Full transcript
0:00In this video, I'm talking about how I
0:01use AI to automate content creation. But
0:03let me tell you this, it's not a
0:05beginner's video. If you haven't worked
0:06it out yet, all of these other videos
0:08that you see out there of people just
0:10giving you a few different tools to use
0:12or a few different prompts and telling
0:14you that you can use them for free, and
0:15it doesn't get you anywhere. So, I'm
0:17sure most people have figured that out
0:18yet. If you haven't, then maybe you need
0:20to go and watch some other videos and
0:22realize that they're not going to get
0:23you anywhere before we actually start
0:25diving into how to build real systems
0:27like this. So, this is a system that we
0:29use to process a ton of content in a
0:32very short period of time. And we do
0:34interviews, and I'm going to get to how
0:36we generate ideas. And interviews are a
0:38great way to produce quality content at
0:40scale without you having to come up with
0:43new ideas all the time. So, if you don't
0:45know me yet, my name is Mitch Assar.
0:47We've been building for clients for the
0:49last 12 months with make big moves. And
0:51we make it clear, we don't build AI
0:52automations. We build AI systems. And
0:54these systems have helped us build big
0:56moves into a brand where we interview a
0:58bunch of the best experts in the AI
1:01space. And we have other brands that in
1:03other different niches as well. So our
1:06everything that we do revolves around
1:07interviews with other people. And to
1:10process those interviews, we need to do
1:12a whole bunch of content on top of those
1:14in order to create really great
1:16products, to promote the content, to do
1:18all of these different things that we
1:19need to do. So, in this video, I'm going
1:21to talk you through a few core concepts
1:24that you need to be thinking about, and
1:26these are advanced concepts. So, if
1:27you're not interested in advanced
1:29concepts that are going to make you a
1:30ton of money, that it's going to free up
1:31a lot of your time, then click off this
1:33video now because now before I continue,
1:35a quick announcement about the AI
1:37content mastery summit. It is coming up.
1:39We've got some epic speakers speaking on
1:41this event and we're going deep into how
1:43to create AI content workflows. Some of
1:45these are super fascinating about how
1:47people are building. And if you want
1:49ideas on how to build faster, better,
1:50and more powerful systems, then make
1:52sure you sign up. It's completely free.
1:54There is an offer inside, but all the
1:55content's free. And I'd really love to
1:57see you there. We're going to dive into
1:59a few things. Number one, smarter way to
2:01find content ideas that don't look like
2:03everyone. Number two, a clear system for
2:05turning any working prompt into
2:06something that runs without them. And
2:08number three, a blueprint for building
2:10an AI operating system for your
2:11business. And I'm actually going to
2:13start there and work my way backwards.
2:15So every business needs an AI operating
2:18system as we move into the new way of
2:21doing things, new way of business. And I
2:23made a video more than two years ago now
2:25about how SAS is dead. And I called it
2:28very early back then. Most of these
2:30tools that we use and including the
2:32thousands of AI tools that are out there
2:34are going to disappear. And what's going
2:36to happen is we you will have your own
2:38AI operating system and we'll simply be
2:41using APIs to pull in all of the data
2:44into a single system. Now the beauty of
2:45this is because AI is getting so great
2:48at coding then you don't even need to
2:50pay somebody to do this. You can simply
2:51prompt it and you can build it all up.
2:53It's possible now but it is a lot harder
2:55than what most people make you think it
2:57is. And we've been talking about this
2:58for a long time is we've need to build
3:00systems where we have full control of
3:02the database where we have full control
3:04of our automations and we can build them
3:06exactly how we want. We prefer a lot of
3:08other people just vibe code with bolt
3:10and lovable and all of these others that
3:12are out there. And this is what's taking
3:14over the old way. So the old way we
3:15would have an email tool, CRM, invoice
3:17software, analytics, project manager,
3:20docs, support tools. You'd have Canva
3:23and all five different other tools in
3:25order to do all of your work. Now, some
3:27of them will remain like video editing
3:29tools probably going to remain, although
3:31Claude's getting close to to replacing
3:33that now as well. But the but what I'm
3:35getting at here, if you haven't realized
3:37it yet, you need an AI operating system.
