# $1 of EBITDA

> Why generating real value with AI requires business insight, technical skill, and the human work of lasting change.

_Jack Soslow · 2026-01-01_

Most AI companies can't generate $1 of real value. Not for their customers, not for themselves. The technology works in demos. The pitch decks are compelling. The topline numbers look enormous. But the EBITDA? The actual profit flowing to actual businesses. Almost nobody can point to it.

On January 1st, 2025, the day [Jack Weissenberger](https://www.linkedin.com/in/jack-weissenberger-9b8549144) and I started [Ciridae](https://www.linkedin.com/company/ciridae), we couldn't either.

In February, we flew to Dallas on our own dime to visit a home restoration company. Twelve employees sandwiched between a two-star hotel, a highway, and a storage facility. The owner, AJ, was 64 and wanted to retire and play golf.

We were there because a young financier named Joe had recently acquired the business and given us free rein to transform it using AI. Joe thought the opportunity was in collections. Restoration companies can wait 120+ days to get paid by insurers, and if you could fix that, you could finance the entire industry.

He wasn't wrong. But the bigger opportunity was something else entirely.

AJ had spent 40 years negotiating with insurers on behalf of homeowners. Over four decades, he'd built up an encyclopedia of tribal knowledge: what each insurer would compromise on, what they wouldn't, which line items came as packages, how to cascade a concession in one room across every room in a house. "If you give it to me here, you should give it to me everywhere."

The result: a 45% gross margin on remodel jobs. The industry average is 15%. A 3x advantage, built entirely on tacit knowledge locked in one man's skull.

The problem: AJ was the competitive advantage and the bottleneck. Every deal required him to print out two estimates, his and the insurer's, and compare them line by line. Hundreds of line items, labeled differently, across different rooms. Reconciling the discrepancies, deciding where to refute, crafting correspondence to insurers, all across 10-15 clients simultaneously, consumed nearly all of AJ's time. The company was turning down work because one man literally couldn't highlight faster.

![A printed insurance repair estimate marked with orange, blue, and green highlighter and handwritten adjustments.](/marketing/blog/1-of-ebitda/aj-insurance-estimate.png "wide")

**FIG. 01. AJ's system: orange marked the insurer's mistakes; blue marked line items to push for. Forty years of margin, in highlighter and pen.**

Source: Ciridae field observation; names have been changed for privacy.

We sat next to him for three hours while he worked. We asked questions. We documented the ruleset he'd never articulated, the logic he'd been running on instinct for 40 years.

Then we turned it into code.

In one day, we built an AI system that outperformed AJ on insurance negotiations. It caught line items he missed. It found cascading opportunities faster. It reduced a 3-hour task to minutes.

We'd de-risked the business, captured the tribal knowledge, and removed the bottleneck preventing growth.

And AJ wouldn't use it.

We'd designed it for him. The output looked exactly like his marked-up printouts, same highlighting, same format. We put it in his email so he wouldn't need another login. We sat with him, trained him, walked him through it. Every time he used it, he made an additional $200+ per deal.

He still wouldn't use it.

We had the mandate from the PE owner. We'd found the perfect problem. We'd built a solution that exactly solved it. We'd trained him personally.

We still didn't generate $1 of EBITDA for his business.

Unsatisfied with failure, we went back to Dallas. We refined the product. We sat with AJ until he trusted it. We collaborated until adoption occurred naturally.

Then we did it again. And again. Finance teams for roofers in Columbus. Sales teams for HVAC businesses in Florida. The largest hospital systems in the world. Roofers, dispatchers, lawyers, healthcare admins. We kept flying out, sitting with operators, codifying tacit knowledge, building systems, and refining until they stuck.

We started generating thousands in value. Then millions.

This is what Silicon Valley doesn't understand about AI transformation. Technology is the easy part. Finding the problem is harder. But the hardest part, the part almost nobody wants to do, is the human work of driving change. Sitting with people. Earning trust. Refining the product until it fits their hands. Pushing until adoption actually happens, workflows change, and meaningful business transformation is realized.

Generating $1 of EBITDA requires three things working together: (1) business acumen to find the real problem, (2) technical skill to build a solution that works, and (3) people skills to drive behavior change. These capabilities rarely exist in one person. They barely exist in most teams.

This is why transformation is so expensive. This is why Palantir charges what they charge. Why Accenture bills what they bill. AI transformation is the hardest problem in the economy: bespoke, human, resistant to scale.

We've spent the past year learning how to scale it anyway.

You can't skip the hard part, but you can get faster at everything around it. We built infrastructure that provisions production applications in 40 minutes, agents that install complex features+integrations, agents that act as on-call engineers, and systems that digest entire data rooms overnight. We built tooling that lets less-technical users do in hours what used to take experts weeks.

All of it exists so we can spend more time in kitchens and hospital floors and cramped back offices. The leverage isn't in removing the human work. It's in creating space for more of it.

There is an ambient feeling in Silicon Valley that AI is not generating real value. That the revenues are fake, the markups are inflated, the whole thing is an elaborate game of passing dollars between AI companies until the music stops.

That feeling is largely correct.

But behind the noise, there are people doing the unsexy work: flying to middle America, sitting in break rooms, watching someone mark up printouts with a highlighter, and slowly, painfully, building the bridge between what AI can do and what the economy actually needs.

$1 of EBITDA. That was the test we initially failed in Dallas, and the one we've since learned to pass. It's the standard to which we hold ourselves: tangible economic value realized by clients on the basis of solutions we provide. That's the test most of this industry cannot pass, because they are unwilling to leave the screen, and do the work.

We named the company Ciridae after the highest order in Patrick Rothfuss's novels. Those chosen for civilization-scale problems, marked by their willingness to do what others won't.

One year in, we have transformed some of the largest private equity portfolios, healthcare systems, and industrial businesses in the country. We generate EBITDA for our customers and for ourselves. Actual earnings from actual businesses that build more roofs, move more freight, save more lives, serve more customers because of what we helped them build.

The work is hard. The work is humble. The work is how AI transformation actually happens.

And we've just begun.

Note: Names have been changed for privacy.