Latest / The Empire of Wealth / The Biggest AI Mistakes Business Owners Are Making

Transcript

The Biggest AI Mistakes Business Owners Are Making · The Empire of Wealth

• 0:00
  Speaker 1: In this video, I'm going to cover the top AI
• 0:02
  mistakes that business owners are currently making. AI has completely
• 0:06
  changed my entire business. Just over the past three weeks,
• 0:10
  I've managed to build an e-commerce store that's only $ 15, 000
• 0:14
  in sales entirely with AI. I've created an AI CFO
• 0:18
  that can order all of my financials. I've even created
• 0:21
  chatbots in Slack that my team can actually use and
• 0:24
  call upon to give them a ton of insights. The
• 0:26
  biggest mistake that I see companies make is they try
• 0:29
  to hire an expert like a developer or a self-proclaimed
• 0:33
  AI expert to come in and teach AI to their
• 0:36
  entire company. This is the wrong approach. What we really
• 0:39
  want to do is we want a leader within the company,
• 0:41
  which is you, to try to teach and create a
• 0:44
  bottoms up culture where the employees of the company start
• 0:48
  using it. So think about it. If you're just the
• 0:50
  founder using AI, sure, you might come 100 times more effective.
• 0:54
  But think about it if there's 20 employees also using it,
• 0:57
  sharing ideas, suddenly they're 10,000 times more effective because it
• 1:01
  creates a culture and it starts to accelerate and everyone
• 1:04
  starts sharing learnings from each other, spinning off agents that
• 1:07
  can do all kinds of tasks. But the truth is
• 1:09
  that it's never going to get adopted again. until you
• 1:12
  do it. Companies always follow the leader. If you're someone
• 1:15
  that needs to hire an AI expert, give a ton
• 1:17
  of permission to do a bunch of things, then it's
• 1:19
  never going to get adopted. You need to create a
• 1:21
  culture where every single week you get together with the
• 1:24
  key employees and you celebrate the wins of AI adoption.
• 1:28
  You as the founder also need to put in guardrails
• 1:31
  such as, what do we do about this data? What
• 1:33
  do we do about budgets? All of this kind of
• 1:35
  stuff before you can actually get them to safely and
• 1:39
  confidently adopt it. Now, here's the issue. How do you
• 1:41
  do that without you yourself as the founder knowing things? Typically,
• 1:45
  whenever I give business advice to founders that come to me,
• 1:48
  I always try to identify the one key thing that's
• 1:51
  going to 10x their business. And usually it's something like
• 1:54
  Facebook ads or marketing or just doubling down on their
• 1:57
  existing acquisition channel. I am very confident that almost every
• 2:01
  single founder, the single highest leverage thing that they could
• 2:04
  learn right now is AI. It's over Facebook ads. It's
• 2:07
  over product development even because things like Facebook ads and
• 2:10
  product are now a subcategory, a subskill of AI. AI
• 2:15
  now amplifies all those skills so much that if we
• 2:18
  learn AI first, we can be far more effective than
• 2:20
  someone that doesn't. The next mistake that I see business
• 2:23
  owners make is they don't start with a goal in mind.
• 2:26
  Businesses are just a series of inputs and outputs. If
• 2:29
  you think about every single department, let's take product. They
• 2:32
  have a series of how to brief in products, how
• 2:35
  to create the product. Then it becomes an output of
• 2:37
  a product. Then it gets sent to marketing. And then
• 2:39
  they have a series of like, how do I take
• 2:41
  photos of this? How do I do this? And it
• 2:43
  keeps going until finance reports on the profit and loss
• 2:46
  statement and tells you how much money you've made. And
• 2:49
  every single thing can be mapped out to that beginning
• 2:52
  initial prompt. Now, if we map that entire process out
• 2:55
  on a board and we look at both what are
• 2:57
  the inputs, outputs, but also what are the objectives and
• 3:00
  the KPIs that each of our employees actually have in
• 3:03
  this subsection, it'll give us the clear places that we
• 3:06
  should actually be applying AI. Most business owners, what they're
• 3:09
  doing is they're downloading something like Chord, And then they're going, wow,
• 3:13
  this is a great copywriting tool. Let's go and let's
• 3:16
  start writing some emails for this. And then going, oh,
• 3:19
  the emails are okay. Yeah, it sped us up. That's great.
• 3:22
  We're just going to use this from now on. That
• 3:23
  is not what we should do. We should be looking
• 3:26
  and benchmarking our email efficiency. We should be looking at
• 3:29
  our total emails that we can actually write. We want
• 3:32
  to be looking at the cost of that department. And
• 3:34
  then we want to build a hypothesis around it that
• 3:36
  if I incorporate AI to my emails, By this process,
• 3:41
  my KPIs will improve in this way. Let me give
• 3:44
  you another example. Let's say we're taking Facebook ad creation.
