 --> 00:14:54.159
But number two, if if if you're not putting a a human face, if you're not putting imperfections out in the world, that's one of the reasons all along I've wanted to, right?
00:14:54.240 --> 00:15:00.240
Uh before I even started everyday AI more than three years ago, I said, I'm not gonna edit this thing.
00:15:00.399 --> 00:15:09.519
I said, I'm gonna go on, I'm gonna do this thing live because I knew at the time, I'm like, in a couple of years, you know, this is all gonna be AI generated.
00:15:09.600 --> 00:15:14.159
So I'm gonna get my imperfections, my, my stuttering, my sniffing, all of it.
00:15:14.240 --> 00:15:16.159
Like that's that's human, that's authenticity.
00:15:16.240 --> 00:15:23.840
And I think that's one of the reasons why, you know, this this podcast has, you know, kind of grown in popularity over the years because it's real, it's authentic.
00:15:23.919 --> 00:15:27.440
And I think that brands need to understand that too, because of the trust crisis, right?
00:15:27.519 --> 00:15:33.519
So, even good example, three major dictionaries named AI Slop their word of the year in 2025.
00:15:33.679 --> 00:15:34.080
Uh, right.
00:15:34.159 --> 00:15:38.159
And work slop, I think, is going to uh trend in that direction.
00:15:38.320 --> 00:15:41.679
If you haven't heard of work slot before, right?
00:15:41.759 --> 00:15:47.519
That's just AI output that's technically competent, but it carries like no domain expertise.
00:15:47.759 --> 00:15:50.080
It just sounds like bland, right?
00:15:50.480 --> 00:15:53.919
And and social media is already a graveyard of enterprise work slot.
00:15:54.000 --> 00:15:58.159
Most companies don't even realize that they are actually contributing to this.
00:16:00.000 --> 00:16:02.320
And here's three reasons.
00:16:02.480 --> 00:16:06.799
I think that companies are still falling into the work slop trap.
00:16:07.039 --> 00:16:11.919
And again, we're gonna get to how you can overcome this uh with human expertise.
00:16:12.080 --> 00:16:16.159
But the three reasons are education, cost, and accuracy.
00:16:16.399 --> 00:16:17.679
All right, so education.
00:16:17.840 --> 00:16:26.320
So uh recent McKinsey study said that only six percent of companies qualify uh as AI high performers, generating mean of meaningful business impact.
00:16:26.639 --> 00:16:38.879
Uh, what that is, is everyone's using AI, but how many people out there have actually learned, have actually taken courses, are actually taught, right?
00:16:39.039 --> 00:16:41.039
We use large language models every day.
00:16:41.519 --> 00:16:45.120
Does anyone know top K, right?
00:16:45.360 --> 00:16:47.440
Top P temperature, right?
00:16:47.679 --> 00:16:52.320
Not that those things are super important, but you should at least know what those things are, right?
00:16:52.399 --> 00:16:55.120
For the things that we're using every single day.
00:16:55.279 --> 00:16:56.559
This is the equivalent.
00:16:56.720 --> 00:17:11.440
This is the like 2010 equivalent of you know, sitting in front of a computer as most knowledge workers do in 2010 and not knowing what an email was, or not knowing uh, I don't know, what a URL was.
00:17:11.599 --> 00:17:13.839
It's like, no, you know what a URL is.
00:17:13.920 --> 00:17:17.039
That's the thing you put in the bar, you type it, and you click the button, right?
00:17:17.200 --> 00:17:19.119
It's it's it's how it operates and how it works.
00:17:19.279 --> 00:17:21.119
Anyways, education is the first reason.
00:17:21.279 --> 00:17:22.319
Number two, cost.
00:17:23.440 --> 00:17:27.039
The economics of AI are irresistible and undeniable.
00:17:27.200 --> 00:17:35.279
AI generates like quality, usually or passable outputs in seconds at a fraction of the human cost.
00:17:35.759 --> 00:17:45.680
Uh, and then accuracy, I think is the third reason that we have so much work slop, because so much of the day-to-day deliverables fall into a gray area, right?
00:17:46.000 --> 00:17:48.799
I think good enough passes for good.
00:17:49.440 --> 00:17:50.000
All right.
00:17:50.160 --> 00:17:51.279
Let me say that again.
00:17:51.440 --> 00:17:59.680
And how many times, if you're being honest with yourself, if you've just blindly used AI output and you've looked at something, how many times have you said good enough?
00:18:00.160 --> 00:18:00.480
Right?
00:18:00.640 --> 00:18:08.640
Whereas before AI, when when when you were doing it manually, I don't think good enough would ever pass for good.
00:18:08.880 --> 00:18:09.680
Now it does.
00:18:10.400 --> 00:18:13.519
In an age of AI, right, where you're like, oh, that's good enough.
00:18:14.960 --> 00:18:29.680
So we've talked about so far why uh work slop is on the rise, the impacts uh maybe personally on consumers, uh, fraud, deep fake.
00:18:30.160 --> 00:18:38.319
But here's the cure it's elevating human expertise inside of your AI setup.
00:18:40.079 --> 00:18:44.079
You need strategic oversight from experts.
00:18:44.480 --> 00:18:54.880
So a Smith OS uh report found that AI content with human strategic oversight performs more than four times better than fully automated outputs out of AI.
00:18:55.279 --> 00:18:55.759
Right?
00:18:56.000 --> 00:19:02.240
This is what makes AI decision makers click yes or no on that proposal.
00:19:03.519 --> 00:19:21.759
Four times higher, a four times better result when you properly put in human expertise, when you properly elevate those domain experts at the right place at the right time in your AI workflows, which unfortunately isn't being done, right?
00:19:21.920 --> 00:19:33.680
Because I think when AI made content creation nearly free, I think don't that instantly elevated domain knowledge to be the competitive moat overnight.
00:19:33.920 --> 00:19:47.039
And I'm talking about basic text-to-text large language models all the way through fully autonomous uh you know, multi-agent orchestration, everything.
00:19:48.319 --> 00:20:06.160
It's elevating the right domain expert at the right time in the process because the winners right now, those that are getting the most out of it, are pouring so much human expertise into their AI that the outputs are unmistakably theirs alone, right?
00:20:07.440 --> 00:20:14.240
And if you listen in our start here series, we talked about a little bit of this in context in the context engineering episode.
00:20:14.640 --> 00:20:19.599
Uh, but I think it's worth maybe repeating one or two things from that.
00:20:22.000 --> 00:20:35.759
It's all about making sure that you have the expert lined up with whoever is building whatever AI flows.
00:20:36.079 --> 00:20:37.119
Here's what I mean by that.
00:20:37.839 --> 00:20:49.359
I think so many times in so many organizations that I talk to, I say, okay, you have this great you know, scheduled AI flow, you know, it's great, it's hands off, cool, great.
00:20:49.519 --> 00:20:50.640
Who set it up?
00:20:50.880 --> 00:20:54.160
It's always usually someone technical, right?
00:20:54.400 --> 00:20:58.960
Usually maybe companies have an AI champion or two, maybe it's someone in IT.
00:20:59.119 --> 00:21:09.599
I think I see it more on the IT side for uh, you know, Windows co-pilot enterprise organizations, uh, but usually it's someone who's technical or an AI champion, but who's actually benefiting from it?
00:21:09.759 --> 00:21:13.359
Who's going in and you know, copying and pasting and updating it?
00:21:13.519 --> 00:21:15.759
It's usually not that person, right?
00:21:16.000 --> 00:21:37.759
I would venture in larger organizations, uh, I would say less than 10% of the output that ultimately gets used uh in a deliverable, in an end artifact, is coming from someone with domain expertise, with very little or zero input in the front end or the back end.
