Ep.230 – The AI Inference Flip, Part 2: Why Cheap Intelligence Makes You Indispensable

Josh Lupresto (00:00)
So a few weeks back, I did an episode called the AI inference flip and how this whole industry pivoted from building intelligence to running it fast and and and and cheap and everywhere. And I got an email from one of our top partners and it honestly reframed the the the the whole thing for me. He said the real takeaway for him was that AI is shifting from just building intelligence to continuously producing and distributing it.

Seems pretty basic. And and then he wrote this line to me that really got me thinking. He said these. He said, the industrial revolution made mechanical power abundant. Fair. The electrical age made energy abundant. The internet made information abundant. And the cloud made computing abundant. See the common theme here? And AI may be the thing.

That makes useful intelligence abundant. Really cool thought process. So if that is true, and I think that it is, then AI stops becoming a product that you buy and it becomes an underlying utility. we’ve started to see this word utility tossed around with AI, right? And we’re getting there. So you think like power, you think bandwidth, something that’s just there, it’s cheap, it’s everywhere, and that you can build on top of. So today

Is part two. Last time we did the technology. Today we do the money, where abundant, cheap intelligence actually creates some business value. And then how you, the advisor, help the end customers find it. So let’s get into it.

Welcome. We are at part two of this. you heard part one, the AI inference flip. Now we got some more stuff for you. So today, we are jumping right back in, I think, in the most important curve in the business right now. So welcome back. I’m your host, Josh Lupresto SVP of sales engineering at Telarus and by popular demand and a little bit of inspiration, here we are. So let’s put a number on this abundance idea because it’s kind of staggering and it’s

Most people’s mentals mental model of this is kind of years out of date, right? So let’s let’s back up a little bit. So GPT-4 launches early 2023. Back then, running a million tokens, and just think of a token as a a word in this in in this example, it costs about 30 bucks. Now here we are, let’s say roughly the middle of 2026, and you could get that same quality or better from open models for under 50 cents.

So some of the analysts, if you read, are putting this drop a thousand times over three years for this kind of fixed level of capability. So let me say that again. Let me let me reframe that just just a little bit. So the exact same unit of intelligence that cost 30 bucks is now costing pennies. So a thousand fold collapse in three years. there’s a good venture firm out there, the Andrews and Horowitz, the A16Z guys, they gave it a name because we need a new name and we need a new acronym in this space, LL Inflation.

So it’s the inverse of inflation. Every year, the same intelligence gets dramatically cheaper. So there’s nothing else really, as you kind of looked around and I thought about this. There’s nothing else really in business that’s behaving like this. Your rent doesn’t fall a thousandfold. We wish it would. your payroll doesn’t. bandwidth got cheaper over decades. This happened in three years. This is the fastest cost decline of any commodity in the history of computing. And here’s the

mental shift that I want to make. The one that that our advisor that sent this over, I think kind of nailed. When something gets powerful, or when something powerful like that gets so cheap that fast, it stops becoming a specialty product and it becomes infrastructure. Nobody buys electricity as a product with a pitch deck. You just plug it in and it’s there. and you you you build a business on the assumption that it will be there.

So intelligence I think is on that exact same path. So soon your customers won’t buy AI. They’ll just assume cheap intelligence is available on tap. And then the question becomes what they build with it, right? That’s where we all come in.

And and that right there, I I think this starts to become the whole ballgame for us because when a powerful thing becomes cheap and abundant, the value doesn’t disappear. It just moves. It moves up to whoever figures out where to apply it. And that’s where I think the advisor’s job is. Okay, so let’s get into the next part here.

Let’s call this the paradox, right? The cheap intelligence, the bigger bills, all that good stuff. Now, I gotta be honest with you. So i i if you walk into a customer and say AI is basically free now, a sharp CFO is gonna go, hmm, I don’t think so, buddy. Let me show you these bills. Because the AI bill is it it’s going up, right? It’s not going down. So how does a price collapse a thousand fold and how does the bill still go up? This is the

single most important thing I think to understand in this whole episode. And it has a name. it’s funny to hear it coming back up again. We talked about this before in in in other presentations. I don’t know if we’ve done it on this podcast or not. But economists call it the Jevons paradox. All right, so so here’s the idea. because I know you were excited to to wake up and listen to a somebody talk about an 1800s economist, but it it makes sense. It all ties together. So just roll with me here.

So back in the 1800s, when steam engines got way more efficient at using coal, people assumed England would use less coal. Well, the opposite happened, if you think about that. because coal power got cheaper, they found a thousand new uses for it. And the total coal use exploded. Great, great for business. So when something something useful gets cheaper, we don’t use less of it. We use dramatically more.

