The $80B AI Write-Down: Why 95% of Projects Fail to Prove Their Worth

Josh Lupresto (00:00)
Let me give you some numbers. Two numbers that just landed and they tell the whole story of where AI is actually at right now. So number one, in just the first half of this year, the most valuable company on the planet, Anthropic, the makers of Claude, who we can’t live without, is heading into one of the biggest IPOs in history. Okay. They posted a net loss of about 80 billion dollars. 80 billion in six months. So

They’ve committed to spend over half a trillion dollars on computing power in the years ahead, right? Half a trillion. It’s crazy. All right, so here’s number two. here’s kind of where all that staggering capability is landing, right, in the real world. So study after study this year found that around 95% of AI projects produced no measurable result. 95%. That’s a problem. And we’ve kind of been waiting for this, right?

There was this initial, my gosh, it’s so easy, it’s just going to implement itself. Well, turns out data matters, numbers matter, and ninety-five percent. no nothing from a measurable result perspective, right? On the bottom line, that’s even worse. So look, the models are brilliant. The business results, they’re kind of missing. So if you put those two numbers together and you get the real headline of

The moment that we’re talking about today. So AI is not some unstoppable force that’s quietly taking over. It is a jaw-dropping technology that still has to do the hard, unglamorous work, which is prove its worth. And, you know, really that it’s worth the money that people are paying for it. And of course, it’s got to prove the value. It’s got to connect technology to a business result. And I think that is the single most valuable thing.

that a trusted advisor does. And that’s exactly what we’re gonna talk about today. So shall we get into it?

Welcome back, everybody. Next level biz tech. I’m your host, Josh Lupresto, SVP of sales engineering at Telarus. Today’s episode is titled The $80 billion write down, right? AI is not just taking over until it proves its worth, right? it’s got to add a lot of value. There’s a lot of hard work to do here. So this episode is less about you know, a specific gadget per se. And I think it’s more about

the business reality underneath all of the AI hype. I just want to uncover that a little bit more because I think the reality is where you make your money, right? This is where the magic happens. So let’s start with let’s start with that crazy anthropic number because I think the headlines made it sound like, you know, the apocalypse. And it’s it’s actually more interesting than that. So yes, they lost about $80 billion in six months.

I don’t know about you guys, but if I lost $80 billion in six months, I might have a different job. But hey, somehow they’re fine. So look, here’s the nuance though that you need because you’re gonna be this calm, informed person in the room, that just brings this sense of reason to the conversation. Okay. So the vast majority of that, around I think 72 billion of it roughly, that wasn’t cash going out the door. It was an accounting charge.

Tied to the rising value of the funding instruments that they have issued. Okay. So basically, the company became so much more valuable that the accounting rules forced them to book a giant paper loss. Strip that out, right? That that was the real operating loss, you know, maybe closer to 9 billion. Still very enormous, but a very different story.

And on the other side of the ledger, I think it’s kind of wild, right? Their revenue went from one billion in the first half of last year to over 16 billion in this year. That’s one of the fastest growing business stories in history. 80% of it from its enterprise customers. So so the honest read isn’t that AI is dying, it’s that AI is simultaneously the fastest growing and the most expensive thing.

That anyone has ever built. Okay. So revenues exploding, costs are exploding, even faster. Okay, that’s a problem. They’re even spending half of a trillion dollars, betting that that value will eventually dwarf the cost. Crazy. Now, here’s why that matters to you and to your customers. Okay. When the people building the technology are spending at that scale, that cost has to get paid for somewhere. Okay.

It it gets paid for when real businesses, your customers, get real measurable value out from that AI, and I think are willing to pay for it, right? that entire trillion dollar AI bet rests on one thing, that this stuff actually is delivering value in ordinary companies. And right now, I I I think the data says that most of them haven’t figured it out. Still a work in progress. that gap is

Kind of this whole episode that we’re going to talk about today. Okay. let’s get into part two. Let’s the the the proof here. Okay. So ninety-five percent can’t show the money. Let’s unpack, let’s unpack a little bit about what that means. Okay. So I I think that’s the most important stat in enterprise technology right now. And I say almost none of your customers have really internalized it yet, but everybody’s starting to become you know, or be held accountable for these things.

So if we look at multiple big studies this year, MIT, McKinsey, KPMG, et cetera, these all landed in the same place. Something like 95% of enterprise AI pilots are showing no measurable impact on the profit and loss statement. That’s crazy. So only a tiny slice, I think it was like five to eight percent, are getting real at, you know, like real, real scale in financial returns. Okay. So McKinsey.

Call McKinsey for example here. So McKinsey found McKinsey found that fewer than four in 10 companies could attribute any profit impact to AI at all. Okay. Now, I want to be really careful on this part because this is easy to kind of misread. This doesn’t mean that AI doesn’t work. We know it works. The technology works great, but in the demo, it’s magic. And people love it.