00:21:37.839 --> 00:21:39.680
And that's where it's the most important.
00:21:39.839 --> 00:21:46.160
It's usually a technical person who's setting it up or an AI champion or something, you know, oh, we found this on the internet.
00:21:46.319 --> 00:21:48.960
This looked good, let's plug it in, and it's good enough.
00:21:49.279 --> 00:21:53.920
But the winners are the ones that are pouring more human expertise.
00:21:54.640 --> 00:21:59.759
Because, like we talked about in our context engineering episode, generic prompts produce generic options.
00:22:00.079 --> 00:22:08.799
Output in your proprietary data, your first person or first company reasoning decision logic that changes everything.
00:22:08.960 --> 00:22:17.440
And you need to start, if you haven't already, you need to document why your company makes decisions, not just what decisions it makes.
00:22:17.599 --> 00:22:22.400
In the same way, I talk so much, and I know you guys are probably annoyed at me saying this.
00:22:22.720 --> 00:22:36.880
You need to spend more time looking at the chain of thought in a large language model than you do on your front end context engineering or whatever you do on the back end, copying, pacing, producing a document, right?
00:22:36.960 --> 00:22:38.880
You need to be able to iterate multiple times.
00:22:39.039 --> 00:22:43.839
That is how you properly insert your company's domain expertise.
00:22:44.000 --> 00:22:46.400
Uh, it's through the proper context engineering process.
00:22:46.559 --> 00:22:53.279
And if you took our, you know, prime prompt polish PPP course uh inside our inner circle community, you already know this.
00:22:53.359 --> 00:23:02.240
But that's where you insert your domain expertise is going through how a certain large language model uh will tackle a certain issue.
00:23:02.319 --> 00:23:07.599
And you have to be using the right model, the right feature, the right mode uh for the right problem, right?
00:23:07.759 --> 00:23:11.119
And that might look a little different depending on what sector you're working in.
00:23:11.359 --> 00:23:13.759
But the chain of thought and sitting down, right?
00:23:13.920 --> 00:23:21.440
Even if you're uh, you know, if you have a certain setup where, hey, someone from IT or I don't know, someone from marketing has to set up a certain thing.
00:23:21.519 --> 00:23:30.799
Well, they need to be sitting down with the actual domain expert, looking at the chain of thought and saying, hey, according to our SOP, according to our skills that we set up, right?
00:23:30.960 --> 00:23:35.119
Um, this is what the input is on the context engineering side.
00:23:35.279 --> 00:23:36.319
This is what the output is.
00:23:36.480 --> 00:23:43.680
But more importantly, let's walk through the chain of thought and let's see how the model tackled this problem, right?
00:23:44.480 --> 00:23:52.000
And documenting the why, that is one of the most powerful things that companies that are winning right now, that's why.
00:23:52.160 --> 00:23:59.119
And that's just context engineering reframed as a trust strategy, not just for productivity improvement.
00:23:59.599 --> 00:24:03.200
Is that going to make your company more productive doing that?
00:24:03.440 --> 00:24:04.960
Absolutely, right?
00:24:05.279 --> 00:24:10.160
But that is ultimately going to create more trust in the end.
00:24:10.319 --> 00:24:13.519
That is going to be that that process right there, right?
00:24:13.680 --> 00:24:19.759
Proper proper context engineering, using the right model, you know, education, training, all that.
00:24:19.920 --> 00:24:35.359
But having your domain expert be involved in the auditing, in the iteration, looking at the chain of thought, that what is going to ultimately lead to less work slop, more trust, and more ROI for your company.
00:24:36.400 --> 00:24:43.200
And this is again why I will continue to call out how bad human in the loop is.
00:24:43.440 --> 00:24:43.759
Right.
00:24:43.920 --> 00:24:50.799
I've been saying this, sorry, uh way before everyone else about how human in the loop is bad.
00:24:50.960 --> 00:25:01.279
It's one of the worst things, I think, that has happened in or around the AI industry, uh, aside from just an overall general lack of education, training, and uh basic knowledge.
00:25:01.519 --> 00:25:03.920
Human in the loop is bad.
00:25:04.240 --> 00:25:07.039
Human in the loop leads to work slop.
00:25:07.599 --> 00:25:16.880
Human in the loop is going to uh continue to contribute to the everything is fake dilemma, and it is going to be bad.
00:25:17.359 --> 00:25:23.119
We need, as again, if you are a longtime listener of this podcast, you know maybe it's the first time listening.
00:25:23.279 --> 00:25:24.720
Well, let me introduce this to you.
00:25:24.960 --> 00:25:27.359
Expert-driven loops are what we need.
00:25:27.599 --> 00:25:34.960
Human in the loop means that you can just put any human passively into any AI system, and that's some sort of guardrail.
00:25:35.039 --> 00:25:36.160
That's garbage.
00:25:36.240 --> 00:25:37.519
Uh, that's failure.
00:25:37.680 --> 00:25:38.960
That is work slot.
00:25:39.279 --> 00:25:40.880
You need expert-driven loops.
00:25:40.960 --> 00:25:47.359
That means the domain uh expertise, the domain expert driving the loop, right?
00:25:47.440 --> 00:25:48.720
And what does that mean?
00:25:48.960 --> 00:25:59.359
Usually when I talk about when anyone talks about uh a loop, we're talking about an agentic or semi-agentic or AI-powered workflow.
00:25:59.599 --> 00:26:05.200
Something that happens in your company repeatedly, uh, right, over and over again.
00:26:05.359 --> 00:26:06.640
That's that's a loop.
00:26:06.799 --> 00:26:11.519
You need an expert driving it proactively, right?
00:26:11.680 --> 00:26:19.440
Because Gardner right now predicts that companies that replaced human agents with generic AI will be forced to rehire by 2028.
00:26:19.680 --> 00:26:22.720
And I think human in the loop is just a passive checkpoint.
00:26:22.799 --> 00:26:29.519
But expert-driven loops mean experts are actively shaping the context and reviewing against professional standards.
00:26:29.759 --> 00:26:30.160
All right.
00:26:30.240 --> 00:26:44.720
This is why I think it's so important, and we've talked about this in previous episodes, to have internal company benchmarks and internal scoping that you are doing not yearly, not quarterly, but probably monthly or bi-monthly.
00:26:46.480 --> 00:26:52.400
So work slop is only going to get worse and more prevalent.
00:26:52.880 --> 00:26:58.880
And I think this is important because we are gonna have AI slop-heavy training data.
00:26:59.200 --> 00:27:03.519
This is something I think uh people very much overlook.
00:27:03.680 --> 00:27:15.920
So, right, if we go back uh to one of our initial stats from Europol that projected that up to 90% of online content may be synthetically generated by the end of this year.
00:27:16.000 --> 00:27:20.240
Uh, and I would say it's probably gonna be more than 90%, if I'm being honest.
00:27:20.400 --> 00:27:22.799
Um, but what does that create?
00:27:23.119 --> 00:27:29.920
I think that creates trust in AI that we probably shouldn't have.
00:27:30.079 --> 00:27:39.440
Uh, because even if you just looked at the relatively quality, like the relative quality, I don't know if there's a a a standard for this or a benchmark.
00:27:39.680 --> 00:27:41.839
If not, I hope someone can create it.
00:27:42.000 --> 00:27:42.480
Uh, right.
00:27:42.640 --> 00:27:52.079
But if you looked at the relative quality of something that was produced on the internet in 20, uh, I don't know, we'll say pre-uh AI, right?
00:27:52.160 --> 00:27:56.880
Because even before Chat GPT, uh, you know, AI slop was a huge problem.