That’s exactly what is happening with intelligence. The price per unit collapsed. So companies went from asking it a simple question here and there to running AI agents that fire off dozens of calls per task all day across the whole business. So one analyst estimated that token consumption growing something like 24 times by the year by by the year 2030. So the price falls off a the price per unit falls off a cliff, and then that usage.

Rockets past it and the total bill goes up. So here’s why this matters for you. and it’s good news, right? the the Jevons paradox thing means the demand for help does not shrink as the intelligence gets cheaper. It grows. So in in in that token, if you understand that, every customer is about to use vastly more intelligence. We’re seeing it every day and in more corners of their business than they ever imagined, right? As it becomes more affordable. So

More usage spreads across more of the business means more complexity, more decisions, more architecture, more wait, where should we be using this and where are we wasting it, right? That to me is is advisor value. That is not less. So the framing for your customer is not AI is cheap now, because it’s not quite yet. It is intelligence just became cheap enough to put almost everywhere. So this entire game.

Is now about putting it in the right places. Okay. cheap doesn’t mean simple either. Cheap just means everywhere, and everywhere means somebody’s got to make some good choices here. And I think that somebody in this equation is you. This is where we help. All right. Let’s go let’s go part three. let’s talk about where this abundance of intelligence creates value. Notice how many word how many times we probably say abundance today.

Okay, so let’s get to the the advisor that inspired this episode’s real question. The one that he said is is is where it comes alive for him. So, where does cheap, abundant intelligence actually create most value in a business? Because use AI is a you know useless piece of advice. Use it here and for this reason. And here’s the return is a consulting engagement. So let me give you a simple way to think about it.

abundant intelligence creates value in three main places. And you can walk a customer through all three. So number one, it’s cost. Taking work that used to require a person and letting some cheap intelligence carry the routine part. Okay. Think document processing. Think first line data entry. That’s the boring high-volume stuff. This is the obvious one. Everybody sees that. It’s real.

But it’s also probably the most crowded and one of the least differentiated. It’s fine as an entry point. Now the second, this is where it gets a little more interesting, is the speed and scale, doing things that were always technically possible, but they were too expensive to do at volume. reviewing every single contract instead of a sample, reviewing every customer support ticket for churn signals instead of just, you know, a little spot check here or there, and then personalizing outreach.

To tens of thousands of customers the way you used to for 10, right? And I’m I’m sure that this is gonna equate to the LinkedIn DMs get a lot more specific and a lot more relevant. So when intelligence is nearly free, you stop rationing it. And the things that were economically impossible last year, they kind of pencil out now as we go forward. That’s not cost cutting, that’s just doing things that you literally could not afford.

To do before. And then the third, the one with the most upside, I think, is the new capability, things that were just not possible at any price before. So a small business getting a level of analysis that used to require a team of a consultants, big money. So real-time translation to opening a new market, for example, a product that adapts maybe to each user.

this is where cheap intelligence doesn’t make the old thing cheaper. It makes it a a genuinely new thing exists, which is cool. Now, here’s the advisor move for you, I think, that ties it together. So your customers’ instinct will be to run to bucket one, cut a cost, automate a task. your value is walking them up the ladder to buckets two.

And buckets three, where that real money is. Anybody can point at a cost. Very few people can kind of sit there with a business owner and say, forget shaving the expense for a second. What could you do now that you couldn’t do at all a year ago? That question’s worth a fortune, right? We don’t hear it asked enough. And it’s a question no vendor selling them a single product is ever going to ask them, right? So you’ve got that advantage, no matter.

How many FDEs, you know, forward deployed engineers from the foundational models are in these accounts or or or whatever, it doesn’t matter. So let me let me connect this back maybe to the abundance idea because it’s the same lesson every prior wave has kind of taught us. So when the cloud made computing abundant, the winners weren’t the people who just moved their old servers to save a few bucks. The winners were the ones who used cheap, unlimited computing.

To build the things that couldn’t exist before, the Netflixes, the Ubers, et cetera. Cheap electricity did not just lower factory bills. Think about it, it created the assembly line. That’s huge. The abundance always creates more more value than the savings. Point your customers at the abundance.

All right, let’s go to part four here. I I always love questions, right? So let’s give you some questions that start to connect these dots a little bit and and and let’s make this usable. So our our advisor that that made some suggestions for this episode, he said that the part that is gold is how to help customers connect the technology to the value. So here are the questions that do exactly that. Write it down, replay it.

Market is your favorite. Don’t forget to go rate the podcast and comment. That always helps. all of those are acceptable. So start by finding the rationing. Okay. Ask what is something valuable that you would love to do for every customer for or for every transaction, but you can only afford to do it for a few. That question kind of goes straight to bucket two, the stuff that was too expensive to do at scale.