• 3:47
  When we're looking at the briefing process, the reaching out
• 3:50
  to creators, the actual shooting, the editing, and the reporting
• 3:54
  on the Facebook ads, we want a clear objective that
• 3:57
  AI is going to come in and improve our cost
• 4:00
  per creative. So the amount we actually have to spend
• 4:02
  per creative, let's say we want to double the volume
• 4:05
  at half the cost. Now we have these clear KPIs
• 4:09
  that AI needs to go and solve. The next mistake
• 4:11
  that business owners really make is they either go way
• 4:13
  too advanced with their tools or way too simple. And
• 4:16
  this fits into what we were talking about before, but
• 4:18
  I constantly wanted to create an AI CFO. I was
• 4:21
  just like, this is really, really important. There's so much
• 4:24
  data there. There's so many mistakes are made from here.
• 4:26
  How do I get AI to fact check my existing
• 4:29
  finance employees? Now I tried ChatDBT. I even tried all
• 4:32
  of these AI specialist spreadsheet tools and whatnot, but nothing
• 4:36
  could handle the amount of data that was going through
• 4:39
  these businesses because they are quite large. So what I
• 4:42
  ended up doing is I finally found that actually running
• 4:45
  through Chord Code and actually locally on the device using
• 4:49
  Excel was far, far better. But the way that I
• 4:52
  found this out was not going to instantly going to
• 4:55
  Chord Code, which is a slightly more advanced compared to
• 4:58
  something like just an interface like ChatGPT. I tested the
• 5:01
  simplest solution first. realized that it failed, tested it thoroughly,
• 5:05
  and then I went to the next simplest. Now, what
• 5:08
  simple might mean for you might actually already just mean
• 5:11
  that it's in your tech stack. You might already have
• 5:13
  this actual tool that you're using, say you're using Google Drive.
• 5:17
  So therefore you want to just try, can I use
• 5:20
  the Gemini extension on these kinds of things? And then
• 5:22
  you go, okay, no, this doesn't work. So now I
• 5:24
  need something more robust. Now let's try something like chord code.
• 5:28
  Now this might seem counterintuitive, but when you're first getting
• 5:31
  started out, having multiple tools for different things based on
• 5:34
  where the data actually is, is not necessarily a bad thing.
• 5:37
  What it does mean though, is that you're probably going
• 5:40
  to have to stitch them together, which creates another complication.
• 5:43
  But that being said, we want to keep it as
• 5:45
  simple as possible because that will allow easy team adoption,
• 5:49
  which again is the key to getting the most leverage
• 5:51
  out of AI. Now, the next mistake is so obvious
• 5:53
  and it is so frustrating. I see a bunch of
• 5:56
  business owners, CEO, founders, even people with just no teams
• 6:00
  just getting started and they go, I tested AI, it
• 6:03
  doesn't work. And then I'm like, how did you actually
• 6:06
  test it? And they're like, I sent this one message
• 6:08
  and look at the response it gave me. That is
• 6:10
  absolutely ludicrous. AI is another employee. We need to train
• 6:15
  the employee. We need to figure out how they like
• 6:17
  to work. We need to feed them the right data.
• 6:20
  And until then, we cannot justify or say that it
• 6:24
  is not better than an existing person or an employee. Sure,
• 6:27
  the employee might be at a low rate. It might
• 6:29
  be a customer service agent or someone that we can
• 6:32
  hire remotely, but that doesn't mean that they are going
• 6:34
  to be more efficient and cheaper long-term. And we need
• 6:37
  to give AI a chance because once all this gets
• 6:40
  stitched together, that's when the compound effect happens. The main
• 6:43
  mental model that we need here is let's say we
• 6:45
  have an employee that's a copywriter. Now, I completely disagree
• 6:48
  with the notion that AI is going to replace all jobs.
• 6:51
  I need a copywriter in my team still, but I
• 6:54
  need a copywriter that understands AI. So just assume that
• 6:58
  we've got this copywriter that is paid $ 100, 000 a year.
• 7:01
  They are making, let's say, a million dollars in revenue
• 7:04
  for the business to make sure that they're actually profitable
• 7:07
  within the business. Our goal is to turn this copywriter
• 7:11
  into a $ 10 million a year copywriter. We actually want
• 7:14
  to look at AI through the lens of revenue expansion
• 7:17
  rather than just cutting costs. But this is the thing.
• 7:20
  If I'm talking about taking a copywriter from a million
• 7:23
  dollars to $ 10 million of value, and then I'm spending
• 7:26
  two hours on my AI tool to be like, wow,
• 7:29
  this doesn't work. Of course, it's not going to work
• 7:32
  like that. You're trying to 10x the value, trying to
• 7:34
  create an extra $ 9 million of incremental value to the company.
• 7:38
  Why would that take two hours? The key here is
• 7:41
  just constantly iterate, create the framework to be celebrated in
• 7:45
  the company, break through hurdles together within the team, and
• 7:48
  just stick with one task until you've figured it out.
• 7:51
  Too many people are running 15 different terminals, five different tools,
• 7:55
  all at the same time, trying to figure things out.
