Latest / AI Fire Daily / #492 Neil: How To Make Money With AI While Others Just Waste It
Transcript
- 0:00So everyone is playing with AI right now. Yeah,
- 0:04but almost no one is actually making real money
- 0:07from it. Why is that? Because they're selling
- 0:10the technology itself. They're not selling the
- 0:13actual outcome. Right. Welcome to the Deep Dive.
- 0:16I'm glad you're here with us. Today we're unpacking
- 0:18a really specific framework. We're looking at
- 0:20how to transition from just from staying busy
- 0:24with AI tools to actually profiting from them.
- 0:27Exactly. We've got a lot of ground to cover for
- 0:29you today. We're going to explain why prompt
- 0:31packs are a complete dead end. Yeah, they really
- 0:34are. And then we'll show you how to structure
- 0:35your inputs using the MAPS framework. Plus, we'll
- 0:38get into the rule of R for automation. And we
- 0:40also need to talk about the warning signs. You
- 0:42know, what happens when you scale up with AI
- 0:44agents. But I think we have to start with the
- 0:46biggest trap in the AI space right now. Oh, absolutely.
- 0:49Which is this... this deep obsession with the
- 0:52tools themselves. It's everywhere. I mean, people
- 0:55are obsessed with the shiny new toys, and they
- 0:58completely miss the business value. It's like
- 1:00they're hypnotized by the tech. They really are.
- 1:03You see everyone trying to sell these like massive
- 1:06prompt packs or basic chat GPT tutorials. Even
- 1:10automation templates kind of fall into this trap.
- 1:13I mean, technically there is a market for those
- 1:15things. Sure. Yeah. But the ceiling is brutally
- 1:18low. Like competition is just overwhelming because
- 1:21everyone has access to the exact same information.
- 1:24The real money always sits one level deeper than
- 1:28the tool itself. Right. It's about finding a
- 1:30broken process. Precisely, because business owners
- 1:33don't care about language models. They have a
- 1:35costly process, right? Say they're losing warm
- 1:38leads because their follow -ups are terrible.
- 1:40They're bleeding money. They just want the bleeding
- 1:42to stop. Exactly. Whoever fixes that specific
- 1:45outcome gets paid. And they can get paid well,
- 1:48regardless of the tools they use. The tool should
- 1:50be a... Totally invisible. Yeah. I like to think
- 1:53of it's like stacking Lego blocks of data. The
- 1:55client doesn't care about the plastic bricks.
- 1:58They just want to buy the finished castle. That
- 1:59is a perfect analogy. Let's look at IG Group
- 2:02from our source material. Massive global financial
- 2:05services company. And finance means heavy compliance.
- 2:08Incredible compliance, yeah. They had strict
- 2:10regulatory requirements and they needed multi
- 2:13-language content fast. They were paying insanely
- 2:16high agency costs just to keep up. So what did
- 2:18they do? They deployed Claude. quietly across
- 2:21their teams. And the results were wild. The analytics
- 2:25team saved 70 hours a week. Marketing saw triple
- 2:28-digit speed -to -market improvements. Wow. Yeah.
- 2:31Full ROI in under three months. But here's the
- 2:35kicker. Claude was just the invisible engine.
- 2:38They didn't sell AI to their customers. They
- 2:40just delivered better financial services. So
- 2:42looking at that, what are the three conditions
- 2:44for spotting a real opportunity? You're looking
- 2:47for a task taking skilled people a long time.
- 2:49The final output has to be very high stakes,
- 2:52and the business must already be spending money
- 2:54to solve it. Look for manual high stakes tasks
- 2:57where businesses already spend money. Exactly.
- 3:00Right. So once you identify the right high stakes
- 3:01problem, you actually have to get the AI to do
- 3:04the work. And if your output is generic, you
- 3:06know, the model isn't broken. Your prompt is.
- 3:09Yeah, your prompt is definitely broken. Vague
- 3:11prompts just give you vague answers. Which brings
- 3:13us to the MAPS framework. Let's break this down
- 3:16for you. M is for mission. You have to give the
- 3:19AI the actual business goal, right? Right. Not
- 3:22just the task. So instead of just saying, find
- 3:24me leads, which is terrible. Because it has no
- 3:27context. Exactly. You say, I need 30 customers
- 3:30to hit my $15 ,000 revenue goal. That's the mission.
- 3:33It grounds the AI. Beat. Then A is for ask. Make
- 3:37one specific request. Right. Don't say, help
- 3:40me with sales. Say, give me a list of 40 U .S.
- 3:44leads with name, company, email, and LinkedIn.
- 3:47Precision is key here. It is. P is for parameters.
- 3:50This is the background context. Your ideal customer,
- 3:53your constraints. And honestly, a massive time
- 3:55saver here is using voice input to just dictate
- 3:58those parameters. You can just talk it out. I
- 4:00have to admit something here. I still wrestle
- 4:02with prompt drift myself when I rush and skip
- 4:05the parameters. It's really easy to get lazy.
- 4:07Oh, we all do it. But skipping parameters guarantees
- 4:10garbage output. The context is what makes it
- 4:12work. Yeah. And the last one is S shape. Tell
- 4:15the AI exactly what the output should look like,
- 4:18like format it as a table. Right. And I have
- 4:21to ask you, why is shape the step everyone seems
- 4:24to forget? Skipping it means wasting 20 minutes
- 4:27reformatting a useless wall of text. It really
- 4:30does. It's such an unforced error. Absolutely.
