ROI from AI
The most difficult AI concept for many entrepreneurs to comprehend is how it applies to them. If you don’t understand how software works, it’s magic. This is amplified if you don’t understand how AI works. This is part of why I feel we are a great trusted partner for our clients; we allow you to focus on what you’re uniquely good at, and we will be your AI team.
The second most difficult concept is measuring success. We coach four ways for measuring success in our Coaching program. Each project is different, and each customer’s goals are different, so we let them choose the method that works best for them.
First, a few definitions:
Task – the series of actions in question. It can be as simple as adding a Stripe customer to a CRM or as complex as doing low-level report analysis and preparing an entire report.
Unit Produced – 1 task being completed
N – the number of people doing that task. This can be 1 or it can be 1000.
All examples assume AI completely automates the task.
Software is high fixed cost, zero marginal cost
Software is valuable because it is scalable. It requires upfront investment to build it, but the cost of duplication is zero.
Think about when you duplicate a file on your computer or you send a photo on your phone. In the atoms-and-elements world, you would need to print another copy of that document or print, then mail, that photo if you wanted to share it, all of which cost money. Buying a phone and a computer is expensive up-front, but it’s free to duplicate a file on your computer or to send a photo to a friend.
In the bits-and-bytes world, replication is free.
The same goes with AI. Once an AI can do a task once, it can do it millions of times for the same (or similar, depending on circumstances) cost. There’s a high fixed cost (the cost of building the technology), but there’s zero marginal cost.
Measurement 1: Time Saved with AI (in labor hours)
(Labor Hours per Unit Produced) x (N) x (Number of Units per year)
+
(Time it takes to train that task) x (N + 1((The Trainee)))
= Labor Hours Saved Per Year.Let’s break this down.
Let’s assume a company inspects buildings for damages. The report team looks at photos, crunches numbers to determine the amount and value of damages, and prepares a report. Each report takes their one employee 2 full workdays. Let’s also assume a company does 100 reports per year.
16 x 1 x 100
+
(Time it takes to train that task) x (1 + 1((The Trainee)))
= Labor Hours Saved Per Year.Already, AI has saved your company 1600 labor-hours per year, not including training time.
Let’s assume it takes 2 weeks to train that employee. That’s 2 weeks for the new employee (trainee) and 2 weeks for the trainer. That’s a total of 80 labor-hours per person.((For simplicity’s sake, it’s easiest to write the formula assuming you retrain the role each year, to account for turnover. I understand that’s not what happens, but you can use the top line in the formula for each subsequent year where training is not needed.))
16 x 1 x 100
+
80 x (1 + 1((The Trainee)))
= Labor Hours Saved Per Year.In this example, in the first year, AI saved the company 1,760 labor-hours.
Let’s put this into perspective.
A one-time investment just saved the company an entire year of payroll. Forever.
This works especially well in high-turnover roles where training is constant and the company spends far too much time and energy hiring.
The unit economics get especially good if the company has two people doing the role.
16 x 2 x 100
+
80 x (2 + 1((The Trainee)))
= Labor Hours Saved Per Year.With two people doing the task and no increase in reports per year, the company saves over 3,440 hours per year. Each year.
Now, let’s assume the company doubles its number of reports, assumedly doubling the revenue.
16 x 2 x 200
+
80 x (2 + 1((The Trainee)))
= Labor Hours Saved Per Year.The company saves 6,640 labor-hours per year.
The beauty of AI saving time is that the AI cost the same no matter if the task is being done once or one million times. Due to the zero-marginal-cost nature of software, scaling is easy and predictable, it requires little-to-no training, and increases gross margin, all at the same time.
Measurement 2: Money Saved (from labor)
Of course, this is not a direct change to the bottom line. This is money saved because your current team is creating new value. The payroll stays the same, but the company’s output drastically increases. Oftentimes, the way we think about this is in the reverse.
“How much money would you have needed to pay to accomplish the things you are now?”
(Labor Hours per Unit Produced) x (N) x (Number of Units per year)
+
(Time it takes to train that task) x (N + 1((The Trainee)))
= Labor Hours Saved Per Year(Labor Hours Saved Per Year) x (Average Labor Hourly Rate)
= Money Saved from LaborLet’s use the first example from above.
16 x 1 x 100
+
80 x (1 + 1((The Trainee)))
= Labor Hours Saved Per Year.The company has saved 1760 labor-hours per year with one AI. Assume that individual is paid $75,000 per year. Their effective hourly rate is $37.50.
(1760 Labor Hours Per Year) x ($37.50) = $66,000That AI, each year, saves the company $66,000. Within 24 months, your ROI is already above 1.5x. Within 36 months, your ROI is above 2.5X. Within five years, your ROI is 4.4
Measurement 3: Money Made (new cashflow)
There are two primary ways that AI can increase cashflow.
1. Increasing the infrastructure of your existing business to handle more clients with the same team.
We’ve briefly experimented with the first point, increasing the number of reports. When we did that, however, we increased the payroll. Let’s assume the same company, but instead of 100 reports per year, they have a great year and have 50 new reports.
(Revenue per Unit) x (Number of New Units) = Money MadeThere’s nothing else in this formula because the Cost of Goods Sold is $0.00! The zero-marginal-cost nature of software made this new revenue all profit.
When we piloted Lede AI, we wrote the code once. On our first football Friday night, we published twice as many articles as our partner had in the entire previous season. In this case, we handled way more article output with the same team.
Now, four years later, Lede AI has published over 1000X more articles than our collaborator would have without Lede AI.
2. Increase the average customer spend by adding new value to your offer.
This happens All. The. Time. Below are a few examples.
A company is already a software company and they increase the prices across the board but add the new features in.
A company adds an “express processing” option because AI works so fast. Customers take this option a lot.
A company adds additional analytics, cloud hosting, or consulting to accompany the new automations and AI.
A company pivots existing, analog offerings into a SaaS solution, and sells that on subscription.
A company uses the AI they built to enter a new market as a SaaS product, and sells that on subscription.
A company leverages the AI they built to turn their competition into their clientele.
Lede AI started as an internal project to cover more high school sporting events. We spun it into a SaaS offering to become a news provider for news providers (#5). We offer one customization that’s mostly automated, and one that’s completely automated with additional features (#1 and #2). In the process, we sold a license to our software to a competitor (#6).
Measurement 4: Hassle Saved
Sometimes, it’s worth it to simply never deal with that task again. It can be more headache than it’s worth, and freedom from that task leads to freedom to do other, more productive things.
Truthfully, oftentimes ROI from AI comes from a combination of all options. There’s time saved and money saved and money made and hassle saved. I think the beauty of software, and AI in particular, is that it has the ability to make 1+1 = 10.