00:27:57.039 --> 00:27:57.519
Uh, right.
00:27:57.599 --> 00:28:09.759
The early GPT technologies that uh predated uh ChatGPT were, you know, already out there polluting the internet with garbage with AI work slot before anyone knew ChatGPT existed.
00:28:09.839 --> 00:28:18.960
Uh, but I think now, right, the average piece of, I don't know, we'll just say blog posts, right?
00:28:19.119 --> 00:28:22.079
Uh, because that's a lot of what goes into trading data.
00:28:22.319 --> 00:28:28.720
The average quality of a blog post, I would say has gone down by two to three X very easily over the last 10 years.
00:28:28.799 --> 00:28:30.720
And it's because of AI, right?
00:28:30.880 --> 00:28:39.359
So that just obviously creates this regurgitated cycle of poor or poorer quality source materials.
00:28:39.519 --> 00:28:47.839
So um, I do think that even more so in the future, yes, the training uh at the companies is getting better, right?
00:28:47.920 --> 00:28:59.759
The reinforcement learning with human feedback and other uh scaling technologies at the big AI Frontier Labs helps with this, you know, helps to make sure that the training data is uh higher quality and the training process is better.
00:29:00.000 --> 00:29:06.720
Yet still it is hard for certain people to understand the differences, uh, right?
00:29:06.799 --> 00:29:16.559
It's a lot of time takes a niche domain expert to be able to differentiate between something that's good enough versus something that is actually humanly good.
00:29:16.880 --> 00:29:17.359
Uh, right.
00:29:17.440 --> 00:29:34.160
And I do think that this creates a future where not only work slop is more prevalent, but the baseline of what you would get out of a large language model with decent context engineering is probably going to be, well, a little sloppy, sloppier than it maybe was a year ago.
00:29:34.480 --> 00:29:36.240
So here's what you need to do.
00:29:36.480 --> 00:29:48.640
Here is the roadmap, all right, on how to get over the everything is fake dilemma and how to properly leverage your company's human expertise to fight AI work slop.
00:29:49.519 --> 00:29:59.039
You need to audit every customer facing or potential client-facing output and flag anything that's generic or unverifiable.
00:29:59.200 --> 00:30:08.400
So that's anything that's currently in production, anything on your website, anything on your pitch decks, uh, anything, you know, in your emails, anything, right?
00:30:08.559 --> 00:30:11.759
And anything in draft version that maybe didn't make it to publication.
00:30:11.920 --> 00:30:15.200
Then you need to identify your strongest experts in those areas.
00:30:15.359 --> 00:30:19.920
You should first uh, you know, flag anything and then categorize all those things.
00:30:20.079 --> 00:30:25.599
Then you need to identify your strongest expert and capture how they would actually say this, right?
00:30:25.839 --> 00:30:32.640
Uh let's say, I don't know, it's a it's a sales page for a certain product or service that's you just started selling.
00:30:32.720 --> 00:30:35.119
Uh, and it sounds kind of generic.
00:30:35.200 --> 00:30:37.759
And it's like, well, this sounds like a lot of nothing, right?
00:30:38.079 --> 00:30:41.200
Get your get your people in there and have them tear this apart.
00:30:41.359 --> 00:30:46.000
You know, they're probably looking at it and be like, look at this garbage the marketing department put out, right?
00:30:46.400 --> 00:30:47.279
I can say that.
00:30:47.359 --> 00:30:49.200
I've I've been in marketing for a while.
00:30:49.359 --> 00:30:53.200
Uh, right there, they they maybe don't like it, right?
00:30:53.279 --> 00:30:56.000
Or maybe it's a company that is producing this for you, right?
00:30:56.079 --> 00:30:58.079
A lot of companies use third-party agencies.
00:30:58.319 --> 00:31:00.720
You need to get you audit everything.
00:31:00.960 --> 00:31:07.839
You need to identify your strongest people and capture how they would actually uh solve that or describe it.
00:31:08.000 --> 00:31:12.400
And then you need to build AI systems that bring those people in routinely.
00:31:12.559 --> 00:31:14.240
It's not a one-time thing.
00:31:14.400 --> 00:31:17.759
It's not a, you know, wash your hands once and you're done.
00:31:18.000 --> 00:31:26.559
Uh, you need to make sure that you build in that domain expertise in your AI operations.
00:31:26.880 --> 00:31:34.000
That is an expert-driven loop, not just having some random human say, yeah, this is good.
00:31:34.079 --> 00:31:37.519
That's work slop and it's crushing your company.
00:31:37.680 --> 00:31:42.079
All right, that's a wrap for this episode in the Start Here series.
00:31:42.240 --> 00:31:47.359
Everything is fake and how your company can leverage human expertise and fight AI work slop.
00:31:47.519 --> 00:31:49.039
I hope this one was helpful.
00:31:49.200 --> 00:31:50.480
If so, let me know.
00:31:50.640 --> 00:31:53.920
Well, let me know by going to startheseries.com.
00:31:54.000 --> 00:31:58.160
That is going to give you free access to our inner circle community.
00:31:58.240 --> 00:32:01.359
Yeah, you're not gonna literally find it anywhere else, FYI.
00:32:01.680 --> 00:32:03.920
Uh, even if you try to find it, you're not gonna find it.
00:32:04.000 --> 00:32:07.519
Uh, that's gonna give you free access to our Start Here series.
00:32:07.599 --> 00:32:10.640
You can go read and listen to all of the episodes in order.
00:32:10.799 --> 00:32:15.519
We have a Spotify playlist uh inside uh the community as well.
00:32:15.759 --> 00:32:17.680
So that's it for today.
00:32:17.759 --> 00:32:20.240
I hope this is helpful and thanks for tuning in.
00:32:20.400 --> 00:32:23.359
Hope to see you back tomorrow and every day for more everyday AI.
00:32:23.440 --> 00:32:24.319
Thanks, y'all.
00:00:01.600 --> 00:00:04.000
Everything is fake.
00:00:04.559 --> 00:00:14.320
The text you read on that landing page, that amazing photo you saw on Instagram, the viral video on Twitter, and all of those comments.
00:00:14.560 --> 00:00:21.199
That customer service rep you talked to last week that was surprisingly chipper and actually helpful.
00:00:21.440 --> 00:00:25.519
There's a good chance all of it was fake, or at least AI.
00:00:25.679 --> 00:00:38.079
And I think we're about to cross over into a very dangerous place where the majority of our day-to-day interactions will be engaging with some type of media or medium that's either partially AI generated or completely AI.
00:00:38.640 --> 00:00:43.359
Yes, there's the fast emerging threat of deep fakes and fraud that's not going to go away.
00:00:43.520 --> 00:00:55.119
And while I think there's a whole other episode to be done on the impacts of the everything is fake disease that has on us as individuals, there's probably a more important chip that you have to write right now.
00:00:55.359 --> 00:01:00.320
That's how your company actually uses AI, but remains human.
00:01:00.640 --> 00:01:11.359
See, I'm someone that encourages AI use like all day, every day, but I know most people and most companies don't take the proper care in elevating human expertise while using AI.
00:01:11.519 --> 00:01:22.319
And that's led to this onslaught of work slop in 2026 or the never-ending barrage of generic sounding, uninspiring outputs that your company may be rubber stamping.
00:01:22.560 --> 00:01:25.680
So, how can you fight back against the year of the fake?
00:01:25.840 --> 00:01:29.840
How can your company resonate in a sea full of mundane work slop?
00:01:29.920 --> 00:01:37.200
And how can you make sure the expertise of your brightest people doesn't get drowned out by agentifying all of your business processes?