The the problem is just it’s it’s it’s something more specific and I think it’s something much more fixable. So here’s the trap. And I think it’s the same trap over and over. A company runs an AI pilot, the team comes back, they’re thrilled, it’s amazing, it saved us two hundred hours a month, right? All this beautiful rainbows and unicorn stuff. And the C CFO asks one simple question. Okay, cool. Where’s that on the income statement? Show me the money. And then the room goes quiet, right?

Because saved hours are not saved dollars. Let that one sink in. Saved hours are not saved dollars. Time saved only becomes real value, I think, in really three ways. Okay. So if you take the cost out, you redeploy that freed up time to something that makes money, or you grow output without growing cost. So if none of those three things happen.

You get a cool demo and a big software bill, right? Great if you’re the software rep that sold the deal, but bad if you’ve got to account for it. That’s the proof problem in a single sentence. Companies are measuring activity, logins, hours saved, prompts instead of outcomes. What do we always talk about? We talk about outcomes. And activity is not value. The CFO knows the difference, and increasingly the CFO is the one signing the checks.

And this is where it gets a little bit structural. And I think where it gets really good for you. Okay. So the studies are showing that you know, it it’s clear that the thing separating the five percent who win from the 95% who stall is almost never the model. It’s the boring stuff. It’s workflow integration, it’s connecting AI to the actual systems and the process of of you know, all these different processes of the business.

It’s picking one specific high value problem instead of trying to boil the ocean, as we say. It’s setting a baseline so you can actually prove the before and after. McKinsey said it, I I think McKinsey said it flat out. He said, the companies seeing real returns are not the ones with the best models. They are the ones who did the integration work. And you know, you see this in some of the earlier episodes, all this talk about why people are deploying forward-deployed engineers.

Because that’s the integration work. That’s the hard stuff. So read what that’s really saying. The hard part of AI was never the intelligence. The hard part is connecting that intelligence to a real business outcome in a real company with real systems, not vibe coding la la land, real systems. That’s not a technology problem. That is an advisory problem. And it’s the one that you in this whole

business model was built for.

Let’s go to part three. Why prove it is this what I’m gonna say is the superpower of the advisors. So let’s connect this all the way to you because I think this is a moment, you know, it’s kind of tailor-made for the trusted advisor. And more importantly, here’s why. For the last two years, AI sold itself on hype. It’s magic, it’ll change everything, just buy it. And a lot of customers did. they bought the tool.

They ran the pilot and then they just kind of hit a wall that we just talked about, right? Now the mood has shifted. The CFO’s in the room. The board is asking, what did we actually get for all this AI spend? The this era of buying AI on vibes, right? Cracks me up. it’s ending. And the era of proving AI is beginning. And proving the value, I think, is the oldest, most

Durable skill that a great advisor has. You don’t sell the shiny object. You connect a technology to a business result and you make the number real. That I think is exactly the muscle that the entire market is suddenly desperate for. I’ve been waiting for this to happen. I’ve been thinking about, wow, you know, this is really gonna show up on the balance sheet before long and everybody’s gonna have to go, what are we getting for that? Right. And and and I think we’re there.

So think about where you sit though for a second. You know, you’re you’re not the you know, you’re not the AI vendor trying to hit a sales number by shoving the model into every corner of the business. You’re not the internal team that fell in love with the cool demo. You’re the new yeah, you’re the neutral person, kind of like right in the middle, who can walk in and ask this unglamorous, incredibly valuable set of questions. what specific problem are we solving?

What does success look like as a number? How does this connect to the systems that you already run? How will we prove that it worked? That is the exact conversation that separates the 5% from the 95%. And almost nobody, nobody inside that customer’s building is having it.

And here’s the part that should kind of get you excited because this isn’t being it’s not about being a wet blanket on AI. It’s kind of the opposite. So the customers who get this right are seeing enormous returns. Some of the agentic deployments that are scoped correctly are reporting very fast, very fast payback, you know, measurable, measurable results, I think in weeks. And look, the opportunity is huge and it’s real. No secret there.

The winners just do it differently. They pick a sharp use case, they integrate it properly, and they measure it honestly, right? No fluff, no BS. So your job is not to slow AI down, it’s to make sure that your customer lands in that 5% and wins instead of the 95% that you know that just kind of spins out there. And and notice how this protects and extends everything that you already do, right?

This isn’t a new product to sell against your book. It’s it’s a lens you put over your in your customer’s entire environment. It’s the connectivity, the security, the cloud, the customer experience tools. AI makes all of it more powerful when it’s connected to a real outcome. So you kind of become this, you become this person who makes sure the customer’s AI spend actually pays off.

Which makes you the person that they call before they spend anything. So in a world where ninety-five percent of AI projects can’t prove their worth, the advisor who can prove worth is the most valuable person in the room. All right. You know where we’re getting to. Part four here in this one is all about the questions. I think this is the this is the set of questions that separate value from the hype. All right.