3:40So, you can start small. We have a bunch
3:42of air table templates on our website.
3:43You can start small and just find one
3:46part of your business in order to get a
3:48feel for it and then begin to do some
3:50more. Now the next problem that we have
3:53is we need to choose whether we when and
3:56where we use an AI agent versus an AI
3:58automation. So if you don't know an AI
4:00agent, they're things like your open
4:02claw, they're things like Menace, all of
4:05these really great tools that they have
4:06their time and place, but I find that
4:08most content that you see online,
4:11they're using AI agents when you could
4:13use an AI automation instead. Now,
4:15what's the difference? An AI automation
4:17is the same process every time. No
4:19thinking required. that it makes people
4:20faster. It's ready in our in minutes,
4:23not hours. And it runs forever as every
4:26single time you trigger it. Plus, you
4:28can make a few changes to the prompts
4:29and like some small optimizations or
4:31adjustments, but you're not giving it a
4:34full job to go out and so 90% of the
4:37workflows that I see in content creation
4:39creator videos can be handled by AI
4:42automation, but they're getting an AI
4:43agent to do it. Now, that's fine. they
4:46still achieve the same result, but an AI
4:48automation would probably cut the cost
4:49by 90 to 99%. So, plus you're going to
4:52get more reliable outputs and you're not
4:54going to have to prompt or chat to your
4:56AI agent as often to steer it in the
4:58right direction. Now, where an AI agent
5:00gets interesting is that it's going to
5:02replace a person, not so much it's going
5:04to make people faster. So, it's actually
5:07going to replace a person. So, if you
5:08think you can just train an AI agent in
5:10a few prompts or a few hours, then
5:12you're probably a little bit naive to
5:15these two different processes. An AI
5:17agent will be super powerful when it
5:19replaces people. That means it's
5:21probably going to take at least a week
5:22to train it fully in order to make sure
5:24it understands how and when to make the
5:26right decisions and what is the
5:28framework for making those decisions.
5:30But overall, 90% of what I see can be
5:32handled by AI automation. So that's
5:34where I spend 90% of my time is in AI
5:37automation. I've seen this play out so
5:39many times in the cycles of AI is that
5:42when a new thing it comes out, it's 12
5:44to 18 months before it actually becomes
5:46viable for real businesses making real
5:49money. So where do we go from here? So
5:50understanding those few concepts is key
5:53and critical. And I'm actually going to
5:54run you through automations and exact
5:56everything that we've got and how it
5:58works. Plus I'm even going to give you
6:01all of the prompts that we use for these
6:03systems. So the prompts will be inside
6:04of our community that you can see here
6:06and you can pull all of these in. These
6:08are just pulled straight from our nan
6:10automations. I've not changed anything.
6:11You get at 100% how we actually create
6:14this content for our internal business.
6:17So we're going to come back to all of
6:19this soon. We're going to break down all
6:20of these automations, including how
6:21we're generating thumbnail images with a
6:23single button click in this system here.
6:26But before we get there, I want to talk
6:27about ideas because this is the number
6:29one problem that I see with content
6:31creation and content creation videos and
6:34the people that come into our community
6:36is everybody wants to generate ideas
6:38with AI. I think that is the worst
6:40possible way to come up with ideas is to
6:43use AI. And let me explain. So the
6:46problem is AI is built to provide you
6:49the mean of all of the data that is out
6:51there. So, no matter how many times
6:52you're prompted about getting an idea,
6:55you'll see the same ideas pop up over
6:56and over again. Now, if you times that
6:58by the hundreds of millions of people
7:00who are using AI, then you're going to
7:02have the same ideas as hundreds of
7:04millions of people. And that's why if
7:06you look online at all the different
7:08content out there in the different
7:09niches, somehow the people who are using
7:11AI badly, they all sound the same.
7:14Whether they're talking about crypto,
7:15AI, business, ice cream, cooking, you
7:18name it, it all somehow sounds the same.
7:20So the that is the problem. So
7:22everybody's generating content without
7:24knowing all of these skills and
7:26concepts. So the solution is you need to
7:28dig a little bit deeper. You can find
7:30content ideas from latest news and
7:32trends. You can find it from viral
7:34content of people who are creative and
7:37actually coming up with their own ideas
7:38and then you can adapt it for your
7:40industry. And you can also create it
7:42through human interviews and stories.