- 4:33We're going to take a quick break here. Don't
- 4:34go anywhere. Insert a clear break here for the
- 4:37mid -roll sponsor read. All right, we're back.
- 4:40So MAPS can use you perfect single outputs. But
- 4:43for tasks that happen every single week, running
- 4:45manual prompts just, it isn't enough. You need
- 4:49automation. But only if it passes the test. Right,
- 4:52the rule of R. We use tools like Zapier or Make
- 4:54for this, which is basically digital glue that
- 4:57automatically moves data. between your different
- 5:00apps. That's exactly what it is. So the first
- 5:02R is repetitive. Does the task happen regularly?
- 5:05Like manually checking ConvertKit every Monday.
- 5:07Yeah, and then pasting those email stats into
- 5:09Notion, that's highly repetitive. The second
- 5:12R is rule -based. Meaning the logic is perfectly
- 5:14predictable. Right. Say a Typeform lead comes
- 5:17in, it goes straight to your HubSpot CRM. A CRM
- 5:21is a digital filing cabinet tracking all your
- 5:23customer relationships. Right, so it hits HubSpot,
- 5:26then triggers a ConvertKit email. The logic never
- 5:29changes. And the final R is return. Will it save
- 5:32you at least one hour a week? Because if yes,
- 5:35the setup effort pays back in a month. Which
- 5:37is huge. So what's the danger of ignoring the
- 5:40return rule? You'll spend days building a machine
- 5:43nobody needs. It's a trap. Spending 60 hours
- 5:46to automate something that saves two minutes
- 5:48is a bad trade. An incredibly bad trade. So what
- 5:52happens when a problem is too complex for a single
- 5:55prompt or, you know, rigid automation? That's
- 5:58when you graduate to AI agents, systems that
- 6:00can actually make decisions mid -task. They adjust
- 6:02on the fly. Exactly. Look at Klarna. They built
- 6:05an OpenAI customer service agent. In its first
- 6:08month, it handled 2 .3 million conversations.
- 6:11Whoa. Imagine scaling to 2 .3 million queries
- 6:15in a single month. It's mind -blowing. It cut
- 6:20resolution time from 11 minutes to 2 minutes,
- 6:23and repeat inquiries dropped by 25%. That saved
- 6:25them a ton of money, right? Service costs per
- 6:27transaction fell from 32 cents to 19 cents. But
- 6:30there is a real warning here. Klarna eventually
- 6:33had to rehire human agents. Prioritizing cost
- 6:37over quality really backfired on them. Right,
- 6:40because agents can hallucinate. You need humans
- 6:42in the loop to review drafts. How do we prevent
- 6:45these long agent workflows from hallucinating
- 6:47or going off track? Well, you have to engineer
- 6:50the process carefully. Build checkpoints where
- 6:53a human or second agent reviews the work mid
- 6:56-process. That's exactly it. A second agent checking
- 6:58the first agent is a game changer. It's a fascinating
- 7:01architecture. So you have this toolkit now. Claude
- 7:05for Thinking with MAPS. Zapier for scheduled
- 7:08tasks with the rule of R, and agents for complex
- 7:11workflows. The final piece is knowing how to
- 7:14charge for this superpower. Yeah, and this is
- 7:16where people mess up. You have to become the
- 7:18orchestrator. You know which tool fits which
- 7:20problem. But there's a big pricing mistake happening.
- 7:22Massive mistake. When AI makes delivery faster,
- 7:26people instinctively drop their prices. But a
- 7:29lead gen service worth $5 ,000 before AI is still
- 7:32worth $5 ,000 after... Because the outcome hasn't
- 7:35changed. Exactly. Klarna didn't pass all their
- 7:3740 % savings back to the customer. They kept
- 7:40the margin. You anchor your price to the client's
- 7:43outcome, not the hours you spent running, Claude.
- 7:46Let the technology stay invisible. Totally invisible.
- 7:49Why is it so hard for creators to keep their
- 7:51prices high when AI does the heavy lifting? Because
- 7:54of guilt. Honestly, it feels like cheating. They
- 7:57wrongly tie their self -worth to hours worked
- 7:59instead of the final results. Exactly that. You
- 8:01have to break that mindset. Let's synthesize
- 8:04this. To actually win with AI, stop selling your
- 8:08knowledge of the tools. Find expensive problems.
- 8:11Master your inputs with MAPS. Filter your busy
- 8:15work through the rule of R. Use agents for big
- 8:17workflows, but keep humans in the loop. And above
- 8:20all, price for the outcome. I want to leave you
- 8:23with a final thought to mull over. If the ultimate
- 8:25goal is to let the AI technology remain completely
- 8:28invisible to the client, What happens in five
- 8:31years when every single business has the exact
- 8:33same invisible AI capabilities? What becomes
- 8:37your ultimate differentiator then? That's the
- 8:39real question. Thank you for joining this Deem
- 8:42Dive. Look at your own workloads today and just
- 8:44find one task that passes the rule of R. Start
- 8:47there. OUTRO music.