So when intelligence is cheap, the rationing goes away. And that’s where you look like a genius. Then maybe find the expensive expertise. Ask where in your business do you rely on scarce, expensive human judgment that creates a bottleneck? I love that one. So wherever there’s a cue waiting on one expert, there’s a place cheap intelligence can widen that pipe. That’s real. That’s

It’s quantifiable value. And then go for the new capability, right? Go for the fun stuff. Ask if if if analysis and and expertise were just nearly free and instant, what could you build or offer that you can’t build today? This I this is one of the funnest questions ever to ask. This is a different version of it than we used to ask, right? Because I think the impossible is exactly that, but it’s not. This is kind of

Bucket three, right? It it most customers have never once been asked it. watch what happens when you do. It’s fun. And then I think ground it in returns. Ask, you know, of of of everything that we just talked about, which one, if it worked, would actually move a number your CFO cares about this year or your COO or who whoever it may be. This is how I think you you you keep it from being a science project. You connect

The shiny idea that’s a real, you know, over to a real owned business metric. Maybe it’s KPIs, you know, whatever it may be. That’s the difference between cool and funded. and we care about funded. Now, just like all the other sessions in this, notice the arc. You’re not selling AI. You’re helping them discover where abundant intelligence connects to actual business value. And then you’re just prioritizing it by return how it returns.

That is the exact skill that our advisor was looking to figure out how to build. And I think that’s the one that makes you indispensable in the AI area. I can’t say this, I I can’t say this enough. the harder this gets, the more they need help and the more they need this kind of coaching. All right, let’s go. This is part five. you connect that.

value and and and we’ll help you build the how. So let me explain kind of what I mean by that. So here’s the real handoff. Same as always, right? You don’t have to be the person that architects the model, the routing, the designs, the deployment. Your genius is in this conversation that we just walked through, right? You’re in those conversations. We are not, to start sometimes. Find out where that cheap intelligence meets the real value and let’s prioritize it as it returns, right? That’s the hard human

high value part. So once you have found the where and the why and the how, that’s exactly what our sales engineering team is here for, right? Which models running where? What costs? What are they integrated to? What do they already own? You know, all that architected so the bill doesn’t blow up in the way that we talked about in the, I think it was part one. When the customer says, okay, I get where this creates value. Now how the heck do we actually build it? That’s your queue to bring us in.

You connect the technology to the business value. We’ll connect the business value to this working system that we have to design. And look, today that’s the whole journey from a cheap, abundant utility to real results in your customers’ PL. and that’s something no model vendor and no website is ever going to do for them.

Okay, so let’s let’s get our crystal ball out. It’s getting harder and harder to see through. It’s getting foggy, but let’s let’s look ahead a little bit. So let me let me bring it home, right? So thanks to our advisor for the email that obviously that that that started this thread. because I think it’s it’s it’s necessary to to continue to connect these dots, right? And again, just like he gave us feedback, I would love your feedback on do you want to see different parts of these, do wanna go deeper in a certain area, those types of things.

every if you if you think about this, every wave of abundance in history followed the same pattern. Cheap mechanical power, cheap energy, cheap information, cheap computing, and now here we are on our silver platter, cheap intelligence. Now, every single time the winners were not the ones who used the cheap new thing to save a little cost. The winners were the ones who saw that abundance and used it to build something new.

The savings are just the small prize. The abundance is the big one. In every one of those waves of of the past, there was a person who helped the business figure out where the new abundance actually mattered for them. And in this wave, in our world, that person is ding ding ding, the trusted advisor, right? That’s you. your job is is is shifting from.

Talking about products and helping customers find where this, you know, nearly free intelligence connects to to how it does that, how it how it pulls that real value in their business. And there’s never been a more valuable job to have, in my mind. This is hard and it’s really, really hard for your customers. So as a you know, you y you kind of look ahead a little bit there, the price of intelligence is just going to keep falling. It is. And the uses are going to keep multiplying.

And then the gap between the companies that apply it well and the ones that don’t is gonna become the whole ballgame. That gap, I think, it’s it’s your opportunity. So this week, go and ask one customer that let’s say that bucket three question. If expertise were nearly free and instant, what would you build that you can’t today? And just stop and just listen. That’s it.

That is the show for today. Pretty quick. But look, if you i i i if part one and and part two helped helped you see this differently, do what our advisor did. Send it to someone that needs to hear it, talk to the customers, and figure out where we can help.

I’m Josh Lupresto This has been Next Level BizTech, part two of the AI inference flip. And we’ll see you next time.