00:01:37.439 --> 00:01:43.439
Well, we're going to talk about that on today's episode of Everyday AI in our Start Here series.
00:01:49.920 --> 00:01:52.239
So here's the big picture.
00:01:52.799 --> 00:01:56.400
Right now, there is an AI trust crisis.
00:01:56.560 --> 00:01:57.599
It's already here.
00:01:58.159 --> 00:02:11.439
So a recent study from Europool uh sorry, Europol projects that up to 90% of online content may be synthetically generated by the end of this year.
00:02:11.680 --> 00:02:15.680
And that means that trust is essentially gone.
00:02:16.240 --> 00:02:25.680
A new Salesforce state of the AI connected customer survey said that 72% of consumers trust companies less than they did a year ago.
00:02:26.000 --> 00:02:34.319
So if almost everything that we see or read online is going to be AI and people don't trust companies, what's the answer?
00:02:34.560 --> 00:02:39.280
Well, it's elevating your human expertise while still using AI.
00:02:39.599 --> 00:02:42.400
And that's what we're going to talk about on today's show.
00:02:42.800 --> 00:02:46.319
So if you stick around for the next 20 or so minutes, here's what you're going to learn.
00:02:46.400 --> 00:02:53.759
You're going to learn why defaulting to everything is AI generated is now the sound professional discipline and not cynicism.
00:02:54.080 --> 00:03:00.960
You're going to learn how AI works slop is silently destroying trust and revenue before any dashboard shows it.
00:03:01.120 --> 00:03:08.080
And you're going to learn what the most AI native companies do to stay human, authentic, and competitive at scale.
00:03:08.479 --> 00:03:09.759
All right, let's get into it.
00:03:09.840 --> 00:03:12.879
Welcome to Everyday AI and our start here series.
00:03:13.039 --> 00:03:22.479
This, if you're new, well, it's the essential podcast series to both learn the basics of AI, but if you are a regular, to double down your knowledge.
00:03:22.879 --> 00:03:31.120
So uh after doing 700 plus episodes of everyday AI, I didn't have a good answer when people always said, like, hey Jordan, you have a lot of information out there.
00:03:31.280 --> 00:03:31.919
Where do I start?
00:03:32.000 --> 00:03:33.840
Well, you start here with the Start Here series.
00:03:33.919 --> 00:03:36.560
I think it's best if you listen to all of these in order.
00:03:36.719 --> 00:03:40.560
They're shorter episodes, usually about 25 to 30 minutes.
00:03:40.719 --> 00:03:46.319
But if you want to catch up on all of them in order, I suggest you go to startheseries.com.
00:03:46.479 --> 00:03:52.639
That is going to give you free access to our inner circle community and in our start here series uh space.
00:03:52.800 --> 00:03:56.240
You can go and listen to all of the episodes in order, read about them.
00:03:56.319 --> 00:03:59.439
There's a playlist that we keep updated there and everything else.
00:03:59.680 --> 00:04:07.199
So if you missed our last episode, we went over the AI labor shift, when it will happen, and what it means for chops.
00:04:07.680 --> 00:04:17.279
So let's talk about why everything is fake and why I think this is actually one of the biggest problems that most people don't know that we're fighting against.
00:04:17.839 --> 00:04:25.680
So again, when I tell you this, I'm not trying to say this is some, you know, oh, like bragging about something.
00:04:25.839 --> 00:04:39.519
Uh in many instances, because I am drowning in AI every day, um, I feel that I pick up on certain trends and topics, maybe a little bit for the before the average, you know, business user.
00:04:39.680 --> 00:04:44.399
And what I've seen over the last six months is this onslaught of work slot, right?
00:04:44.560 --> 00:04:50.000
Um, and I don't know if the average everyday business consumer has realized this.
00:04:50.160 --> 00:04:50.879
Maybe you have.
00:04:51.040 --> 00:04:56.800
If you spend any time on, you know, LinkedIn, Twitter, you've probably already seen this.
00:04:56.879 --> 00:05:01.600
Uh so I think it's already impacted the written word, uh, but I think it's gonna impact everything, right?
00:05:01.680 --> 00:05:08.079
So as vibe coding uh becomes very commonplace, it's gonna impact, you know, all of the apps we use.
00:05:08.160 --> 00:05:11.680
There's gonna be app slop, uh, there's gonna be video slop, right?
00:05:11.759 --> 00:05:18.399
Uh, but essentially, I think the smart business consumer needs to realize that this is just the future.
00:05:18.560 --> 00:05:21.279
And there are things that we can do to fight against it.
00:05:21.439 --> 00:05:22.959
So let's start here.
00:05:24.079 --> 00:05:32.399
Right now, no one really knows when or if something is AI generated, which is maybe a good thing, right?
00:05:32.480 --> 00:05:35.600
If you're thinking, oh, well, I can use AI at scale.
00:05:35.839 --> 00:05:36.639
Yes, you can.
00:05:36.800 --> 00:05:51.360
So when something is done correctly, it is very hard uh to tell the difference between uh, you know, an expert who is using AI correctly at scale versus a team of maybe 20 humans who aren't using AI, uh, right.
00:05:51.439 --> 00:05:56.959
As a former uh, you know, journalist, I was a Pulitzer uh fellow, right?
00:05:57.120 --> 00:05:59.279
I won the ACP Story of the Year.
00:05:59.439 --> 00:06:03.519
Um, so I was a pretty decent journalist way back in the day, pretty good at writing, right?
00:06:03.600 --> 00:06:09.120
I I think writing, we've already far surpassed what the average large language model can do.
00:06:09.279 --> 00:06:10.800
But there's a big gap there.
00:06:10.959 --> 00:06:13.040
Because when I say that, a lot of people won't believe me.
00:06:13.120 --> 00:06:19.439
And they're gonna be like, no, what comes out of you know, Chat GPT or Gemini or Claude or, you know, Copilot, whatever, is kind of garbage.
00:06:19.519 --> 00:06:20.720
No, it's it's not.
00:06:20.879 --> 00:06:26.160
It's better than what award-winning writers like myself can do if you know what you're doing.
00:06:26.240 --> 00:06:27.199
But that's a big if.
00:06:27.279 --> 00:06:30.240
And that is elevating the human expertise, uh, right.
00:06:30.319 --> 00:06:40.079
Because if someone goes in and doesn't put a lot of care, they don't understand the basics of context engineering or understand how large language models work, they're not going to be able to produce something that's economically valuable.
00:06:40.160 --> 00:06:41.519
They're gonna produce work slop.
00:06:41.600 --> 00:06:49.519
So that is the uh generic output that you get from you know, just trying to either take the shortest way out or try to get the quickest output.
00:06:49.680 --> 00:07:01.680
Uh, but right now, humans can't tell the difference between AI content that's produced at by an expert and actual, you know, expert uh created good human-only AI content.
00:07:01.759 --> 00:07:12.480
Uh, in a study from Beringa last year showed that people claiming to confidently be able to spot AI images only scored about 30% accuracy.
00:07:12.639 --> 00:07:18.959
So, yeah, uh, even when people are like, oh, I'm sure that this is or is not an AI image, no.
00:07:19.120 --> 00:07:27.199
Uh also self-reported confidence in AI detection is rising every year while actual accuracy keeps just uh declining.
00:07:27.759 --> 00:07:33.120
And this is not something that your team can just solve with better instincts or better tools.
00:07:33.279 --> 00:07:34.560
So, what do I mean by this?
00:07:34.720 --> 00:07:37.680
The quality of what AI can produce.
00:07:38.160 --> 00:07:42.399
I keep, I've I've been saying this over and over the last couple of months.