We’re we’re recording this on a Friday. Let’s think about how to make this usable Monday morning for you. Okay. But you’re gonna see it on a Wednesday. these are the questions that turn or doing AI into real provable projects. And I think they’re the exact questions that the ninety-five percent skip. All right. Write them down, pause it, however you wanna do it. Let your bots scrape this recording and and and convert it to text for you. All of the above are fine.

So start with the proof question. Ask this one. Hey, you’ve got AI in the building somewhere. Can anyone show you where it’s actually moved a number that the CFO cares about? This is the whole ballgame in one question, right? If if they can’t answer it, and in and most can’t, you’ve just shown them the gap, right? You’ve shown it gently, and you’re the one standing there ready to help close it.

And then scope it down, right? Take it, take it one more level. i ask them this. If you had to pick one specific expensive, repetitive problem in your business to point to and to point AI to, what would it be? So so 95%, those the 95% fail that because they try to do everything. let’s let the tool do this and this and this. The five percent win because they pick one sharp thing. You just pointed them at the winning move. That’s all.

Then I’d say test for the baseline. Ask before you turned on any AI tool, did anyone write down what that process cost in time or dollars so that you could prove that it got better? Right. Just get your baseline set. Almost nobody does this. No baseline means no proof ever. That is a real gap that you can help fix. You’re you’re a whiz kid for coming in with that kind of solution. And I

I mean, honestly, it makes you look like the adult in the room in the midst of all this tool purchasing. So then find the integration gap. Ask this one. Is your AI actually connected to the systems where the work happens? Or is it just this kind of smart tool sitting off to the side that people have to go use? That’s what we see a lot of. That I think also is it’s it’s the disconnect. It’s the number one reason that these pilots stall. Naming it is you diagnosing the real problem. And landed on ownership.

Ask this one. Who in your organization actually owns the number AI is supposed to move? And would they stake their name on it? No, no owner, no outcome. Forget it. It ain’t gonna happen, right? If everybody owns it, nobody owns it. I think that question alone reframes AI from this toy into this managed investment, which is exactly how the winners are gonna be treating it. It’s how they’re treating it right now. So think about.

Step back for a second. Think about this whole arc. You’re not selling AI and you’re not banishing AI. You’re helping them do AI in a way that actually pays off. And when they say, okay, help us actually build these and prove one of these out, that is the moment that you don’t go it alone. Of course. You know our sales engineering team is here, right? You’re finding the value. we will help you build it, help you prove it. So look, you’re

You’ve spotted this opportunity. You’ve kind of taken this established base and you’ve asked those sharp questions and you’ve connected it to a business outcome. That is the high value, really freaking hard human part. And that’s yours. You don’t need to personally architect the integration or build the measurements, all of those things, right? The deep work, the integrating AI into the the the systems work actually happens, designing it. you know, the the value in that is it it’s real. It’s not just

you know, hours on a slide wiring up that proof. That’s exactly what the sales engineering team is here to help you do. The suppliers engineering resources are here to do. That’s what some of the scoping engagements are here to help you do. So when the customer says, hey, help us pick the right use case and actually make it deliver, that’s your cue to bring us in. We’ll go as deep as it takes. You got a plethora of people on the engineering team here across the whole solution map. We just want to take that into, you know, take this promising idea into a result. That’s it.

So you connect AI to this business value, right? We’re gonna help build the thing that proves it. Together, we’re not just helping them buy AI, you’re helping them become you know, one of the few people that can show that it was worth it. And I think in this market, that’s the most valuable thing anyone can do for them. All right. Let’s try to let’s try to final thoughts here. let’s let’s look ahead just a little bit. Let me bring it home. So

The biggest AI company on earth is spending half a trillion dollars on a bet that this technology will prove its worth in a real business. And right now, most of those businesses, like we talked about, can’t show that it has. That’s not a story about AI failing. It’s a story about AI entering the grown-up phase where the hype is worn off, and the only thing that matters is whether it actually moved a number. And that is the best.

Possible news for you. Because when the whole market shifts from is it cool to is it worth it? The person who connects technology to the value stops being a nice to have and becomes the whole game. Okay. You’re not in the business of selling AI. You’re in the business of making sure that it pays off. There has never been a better time to be exactly that person.

So let’s look ahead, just a tad here as we kind of wrap this up. watch the IPOs, watch the CapEx numbers because the pressure to prove real world value is only going to intensify. Get ahead of it. Watch the winners pull away from the pack as they kind of figure out integration and proof and watch the, you know, watch this whole show me the ROI become the first question in every AI conversation. You want to be the one with the answers.

And I think if you follow some of this methodology, hopefully this helps make it easier for you. So this week, go pick one customer, ask them the proof question. Can anyone show you where AI has actually moved a number that you care about? Then listen to that pause. That pause is the key to your next engagement. So that’s the show for today. Hey, do me a favor, right? Wherever you’re listening to this.

Follow us on Spotify, subscribe, Apple Podcasts, wherever you’re at, so that you can see them every week. I’m Josh Lupresto, SVP of Sales Engineering at Telarus, and this has been Next Level BizTech.