7:44These are your podcasts. podcast is
7:45going going to continue to do well and
7:48because people want people on people
7:50interaction. They want a feeling and
7:52it's very hard to get a feeling from AI
7:54and AI generated content and we've
7:56created it all. I'm not against it but I
7:59think the way that we're going is that
8:01you always have to do the opposite to
8:03what is popular right now. So that is
8:06number one is you need to dig a little
8:07bit deeper and come up with your own
8:09ideas. latest trends, just getting out
8:11in the world and being in the real
8:13world. You come up with a lot of great
8:14ideas just from speaking with people and
8:17seeing how people actually live their
8:18life. Once you do that, you then start
8:20to create the content that you desire.
8:23Then you have a whole bunch of other
8:24things that you need to do with that
8:26content like creating the thumbnails and
8:27the descriptions if you're a YouTuber.
8:29So then we move from okay, we've done
8:32we've got our idea, we've created our
8:33content, now we need to do a few other
8:35things in order to schedule that content
8:37and promote it to get it out there. so
8:40that people start to engage with us
8:41whether we're content creators, a
8:43business or what have you in order to
8:46generate eyeballs on that content that
8:48we've created. So when doing that then
8:51we do it manually first. So we create
8:53all of our prompts manually in chat GBT
8:56or in Claude or in Nano Banana. We do it
8:59manually to find our style that we want
9:01and once we have it optimized we then
9:04automate it with tools like N. From
9:06there, you can then scale it so you get
9:0810x output and that's exactly what we've
9:10done. So, this is a system that we've
9:13created and you can see all of these
9:15different interfaces down here for
9:17creating entire virtual summits. So,
9:19this is just one part of the system. I'm
9:21not going to go through all of it, but
9:22this is one of this now runs our entire
9:24business. And we just produced a summit
9:27and got it live last week. And I'm going
9:30to show you a screenshot from in that
9:32last week where most of the work gets
9:33done for that summit. If you see this is
9:35an insights from our N account, we've
9:38saved 52 hours in the last week up 19
9:42hours from the week before because of
9:44the system that we've built. Now, what
9:46how did I manage how did I measure that?
9:49Basically, you just add in how long it
9:51would normally take you to do things
9:52manually if you were to do this with a
9:55chat GBT CL dashboard, a clawed
9:57dashboard, having to post the content
9:59yourself, having to move it through
10:00different systems yourself. I added in
10:03how long we would normally take to do
10:05those things and then every time the
10:07system runs it then logs the time saved
10:10for that workflow. So 52 hours is more
10:13than a wage that we're saving per year
10:15now just for a single system around
10:18creating virtual summits. So what does
10:20that system look like? Basically it's a
10:22pretty huge system. It's taken me a
10:24little while to develop, but now we can
10:26handle three to four times more work
10:29with this system using the same team
10:32that we had three or four years ago
10:34doing one summit per month. So basically
10:36what we could do now with this system is
10:38we could build four maybe even eight
10:40summits per month with six people
10:43compared to one summit. So that means
10:46potentially four to eight times more
10:48revenue for the brand. And also it means
10:51four to eight times more revenue with no
10:53extra operational costs which is pretty
10:55huge. And this is just a small niche
10:57business. We're not doing anything
10:58crazy. We're not developers. We're just
11:00using the tools that we have available
11:02and thinking about them in a very smart
11:05way in order to produce maximum
11:08productivity for the entire business and
11:10team. So how does this work? Basically
11:12we have all of these different linked
11:14records set up. So it links it to the
11:16right summit that links it to the right
11:18guest name. And then we upload the
11:20content to YouTube. Once it's been
11:22uploaded, it takes a little while for
11:24YouTube to process it and transcribe it.
11:26And then once it's transcribed, then we
11:28hit import YouTube transcript. This is
11:31the basis for all of our automations in
11:33this workflow. So we just drop that in
11:35down below. You can see here it's
11:36timestamped. And then that allows us to
11:38do everything else that we need to do.