00:07:42.720 --> 00:07:52.000
The quality and the the just scale that we've seen in the last three months has far surpassed what we got the three years prior.
00:07:52.160 --> 00:07:57.759
And that's what I think has gotten us to this crisis of, well, everything's fake, because you can't tell, right?
00:07:57.920 --> 00:08:08.319
If someone knows what they're doing uh in, you know, Nano Banana Pro or you know, Seed Dance for video generation, Vio for uh right, all of these platforms.
00:08:08.480 --> 00:08:23.120
If you're using the best platform for image editing, video editing, text generation, web, right, for software engineering, agentic coding, the output ultimately, if you have an expert driving it, is pretty much indistinguishable, which is crazy to say that.
00:08:23.279 --> 00:08:24.639
Even video, right?
00:08:24.720 --> 00:08:25.360
We never thought.
00:08:25.439 --> 00:08:34.000
We thought, you know, back when we in the early Dolly one, Dolly two days, you would have said, oh, you know, AI video is 20 years away, right?
00:08:34.159 --> 00:08:38.240
But here we are, you know, five years after that point in time.
00:08:38.320 --> 00:08:46.240
And well, no, it's already, you know, gotten to the point of human level again if you have an expert human driving it.
00:08:47.679 --> 00:08:53.200
So when everything is fake, you can't look over fraud.
00:08:53.360 --> 00:08:54.559
That's only one concern.
00:08:54.639 --> 00:08:56.159
But I think you have to understand this.
00:08:56.320 --> 00:08:59.440
So I am speaking to this from a couple of different ways.
00:08:59.679 --> 00:09:08.000
Um, for consumers, I think today's show may be helpful that you need to change your mindset and assume literally everything is fake.
00:09:08.080 --> 00:09:09.200
Everything is AI, right?
00:09:09.360 --> 00:09:12.000
I've gotten a lot more comments recently, right?
00:09:12.080 --> 00:09:16.159
That uh, you know, this podcast or myself, uh AI, no.
00:09:16.320 --> 00:09:21.600
Uh I've literally had a cold for off and on for like three months, yeah, right.
00:09:21.679 --> 00:09:23.120
But I've gotten all these comments recently.
00:09:23.279 --> 00:09:25.519
It's like, oh no, this is this is no.
00:09:25.600 --> 00:09:37.600
It's uh I don't think AI can uh quite ramble on like I do, uh, not yet, or uh yet simulate my stuffy nose or horsey throat, right?
00:09:37.679 --> 00:09:42.000
So no, uh if if you're uh listening on the podcast, this is not AI.
00:09:42.080 --> 00:09:45.120
I'm a real human doing this live and unedited, unscripted.
00:09:45.200 --> 00:09:48.080
My voice has just been gone off and on for three months.
00:09:48.159 --> 00:09:51.120
But I think it's a good thing for consumers to have, right?
00:09:51.200 --> 00:09:55.120
When you are having that conversation with uh someone over the phone, right?
00:09:55.200 --> 00:10:02.639
When you're watching something or, you know, probably the biggest thing that no one's talking about, just the right agentic AI, right?
00:10:02.720 --> 00:10:05.600
With all these, you know, the open claw variants out there.
00:10:05.759 --> 00:10:08.639
Probably anything you read online, even discourse, right?
00:10:08.799 --> 00:10:19.120
People debating things on, you know, Twitter, LinkedIn, Reddit, Quora, it's probably mostly going to be bot driven, I would say, uh, within a couple of months.
00:10:19.200 --> 00:10:20.799
And people aren't going to understand that.
00:10:21.039 --> 00:10:27.600
So, yes, I think this is important for consumers to understand that pretty much anything you interact with will probably be fake.
00:10:27.759 --> 00:10:31.759
But I think it's also as business leaders, I think it's important to understand, right?
00:10:31.840 --> 00:10:35.440
That that email you got from a client, is it real or is it not?
00:10:35.600 --> 00:10:36.000
Right.
00:10:36.159 --> 00:10:39.120
Um, you know, pretty, pretty soon, right?
00:10:39.279 --> 00:10:40.639
Uh, I'm I'm getting this all the time.
00:10:40.799 --> 00:10:43.120
I'm getting emails from AI agents.
00:10:43.279 --> 00:10:45.120
Sometimes they disclose it, sometimes they don't.
00:10:45.200 --> 00:10:51.519
I've gotten video pitches from, you know, clearly things that are AI and people think that I maybe I can't tell the difference.
00:10:51.600 --> 00:10:52.080
Uh, right.
00:10:52.240 --> 00:10:59.759
So, yes, this does lead to problems on the consumer side and on the business decision maker side, but also fraud, right?
00:10:59.840 --> 00:11:03.120
Company wide, I think you have to be uh paying attention to this.
00:11:03.360 --> 00:11:14.320
So uh in the experienced future of fraud forecast that came out this year, 72% of business leaders identified AI-enabled fraud as a top operational challenge.
00:11:14.639 --> 00:11:23.120
Uh, I think maybe two years ago, I think most people understood that AI was was a huge threat, right?
00:11:23.200 --> 00:11:30.159
Because what you could do with with deep fakes, uh things like that, voice cloning, um, obviously a huge concern.
00:11:30.240 --> 00:11:36.000
But now the barrier to do this, the technical barrier in the technical know-how is essentially zero.
00:11:36.159 --> 00:11:42.879
Uh, you know, anyone with that can read within 30 minutes could probably figure out how to do this at scale.
00:11:43.039 --> 00:11:47.120
Voice cloning, uh, you know, AI videos that look very real.
00:11:47.360 --> 00:11:49.120
Um, it's it's very easy, right?
00:11:49.200 --> 00:11:50.639
These fraud tools are now free.
00:11:50.720 --> 00:11:57.360
They don't require any technical skill, then they allow complete autonomy and anonymity as well.
00:11:57.679 --> 00:12:04.639
So anything that you is important for your day-to-day business operations.
00:12:04.720 --> 00:12:12.480
So vendor proposals, you know, job candidates, uh, executive communications, they can no longer just be verified by site alone.
00:12:12.639 --> 00:12:20.720
I think it's important that we come to that realization that, you know, these routines that we've gone through throughout the years, oh, checking email, right?
00:12:20.879 --> 00:12:21.440
Hiring.
00:12:21.679 --> 00:12:24.240
Oh, yes, looking at these resumes, right?
00:12:24.399 --> 00:12:26.559
Probably we've already seen this over the past two years.
00:12:26.720 --> 00:12:32.080
Now you assume every single resume is written by, you know, ChatGPT or Claude or Gemini or something, right?
00:12:32.159 --> 00:12:38.159
But the same thing with videos uh as well, audio, you know, interviews, similar.
00:12:39.759 --> 00:12:43.440
There's a flip side to that, to this deep fake issue, right?
00:12:43.519 --> 00:12:49.600
And this is again, maybe good conversation for another uh episode where we can go deeper on this.
00:12:49.759 --> 00:12:52.879
But even proof of things that actually happen.
00:12:53.200 --> 00:12:55.440
I think that's gonna be hard to prove, right?
00:12:55.519 --> 00:12:57.360
That's called the liar's dividend.
00:12:57.440 --> 00:13:01.360
That's when real data and real expertise gets lumped in with AI spam, right?
00:13:01.440 --> 00:13:08.799
Kind of how, you know, people have been saying, oh, Jordan, you're, you know, because my voice has been going in and out and my audio is a little weird because of that.
00:13:08.960 --> 00:13:11.279
People are saying, oh, no, you're AI, right?
00:13:11.360 --> 00:13:12.639
No, that's the liar's dividend.
