For our first 18 months, OutcomeCatalyst was a services company. My whole career before that was in services too, where you price by the hour and the client pays for the effort the work takes.

When we launched our platform a year ago, I kept selling the same way. I would give a price on the first call before we had quantified anything for the buyer. It landed the same way almost every time: "That's expensive." When I later analyzed why deals died, pricing was the reason 27% of the time.

A client complaint finally showed me why. We’d just finished delivering their first AI workflow, and I put an expansion proposal in front of them right away. They told me it felt like I was pushing it on them, and that we hadn't spent time helping them adopt what they had already bought.

They couldn’t see what the first workflow was worth to them yet, and I was already asking for more money. It was the same mistake I was making on first calls. A buyer pays for what something is worth to them, and until they can see that, any price looks expensive.

So I changed the order. Now every time I sell a client, I build an ROI model with them before I quote a price, in their own unit economics. My prices are about the same as they were before, but pricing has gone from 27% of my lost deals to zero. Nobody has told me it's expensive since.

The buyers I talk to have rarely done this math before. Most say AI is making their team more productive, but when I ask how they measure it, they can't tell me. About 90% of them have never quantified what the manual work costs them, or what they would do with the time back.

Here’s the framework I use to build an ROI model on any AI workflow, plus three real models from commercial real estate, manufacturing, and healthcare that show what the return looks like in each.

I tell every client the same thing. Get Claude or ChatGPT subscriptions and experiment as much as you want, but don't make a real investment in AI until you understand the unit economics.

An ROI model for one AI workflow takes four steps:

Let's walk through what the math looks like in three industries below:

Below are three anonymized ROI models I built with buyers, each from numbers they gave me. The questions I ask are always some version of these:

Most buyers have never been asked. Even the numbers they give me usually start with "I think it takes about this long…"

The return shows up differently in each business. In commercial real estate, it's a better price on each deal. In manufacturing, it's more output from the same team. In healthcare, it's more patients served.

One of my prospects invests in industrial real estate. They own about 4 million square feet across the country with very few people, so every building they underwrite to acquire is their unit of value.

When I asked how many offering memos they could screen in a week, they'd never actually quantified it. So we worked it out together. A quick screen takes the team about 36 minutes each, and they can get through 10 to 20 a week. That's the ceiling, and it caps how many deals they ever get to look at.

The constraint is coverage. The team knows how to judge a deal, but they can't judge enough of them. The system we proposed codifies how they already screen and runs it across thousands of OMs instead of the hundreds they can review by hand, so a deal worth pursuing never gets lost in the unread pile.

The math is basis points on a single acquisition. Their average deal is $10 to $30 million, so on a $20 million building, a better entry of even half a percent is $100,000. Our platform costs a fraction of that. If screening every OM instead of a tenth of them finds or improves the entry on at least one deal a year, the model has paid for itself many times over.

One of our existing clients is a manufacturer, and this is the model for an expansion we're doing with them.

When a quote comes in, someone keys it into their ERP by hand to build the fabrication package. It takes about 20 minutes each time, across 10 engineers on the floor. They have done it that way for 35 years, so nobody had asked whether it needed a person at all.

When we added it up across a full year, it came to about 10,000 hours of manual work.

Our platform can extract all specs of each quote and route it straight into the ERP the moment it comes in, so nobody on the floor has to touch it. They get more requests than they have the throughput to handle today, so 10,000 hours back is 10,000 more hours to produce against that demand.

If you have a backlog, count the hours your team spends on manual work first. Time back becomes more output.

Another prospect runs a home infusion pharmacy. They're in the tens of millions in revenue and growing fast.

Hospitals are their distribution. After surgery, a patient gets a script and a referral to the pharmacy for infusions at home, which adds up to several hundred referrals a month. Each patient is the unit of value.

For every referral, a pharmacist reads through the patient's medical history, which can run 80 pages, to pull out what the case manager needs to take up with the insurance company. In some cases that back-and-forth takes two hours.

We added up that time across every pharmacist and every patient. In the model, automating the review gives each pharmacist 20 to 40 minutes a day back, which is like growing the team by 30% without hiring anyone. Then we multiplied the extra patients they could serve by the revenue per patient and found the model shows 50x enterprise value for every dollar spent with us.

If your growth is capped by skilled staff, your unit is one patient or client served.

Cutting cost and adding revenue both raise EBITDA, which is why every ROI model I build measures the return there.

The inputs are estimates. When I ask buyers how long a task takes, most of them tell me what they think it takes. Time real instances before you trust the model, so it runs on real numbers.

Extra capacity only adds revenue if there's demand to fill. The manufacturer can turn 10,000 hours into output because they have more requests than they can handle. Without a backlog or a waitlist, only the cost side of the model holds up.

You can only model the return if you understand what the workflow will do and what it costs. Most operators don't, so they have nothing to model against and never build one.

We build the model with buyers because we know the workflow’s impact and what it costs, so they can see the return before they commit to anything.

Zach

Founder & CEO, OutcomeCatalyst
Book 30 minutes with me →