11:40So number one is because we're uploading
11:42it to YouTube. Then we want to generate
11:44all the things we need for YouTube. We
11:45need a title. We need a description. We
11:47have a default description that we add
11:49in. We need some timestamps for the
11:51video. We need a first comment call to
11:53action. And we've got some key points
11:55here that we use on the website for
11:57this. So in order to do that, I have a
11:59prompt library attached in this system.
12:01Now if I open up this, here's our prompt
12:03for the description. Here's our prompt
12:06for all of the different timestamp.
12:08Here's a prompt for the thumbnail titles
12:10which get generated and so forth. So,
12:12we've got all of our prompts saved in a
12:14prompt library and we simply just import
12:15them into this workflow. Choose the
12:17model that we want to run. So, we're
12:19going to run Claude Sonnet 4.6 and then
12:22hit this button here. Within 30 to 60
12:24seconds, it then drops in the output
12:27over here. So, here's our title
12:28variations for the video. Here's our
12:30description for the video or here's
12:32everything else for the video.
12:33description, timestamps, title for the
12:35thumbnail, first comment, and key
12:37points. So, all of that gets generated
12:40based on the YouTube transcript that
12:41we've added in based on these prompts
12:44and then based on some extra brand
12:46context that we give it for this
12:47particular brand. From there, we add all
12:49of these into these fields and then I
12:51hit this button here and uploads it
12:54directly to YouTube. So, it fills in all
12:56of the content on YouTube. So, I don't
12:58need to go into YouTube and structure it
13:00all. So, that's number one. Next, we
13:03remember we then had thumbnail titles.
13:05So, we come down here and we add in as
13:07many titles as we want into this section
13:10here, split it by a comma, and then that
13:12will now generate thumbnails. So, how do
13:16we generate the thumbnails? We have
13:17another database where we're collecting
13:20thumbnails from all of the biggest
13:21podcasts in the world. So, we've got the
13:23J Shetty one, we've got Chris
13:24Williamson, we've got Mel Robbins, the
13:26Diary of a CEO. We're pulling in all of
13:29those as a training image. And then in
13:31the prompt in the back end, which you
13:33guys can get access to in our community
13:36here. So remember, I'll give you access
13:38to the thumbnail titles and the images
13:40and so forth. So the prompt in the
13:42background then says, take this style
13:45here, reference it, and then replace it
13:49with our speaker head shot, replace it
13:51with these thumbnail titles here, and it
13:54generates all of these titles,
13:56thumbnails here that we can then choose
13:58from. So again, this is a single click
14:00of the button. It goes off for image
14:02generation takes a little bit longer. So
14:04it might take two or three minutes to
14:05generate and then we end up with all of
14:07these that we can then choose from. And
14:09it may get only 80% of them, but for 8
14:12cents an image, then generally what I'll
14:14do is I'll generate three versions per
14:16thumbnail. And then I'll choose the ones
14:19that are correct in terms of the style,
14:22in terms of the writing, in terms of the
14:25speaker image. I just you just choose
14:27from them all because what's the
14:28alternative? The alternative is you go
14:30into Canber and you manually create all
14:32of them. You're paying somebody $10 an
14:35hour or you're doing it yourself and
14:37your time's worth way more than that. So
14:39the other option is you spend way too
14:41much time in there trying to get it
14:42perfect. Whereas now we just generate a
14:44whole bunch and choose the best ones
14:46from a big bunch and then it's done. And
14:48I'm also not sitting here waiting for
14:50three or four minutes for all these to
14:52generate. I'm then continuing on this
14:54workflow and doing all other parts of
14:56the workflow. So, I literally spend
14:59maybe 30 seconds generating the images
15:02here that I get to choose from for the
15:03thumbnails. We then have a social media
15:05post workflow which we're optimizing a
15:07little bit more. We have one here called
15:09session resources. So, we pull out all
15:11of the different resources mentioned in
15:14our tools and then we can upload them
15:16into different systems for we make it
15:18easy for our members to get access to
15:20these tools. We also then have we're
15:23building an integration that will
15:24connect it to our tools review website
15:27that then pulls in the affiliate links.
15:29It will soon then create different
15:31resources that we can use for lead
15:33magnets and so forth. We have another
15:36workflow here where we can generate
15:37other topic ideas based on this topic.