00:13:12.799 --> 00:13:15.600
But I think it's important to realize and understand as well.
00:13:15.840 --> 00:13:18.879
And and because of AI content detection, right?
00:13:18.960 --> 00:13:20.480
Which is not real, right?
00:13:20.559 --> 00:13:35.120
There's obviously things that are a little bit better and a little more reliable on the video and the image side, uh, you know, audio as well in terms of invisible uh watermarks in some type of media, but in written media, it's there's no such thing, right?
00:13:35.279 --> 00:13:44.960
Uh so when you have AI content detection, you know, coming out and say, hey, you know, your company's proposal, you know, couldn't be accepted for this RFP because it was AI generated, right?
00:13:45.120 --> 00:13:46.000
Well, no.
00:13:46.159 --> 00:13:55.759
Uh you know, and that's a problem because I think that your company's real expertise can actually be dismissed as AI generated, uh, just because there's zero recourse or proof.
00:13:56.879 --> 00:13:59.200
And that's led to this trust crisis.
00:13:59.279 --> 00:14:03.519
That if you don't know it's happening right now, it is, right?
00:14:03.679 --> 00:14:16.879
So, like I said, that Salesforce study that said that 72% of people trust companies less this year than they did last year, that is a drastic drop.
00:14:17.519 --> 00:14:18.000
All right.
00:14:18.240 --> 00:14:25.279
And that has led to, I think, or is the result of maybe work slop.
00:14:25.600 --> 00:14:25.840
Right.
00:14:26.159 --> 00:14:30.159
Um, I think consumers are trusting companies less.
00:14:31.200 --> 00:14:40.960
Number one, if you are putting out low effort content at scale, which I think so many, uh, so many people are.
00:14:41.120 --> 00:14:43.360
And that has led to uh work slot.
00:14:43.440 --> 00:14:54.159
But number two, if if if you're not putting a a human face, if you're not putting imperfections out in the world, that's one of the reasons all along I've wanted to, right?
00:14:54.240 --> 00:15:00.240
Uh before I even started everyday AI more than three years ago, I said, I'm not gonna edit this thing.
00:15:00.399 --> 00:15:09.519
I said, I'm gonna go on, I'm gonna do this thing live because I knew at the time, I'm like, in a couple of years, you know, this is all gonna be AI generated.
00:15:09.600 --> 00:15:14.159
So I'm gonna get my imperfections, my, my stuttering, my sniffing, all of it.
00:15:14.240 --> 00:15:16.159
Like that's that's human, that's authenticity.
00:15:16.240 --> 00:15:23.840
And I think that's one of the reasons why, you know, this this podcast has, you know, kind of grown in popularity over the years because it's real, it's authentic.
00:15:23.919 --> 00:15:27.440
And I think that brands need to understand that too, because of the trust crisis, right?
00:15:27.519 --> 00:15:33.519
So, even good example, three major dictionaries named AI Slop their word of the year in 2025.
00:15:33.679 --> 00:15:34.080
Uh, right.
00:15:34.159 --> 00:15:38.159
And work slop, I think, is going to uh trend in that direction.
00:15:38.320 --> 00:15:41.679
If you haven't heard of work slot before, right?
00:15:41.759 --> 00:15:47.519
That's just AI output that's technically competent, but it carries like no domain expertise.
00:15:47.759 --> 00:15:50.080
It just sounds like bland, right?
00:15:50.480 --> 00:15:53.919
And and social media is already a graveyard of enterprise work slot.
00:15:54.000 --> 00:15:58.159
Most companies don't even realize that they are actually contributing to this.
00:16:00.000 --> 00:16:02.320
And here's three reasons.
00:16:02.480 --> 00:16:06.799
I think that companies are still falling into the work slop trap.
00:16:07.039 --> 00:16:11.919
And again, we're gonna get to how you can overcome this uh with human expertise.
00:16:12.080 --> 00:16:16.159
But the three reasons are education, cost, and accuracy.
00:16:16.399 --> 00:16:17.679
All right, so education.
00:16:17.840 --> 00:16:26.320
So uh recent McKinsey study said that only six percent of companies qualify uh as AI high performers, generating mean of meaningful business impact.
00:16:26.639 --> 00:16:38.879
Uh, what that is, is everyone's using AI, but how many people out there have actually learned, have actually taken courses, are actually taught, right?
00:16:39.039 --> 00:16:41.039
We use large language models every day.
00:16:41.519 --> 00:16:45.120
Does anyone know top K, right?
00:16:45.360 --> 00:16:47.440
Top P temperature, right?
00:16:47.679 --> 00:16:52.320
Not that those things are super important, but you should at least know what those things are, right?
00:16:52.399 --> 00:16:55.120
For the things that we're using every single day.
00:16:55.279 --> 00:16:56.559
This is the equivalent.
00:16:56.720 --> 00:17:11.440
This is the like 2010 equivalent of you know, sitting in front of a computer as most knowledge workers do in 2010 and not knowing what an email was, or not knowing uh, I don't know, what a URL was.
00:17:11.599 --> 00:17:13.839
It's like, no, you know what a URL is.
00:17:13.920 --> 00:17:17.039
That's the thing you put in the bar, you type it, and you click the button, right?
00:17:17.200 --> 00:17:19.119
It's it's it's how it operates and how it works.
00:17:19.279 --> 00:17:21.119
Anyways, education is the first reason.
00:17:21.279 --> 00:17:22.319
Number two, cost.
00:17:23.440 --> 00:17:27.039
The economics of AI are irresistible and undeniable.
00:17:27.200 --> 00:17:35.279
AI generates like quality, usually or passable outputs in seconds at a fraction of the human cost.
00:17:35.759 --> 00:17:45.680
Uh, and then accuracy, I think is the third reason that we have so much work slop, because so much of the day-to-day deliverables fall into a gray area, right?
00:17:46.000 --> 00:17:48.799
I think good enough passes for good.
00:17:49.440 --> 00:17:50.000
All right.
00:17:50.160 --> 00:17:51.279
Let me say that again.
00:17:51.440 --> 00:17:59.680
And how many times, if you're being honest with yourself, if you've just blindly used AI output and you've looked at something, how many times have you said good enough?
00:18:00.160 --> 00:18:00.480
Right?
00:18:00.640 --> 00:18:08.640
Whereas before AI, when when when you were doing it manually, I don't think good enough would ever pass for good.
00:18:08.880 --> 00:18:09.680
Now it does.
00:18:10.400 --> 00:18:13.519
In an age of AI, right, where you're like, oh, that's good enough.
00:18:14.960 --> 00:18:29.680
So we've talked about so far why uh work slop is on the rise, the impacts uh maybe personally on consumers, uh, fraud, deep fake.
00:18:30.160 --> 00:18:38.319
But here's the cure it's elevating human expertise inside of your AI setup.
00:18:40.079 --> 00:18:44.079
You need strategic oversight from experts.
00:18:44.480 --> 00:18:54.880
So a Smith OS uh report found that AI content with human strategic oversight performs more than four times better than fully automated outputs out of AI.
00:18:55.279 --> 00:18:55.759
Right?
00:18:56.000 --> 00:19:02.240
This is what makes AI decision makers click yes or no on that proposal.
00:19:03.519 --> 00:19:21.759
Four times higher, a four times better result when you properly put in human expertise, when you properly elevate those domain experts at the right place at the right time in your AI workflows, which unfortunately isn't being done, right?
00:19:21.920 --> 00:19:33.680
Because I think when AI made content creation nearly free, I think don't that instantly elevated domain knowledge to be the competitive moat overnight.
00:19:33.920 --> 00:19:47.039
And I'm talking about basic text-to-text large language models all the way through fully autonomous uh you know, multi-agent orchestration, everything.