• 7:57
  And then the output's terrible. Just like go from, I
• 8:00
  have this objective to fix this KPI, and I'm going
• 8:02
  to stick with this until it does that. The next
• 8:04
  mistake that I see business owners make is they don't
• 8:06
  respect data hygiene. Now, a lot of people are either
• 8:10
  feeding it way too much data or not enough. Now,
• 8:13
  the key here is that LLMs have a context window.
• 8:16
  So if you give it way, way too much, it's
• 8:18
  going and you might have done this before you created
• 8:20
  this giant huge chat with random things about cooking, but
• 8:22
  then also your running program and then asking it business questions.
• 8:25
  And it's just like has no idea what's going on.
• 8:27
  That's probably way too much. Now, a lot of other people,
• 8:30
  they're doing really short prompts with zero context as well
• 8:32
  and expecting it to solve their business problems, which is
• 8:35
  just not how this works. Now, the key here is,
• 8:37
  let's say you've already got a big company and you've
• 8:39
  got some really, really sensitive stuff. What you want to
• 8:42
  do is you want to sandbox your data. Now, there's
• 8:44
  a couple of ways that we can do this. We
• 8:46
  could buy a new laptop. A lot of people are
• 8:48
  buying Mac minis. We can use a VPS. So it
• 8:51
  is slightly more technical. If you're not technical, I wouldn't
• 8:53
  be doing that. Remember, everything that I'm teaching you in
• 8:56
  this video, you can copy my transcript and you could go,
• 8:58
  what does Davey mean by this? When I get stuck
• 9:00
  in AI, I am just constantly asking questions. But let's
• 9:03
  just say that you want to go to AI right
• 9:05
  now and explain your situation and say, how do I
• 9:07
  ensure security? Also give it the data that you actually
• 9:10
  need to get the answer. You're going to get a really,
• 9:12
  really good response. Now, the way that I've approached this
• 9:14
  is I've sandboxed. I've created a laptop that I can
• 9:17
  feed it the data that I'm actually comfortable with it having.
• 9:20
  And then I've slowly progressed over time with giving it
• 9:23
  more and more permission. because we don't want to give
• 9:25
  it all and then it hallucinates or you do something really,
• 9:27
  really stupid and everything breaks. Then what you want to
• 9:29
  do is you might want to make sure that you
• 9:31
  have constant backups and you even redact information that you really,
• 9:34
  really need. Now, here's my key thing and my key
• 9:37
  lesson around data and how to get the most out
• 9:39
  of AI is you want to create consistent loops of
• 9:43
  new data injection for certain tasks. So for example, let's
• 9:47
  say we want to create a Facebook ad briefing bot
• 9:50
  where it gives you all of the best briefs. It's
• 9:52
  a creative strategist. Let's say we give it all of
• 9:55
  our top performing ads that have worked. We give it
• 9:57
  our website. We just give it a bunch of information.
• 9:59
  Maybe we give it customer service tickets redacted about our
• 10:03
  brand itself. And then we go, can you write a
• 10:05
  Facebook ad for this? And then it gives you some really,
• 10:08
  really good ads and you're really excited. And then what's
• 10:10
  it going to do next week? Next week, it's going
• 10:12
  to write the exact same ads because it's got no
• 10:14
  new data. So what you need to do is you
• 10:16
  need to both keep injecting what is working for you,
• 10:19
  but you also want it to look broader. Now, the
• 10:21
  Facebook ads library at Current Speaking is public at the moment.
• 10:24
  And then you can sort by impressions, right? You've got
• 10:26
  the top performing ads for every single brand sitting at...
• 10:29
  the top so what we need to do is we
• 10:31
  can create a scraper and now a lot of people
• 10:34
  have actually used certain apis and complex things to actually
• 10:38
  access facebook ads and whatnot i'm a little bit scared
• 10:41
  about that they've just recently purchased menace at this time
• 10:44
  meaning that i want to keep things not logged in
• 10:47
  because i don't want my facebook ad account is banned
• 10:48
  lucky it is public so you can create a facebook
• 10:51
  ad scraper to look at adjacent niches what is currently
• 10:54
  performing then that can feed in the data for you.
• 10:57
  Now you've got all of the top performing ads on
• 10:59
  Facebook as a whole. Plus you've got your Facebook ads.
• 11:02
  And now this is how we create this dynamic flywheel.
• 11:04
  It will automatically grab the top performers and then automatically
• 11:07
  create what your ad should be next week. So in short,
• 11:10
  create automations to make sure that your data is constantly
• 11:13
  changing and run regular scheduled automations to make sure that
• 11:16
  the outputs that you're getting are also different. If you
• 11:19
  liked today's video, I've created this really advanced AI roadmap
• 11:23
  for e-commerce brands about what I think is the most important,
• 11:26
  impactful AI things that you could actually roll out in
• 11:29
  your company. It's completely free. Just go sign up and
• 11:31
  I'll send you what I think you should be rolling
• 11:33
  out using AI.