15:40We didn't run this one for this example,
15:41but we could do that. And now we can get
15:43to some really cool stuff, which is like
15:45session notes. So again, like all of
15:47these things are getting created with a
15:49click of a button. I'm for me to process
15:52all of this content, it takes less than
15:5320 minutes. The we can generate
15:55infographics. So again, this infographic
15:57gets generated for this. We use it as a
16:00a bonus download for our members. We use
16:03it as a lead generation tool to generate
16:06leads from this event, this interview.
16:08We then create this ebook here. So this
16:11requires a little bit of an extra step.
16:13So, you can see here this book provides
16:16superdetailed notes and frameworks and
16:19step-by-step guides. It also creates
16:22some goals and targets. It creates an
16:24action checklist. So, we're building
16:26these in order to help the audience
16:28start take action from the content that
16:31we're creating. Like, we consider these
16:33workshops. We don't consider them
16:34interviews or or like things that you
16:36just sit back and passively watch. Like,
16:38these are high value workshops that
16:39we're creating and we're throwing on
16:41YouTube. So we generate that by creating
16:43a HTML file. We download all of those
16:45things and upload it into Claude in
16:47order to create this final PDF. There is
16:49there's been no design that has been
16:51done by this. There's been no prompting
16:53that has been done by this. 100% of this
16:55has been done from the AI automation
16:57with just one final manual step. We can
17:00also then generate carousels. So we have
17:03an option here to generate this carousel
17:05where it'll generate a hero slide and a
17:07call to action slide. Plus then we just
17:10do a quick video a couple of video clips
17:12just inside of Dcript and it takes us
17:15maybe two minutes to find these and just
17:17do a quick edit and we do it tell it's
17:19telling us the exact timestamps we need
17:21to use and it creates this entire
17:22carousel for us. So, when we post this
17:25online, there's a hero image, video one
17:28that looks like this, video two that
17:30looks like this, and then our call to
17:32action slide here at the end where we're
17:34asking people to comment and then we'll
17:36send them a link through to get access
17:39to this talk. Plus, we're not even
17:41scheduling it on Instagram. We schedule
17:43it from here. So, this is connected to
17:45go high level. We use that to schedule
17:47our content. We give it a date and a
17:49time and hit schedule to Instagram and
17:51it just schedules the thing directly on
17:54there for us. So, this is a fast
17:56overview to show you what's possible.
17:58And the whole point of this video is to
18:00give you a paradigm shift into what's
18:02actually possible. If we had to do all
18:04of what I just explained manually using
18:06chat GPT and Claude and the dashboards
18:09and the different tools provided, then
18:10it's probably going to take an hour to
18:12an hour and a half per person, which if
18:15you're not doing too many interviews,
18:16then it's not a big deal. But we're
18:18scaling up to be doing 20 to 50
18:20interviews per month by the end of the
18:22year. So that means we're saving at
18:24least an hour per interview, which means
18:26we're saving at least 20 to 50 hours per
18:29month just on this tiny part of the
18:31workflow that we do in order to create
18:33this type of content. So again, the
18:35whole point of this video is to help you
18:37shift your mind into what's actually
18:39possible and what's coming and where you
18:41want to move to. And if you are
18:43interested into diving into more of
18:46types of concepts, then you can sign up
18:48for our AI content mastery summit. At
18:50the very least, you can join for free,
18:52but there's some great premium tools in
18:54there as well after you do sign up. But
18:56100% of the content is free. And a lot
18:58of these experts are just going into
18:59deep into how to shift your mindset when
19:02it comes to building business. Like the
19:04very first session here by Rosco is
19:06incredible session about how he's
19:08shipping three apps per week these days
19:10because he has built the system in order
19:14to move so quickly using these tools
19:16like lovable and claw code and so forth.
19:19So there's a lot of great content that's
19:20happening here. I'd love to see you
19:22there. If you prefer to wait for this
19:23content to come out on YouTube, then
19:25that's possible as well. Just make sure
19:26you subscribe to Big Moves AI and you'll
19:29get all of the updates over there very
19:31soon. So, I hope this video is valuable
19:33and I'll see you in the next