00:19:48.319 --> 00:20:06.160
It's elevating the right domain expert at the right time in the process because the winners right now, those that are getting the most out of it, are pouring so much human expertise into their AI that the outputs are unmistakably theirs alone, right?
00:20:07.440 --> 00:20:14.240
And if you listen in our start here series, we talked about a little bit of this in context in the context engineering episode.
00:20:14.640 --> 00:20:19.599
Uh, but I think it's worth maybe repeating one or two things from that.
00:20:22.000 --> 00:20:35.759
It's all about making sure that you have the expert lined up with whoever is building whatever AI flows.
00:20:36.079 --> 00:20:37.119
Here's what I mean by that.
00:20:37.839 --> 00:20:49.359
I think so many times in so many organizations that I talk to, I say, okay, you have this great you know, scheduled AI flow, you know, it's great, it's hands off, cool, great.
00:20:49.519 --> 00:20:50.640
Who set it up?
00:20:50.880 --> 00:20:54.160
It's always usually someone technical, right?
00:20:54.400 --> 00:20:58.960
Usually maybe companies have an AI champion or two, maybe it's someone in IT.
00:20:59.119 --> 00:21:09.599
I think I see it more on the IT side for uh, you know, Windows co-pilot enterprise organizations, uh, but usually it's someone who's technical or an AI champion, but who's actually benefiting from it?
00:21:09.759 --> 00:21:13.359
Who's going in and you know, copying and pasting and updating it?
00:21:13.519 --> 00:21:15.759
It's usually not that person, right?
00:21:16.000 --> 00:21:37.759
I would venture in larger organizations, uh, I would say less than 10% of the output that ultimately gets used uh in a deliverable, in an end artifact, is coming from someone with domain expertise, with very little or zero input in the front end or the back end.
00:21:37.839 --> 00:21:39.680
And that's where it's the most important.
00:21:39.839 --> 00:21:46.160
It's usually a technical person who's setting it up or an AI champion or something, you know, oh, we found this on the internet.
00:21:46.319 --> 00:21:48.960
This looked good, let's plug it in, and it's good enough.
00:21:49.279 --> 00:21:53.920
But the winners are the ones that are pouring more human expertise.
00:21:54.640 --> 00:21:59.759
Because, like we talked about in our context engineering episode, generic prompts produce generic options.
00:22:00.079 --> 00:22:08.799
Output in your proprietary data, your first person or first company reasoning decision logic that changes everything.
00:22:08.960 --> 00:22:17.440
And you need to start, if you haven't already, you need to document why your company makes decisions, not just what decisions it makes.
00:22:17.599 --> 00:22:22.400
In the same way, I talk so much, and I know you guys are probably annoyed at me saying this.
00:22:22.720 --> 00:22:36.880
You need to spend more time looking at the chain of thought in a large language model than you do on your front end context engineering or whatever you do on the back end, copying, pacing, producing a document, right?
00:22:36.960 --> 00:22:38.880
You need to be able to iterate multiple times.
00:22:39.039 --> 00:22:43.839
That is how you properly insert your company's domain expertise.
00:22:44.000 --> 00:22:46.400
Uh, it's through the proper context engineering process.
00:22:46.559 --> 00:22:53.279
And if you took our, you know, prime prompt polish PPP course uh inside our inner circle community, you already know this.
00:22:53.359 --> 00:23:02.240
But that's where you insert your domain expertise is going through how a certain large language model uh will tackle a certain issue.
00:23:02.319 --> 00:23:07.599
And you have to be using the right model, the right feature, the right mode uh for the right problem, right?
00:23:07.759 --> 00:23:11.119
And that might look a little different depending on what sector you're working in.
00:23:11.359 --> 00:23:13.759
But the chain of thought and sitting down, right?
00:23:13.920 --> 00:23:21.440
Even if you're uh, you know, if you have a certain setup where, hey, someone from IT or I don't know, someone from marketing has to set up a certain thing.
00:23:21.519 --> 00:23:30.799
Well, they need to be sitting down with the actual domain expert, looking at the chain of thought and saying, hey, according to our SOP, according to our skills that we set up, right?
00:23:30.960 --> 00:23:35.119
Um, this is what the input is on the context engineering side.
00:23:35.279 --> 00:23:36.319
This is what the output is.
00:23:36.480 --> 00:23:43.680
But more importantly, let's walk through the chain of thought and let's see how the model tackled this problem, right?
00:23:44.480 --> 00:23:52.000
And documenting the why, that is one of the most powerful things that companies that are winning right now, that's why.
00:23:52.160 --> 00:23:59.119
And that's just context engineering reframed as a trust strategy, not just for productivity improvement.
00:23:59.599 --> 00:24:03.200
Is that going to make your company more productive doing that?
00:24:03.440 --> 00:24:04.960
Absolutely, right?
00:24:05.279 --> 00:24:10.160
But that is ultimately going to create more trust in the end.
00:24:10.319 --> 00:24:13.519
That is going to be that that process right there, right?
00:24:13.680 --> 00:24:19.759
Proper proper context engineering, using the right model, you know, education, training, all that.
00:24:19.920 --> 00:24:35.359
But having your domain expert be involved in the auditing, in the iteration, looking at the chain of thought, that what is going to ultimately lead to less work slop, more trust, and more ROI for your company.
00:24:36.400 --> 00:24:43.200
And this is again why I will continue to call out how bad human in the loop is.
00:24:43.440 --> 00:24:43.759
Right.
00:24:43.920 --> 00:24:50.799
I've been saying this, sorry, uh way before everyone else about how human in the loop is bad.
00:24:50.960 --> 00:25:01.279
It's one of the worst things, I think, that has happened in or around the AI industry, uh, aside from just an overall general lack of education, training, and uh basic knowledge.
00:25:01.519 --> 00:25:03.920
Human in the loop is bad.
00:25:04.240 --> 00:25:07.039
Human in the loop leads to work slop.
00:25:07.599 --> 00:25:16.880
Human in the loop is going to uh continue to contribute to the everything is fake dilemma, and it is going to be bad.
00:25:17.359 --> 00:25:23.119
We need, as again, if you are a longtime listener of this podcast, you know maybe it's the first time listening.
00:25:23.279 --> 00:25:24.720
Well, let me introduce this to you.
00:25:24.960 --> 00:25:27.359
Expert-driven loops are what we need.
00:25:27.599 --> 00:25:34.960
Human in the loop means that you can just put any human passively into any AI system, and that's some sort of guardrail.
00:25:35.039 --> 00:25:36.160
That's garbage.
00:25:36.240 --> 00:25:37.519
Uh, that's failure.
00:25:37.680 --> 00:25:38.960
That is work slot.
00:25:39.279 --> 00:25:40.880
You need expert-driven loops.
00:25:40.960 --> 00:25:47.359
That means the domain uh expertise, the domain expert driving the loop, right?
00:25:47.440 --> 00:25:48.720
And what does that mean?
00:25:48.960 --> 00:25:59.359
Usually when I talk about when anyone talks about uh a loop, we're talking about an agentic or semi-agentic or AI-powered workflow.
00:25:59.599 --> 00:26:05.200
Something that happens in your company repeatedly, uh, right, over and over again.
00:26:05.359 --> 00:26:06.640
That's that's a loop.
00:26:06.799 --> 00:26:11.519
You need an expert driving it proactively, right?
00:26:11.680 --> 00:26:19.440
Because Gardner right now predicts that companies that replaced human agents with generic AI will be forced to rehire by 2028.
00:26:19.680 --> 00:26:22.720
And I think human in the loop is just a passive checkpoint.
00:26:22.799 --> 00:26:29.519
But expert-driven loops mean experts are actively shaping the context and reviewing against professional standards.
00:26:29.759 --> 00:26:30.160
All right.
00:26:30.240 --> 00:26:44.720
This is why I think it's so important, and we've talked about this in previous episodes, to have internal company benchmarks and internal scoping that you are doing not yearly, not quarterly, but probably monthly or bi-monthly.
00:26:46.480 --> 00:26:52.400
So work slop is only going to get worse and more prevalent.
00:26:52.880 --> 00:26:58.880
And I think this is important because we are gonna have AI slop-heavy training data.
00:26:59.200 --> 00:27:03.519
This is something I think uh people very much overlook.
00:27:03.680 --> 00:27:15.920
So, right, if we go back uh to one of our initial stats from Europol that projected that up to 90% of online content may be synthetically generated by the end of this year.
00:27:16.000 --> 00:27:20.240
Uh, and I would say it's probably gonna be more than 90%, if I'm being honest.
00:27:20.400 --> 00:27:22.799
Um, but what does that create?
00:27:23.119 --> 00:27:29.920
I think that creates trust in AI that we probably shouldn't have.
00:27:30.079 --> 00:27:39.440
Uh, because even if you just looked at the relatively quality, like the relative quality, I don't know if there's a a a standard for this or a benchmark.
00:27:39.680 --> 00:27:41.839
If not, I hope someone can create it.
00:27:42.000 --> 00:27:42.480
Uh, right.
00:27:42.640 --> 00:27:52.079
But if you looked at the relative quality of something that was produced on the internet in 20, uh, I don't know, we'll say pre-uh AI, right?
00:27:52.160 --> 00:27:56.880
Because even before Chat GPT, uh, you know, AI slop was a huge problem.
00:27:57.039 --> 00:27:57.519
Uh, right.
00:27:57.599 --> 00:28:09.759
The early GPT technologies that uh predated uh ChatGPT were, you know, already out there polluting the internet with garbage with AI work slot before anyone knew ChatGPT existed.
00:28:09.839 --> 00:28:18.960
Uh, but I think now, right, the average piece of, I don't know, we'll just say blog posts, right?
00:28:19.119 --> 00:28:22.079
Uh, because that's a lot of what goes into trading data.
00:28:22.319 --> 00:28:28.720
The average quality of a blog post, I would say has gone down by two to three X very easily over the last 10 years.
00:28:28.799 --> 00:28:30.720
And it's because of AI, right?
00:28:30.880 --> 00:28:39.359
So that just obviously creates this regurgitated cycle of poor or poorer quality source materials.
00:28:39.519 --> 00:28:47.839
So um, I do think that even more so in the future, yes, the training uh at the companies is getting better, right?
00:28:47.920 --> 00:28:59.759
The reinforcement learning with human feedback and other uh scaling technologies at the big AI Frontier Labs helps with this, you know, helps to make sure that the training data is uh higher quality and the training process is better.
00:29:00.000 --> 00:29:06.720
Yet still it is hard for certain people to understand the differences, uh, right?
00:29:06.799 --> 00:29:16.559
It's a lot of time takes a niche domain expert to be able to differentiate between something that's good enough versus something that is actually humanly good.
00:29:16.880 --> 00:29:17.359
Uh, right.
00:29:17.440 --> 00:29:34.160
And I do think that this creates a future where not only work slop is more prevalent, but the baseline of what you would get out of a large language model with decent context engineering is probably going to be, well, a little sloppy, sloppier than it maybe was a year ago.
00:29:34.480 --> 00:29:36.240
So here's what you need to do.
00:29:36.480 --> 00:29:48.640
Here is the roadmap, all right, on how to get over the everything is fake dilemma and how to properly leverage your company's human expertise to fight AI work slop.
00:29:49.519 --> 00:29:59.039
You need to audit every customer facing or potential client-facing output and flag anything that's generic or unverifiable.
00:29:59.200 --> 00:30:08.400
So that's anything that's currently in production, anything on your website, anything on your pitch decks, uh, anything, you know, in your emails, anything, right?
00:30:08.559 --> 00:30:11.759
And anything in draft version that maybe didn't make it to publication.
00:30:11.920 --> 00:30:15.200
Then you need to identify your strongest experts in those areas.
00:30:15.359 --> 00:30:19.920
You should first uh, you know, flag anything and then categorize all those things.
00:30:20.079 --> 00:30:25.599
Then you need to identify your strongest expert and capture how they would actually say this, right?
00:30:25.839 --> 00:30:32.640
Uh let's say, I don't know, it's a it's a sales page for a certain product or service that's you just started selling.
00:30:32.720 --> 00:30:35.119
Uh, and it sounds kind of generic.
00:30:35.200 --> 00:30:37.759
And it's like, well, this sounds like a lot of nothing, right?
00:30:38.079 --> 00:30:41.200
Get your get your people in there and have them tear this apart.
00:30:41.359 --> 00:30:46.000
You know, they're probably looking at it and be like, look at this garbage the marketing department put out, right?
00:30:46.400 --> 00:30:47.279
I can say that.
00:30:47.359 --> 00:30:49.200
I've I've been in marketing for a while.
00:30:49.359 --> 00:30:53.200
Uh, right there, they they maybe don't like it, right?
00:30:53.279 --> 00:30:56.000
Or maybe it's a company that is producing this for you, right?
00:30:56.079 --> 00:30:58.079
A lot of companies use third-party agencies.
00:30:58.319 --> 00:31:00.720
You need to get you audit everything.
00:31:00.960 --> 00:31:07.839
You need to identify your strongest people and capture how they would actually uh solve that or describe it.
00:31:08.000 --> 00:31:12.400
And then you need to build AI systems that bring those people in routinely.
00:31:12.559 --> 00:31:14.240
It's not a one-time thing.
00:31:14.400 --> 00:31:17.759
It's not a, you know, wash your hands once and you're done.
00:31:18.000 --> 00:31:26.559
Uh, you need to make sure that you build in that domain expertise in your AI operations.
00:31:26.880 --> 00:31:34.000
That is an expert-driven loop, not just having some random human say, yeah, this is good.
00:31:34.079 --> 00:31:37.519
That's work slop and it's crushing your company.
00:31:37.680 --> 00:31:42.079
All right, that's a wrap for this episode in the Start Here series.
00:31:42.240 --> 00:31:47.359
Everything is fake and how your company can leverage human expertise and fight AI work slop.
00:31:47.519 --> 00:31:49.039
I hope this one was helpful.
00:31:49.200 --> 00:31:50.480
If so, let me know.
00:31:50.640 --> 00:31:53.920
Well, let me know by going to startheseries.com.
00:31:54.000 --> 00:31:58.160
That is going to give you free access to our inner circle community.
00:31:58.240 --> 00:32:01.359
Yeah, you're not gonna literally find it anywhere else, FYI.
00:32:01.680 --> 00:32:03.920
Uh, even if you try to find it, you're not gonna find it.
00:32:04.000 --> 00:32:07.519
Uh, that's gonna give you free access to our Start Here series.
00:32:07.599 --> 00:32:10.640
You can go read and listen to all of the episodes in order.
00:32:10.799 --> 00:32:15.519
We have a Spotify playlist uh inside uh the community as well.
00:32:15.759 --> 00:32:17.680
So that's it for today.
00:32:17.759 --> 00:32:20.240
I hope this is helpful and thanks for tuning in.
00:32:20.400 --> 00:32:23.359
Hope to see you back tomorrow and every day for more everyday AI.
00:32:23.440 --> 00:32:24.319
Thanks, y'all.
