The Automation Payback Window

Let me tell you about the dumbest smart thing I ever built.

A few years back, when I was still knee deep in the hedge fund world, I spent the better part of three weeks building an automation that pulled overnight market data, ran it through a model, and spat out a formatted report every morning at 6 a.m. It was gorgeous. It had error handling. It had retries. It logged everything. I was proud of it the way you are proud of a thing that took way too long to build.

Then one morning I actually did the math. The report saved me maybe eight minutes a day. I had spent something like sixty hours building it. At my billing rate back then, that automation would pay for itself in roughly two and a half years, assuming nothing broke, assuming the data source never changed its format, assuming I still wanted the report in two and a half years. Spoiler: the data source changed its format four months later and the whole thing quietly died. Nobody noticed for a week.

That is the trap. Automation feels like progress. Building it feels like leverage. And most of the time, if you actually run the numbers, you are just doing an expensive hobby with a productivity costume on.

So today I want to hand you the exact filter I use now before I automate anything. I call it the payback window, and once you start using it, you will kill half the automations on your wish list and triple down on the other half. That is a good trade.

Here is the whole idea in one sentence. Every automation has a payback window, which is the amount of time it takes for the thing to save you more than it cost you to build and run. If that window is short, build it today. If that window is long, walk away and go do something that matters.

The formula is not fancy. You do not need a finance degree. You need four numbers.

First, the build cost. This is the hours it takes to build the automation multiplied by what your time is actually worth. Not what you wish it was worth. What it is worth. If you would happily pay someone 50 dollars an hour to take a task off your plate, then your time on that task is worth 50 dollars an hour. Use that.

Second, the run cost. This is what it costs to keep the thing alive every month. Software subscriptions, API fees, the occasional fix when it breaks. Most people forget this one entirely, and it is the number that quietly bankrupts a stack of automations.

Third, the time saved. How many minutes does this actually give you back, per week, honestly. Time yourself doing the manual version once. Do not estimate. People overestimate time saved by about triple, in my experience.

Fourth, the value of that saved time. Same rate as before. The saved minutes times your hourly rate gives you weekly savings.

Now you divide. Build cost divided by weekly savings gives you the number of weeks until the thing breaks even. That is your payback window. Everything after that break even point is pure profit. Everything before it is you in the hole.

Let me run a real one so this stops being abstract.

Say you run a small service business and every new client triggers the same five step onboarding: send a welcome email, create a folder, add them to your project tool, schedule a kickoff call, and drop their details into your CRM. Right now you do it by hand. It takes you about twenty minutes per client, and you onboard roughly eight clients a month.

Twenty minutes times eight clients is 160 minutes a month, call it 2.7 hours. If your time is worth 75 dollars an hour, that manual onboarding is costing you about 200 dollars a month in your own labor. Not nothing.

Now you decide to automate it. You wire it up in a tool like Make.com, which is where I do most of this kind of plumbing because it connects to basically everything and you can see the whole flow laid out visually instead of guessing. Let us say it takes you six hours to build and test the whole onboarding sequence. At 75 dollars an hour, that is a 450 dollar build cost. The subscription and operation costs you maybe 20 dollars a month.

Weekly savings: 200 dollars a month is about 46 dollars a week, and you subtract the run cost of roughly 5 dollars a week, so you are netting about 41 dollars a week.

Payback window: 450 divided by 41 is about eleven weeks.

Eleven weeks. Under three months. After that, this thing prints roughly 41 dollars a week for you forever, or at least until your process changes. That is an automation worth building. That is not a hobby. That is leverage.

Compare that to my beautiful morning report from earlier, which had a payback window of two and a half years. Same skill, same tools, wildly different decision. The difference was not the code. The difference was the math.

Once you have run a few of these, you start to see automations fall into three buckets. I sort every candidate into one of them before I touch a keyboard.

Tier one is anything with a payback window under three months. Build these immediately. Do not overthink them. Do not gold plate them. Get them live and let them start paying you back. Client onboarding, invoice reminders, lead intake, content repurposing, anything you do more than a few times a week that follows the same steps every time.

Tier two is anything with a payback window between three months and a year. These are worth building, but only if the underlying process is stable. If you are going to change how you onboard clients next quarter, do not automate this quarter's version. You will just be automating something you are about to throw away. Build tier two only when you are confident the process will still look the same a year from now.

Tier three is anything with a payback window over a year. Walk away. I mean it. These are the automations that feel impressive at the dinner party and lose you money in real life. The only exception is if the automation removes a task you genuinely hate so much that avoiding it has value beyond the dollars. Sometimes buying your own sanity is worth a longer payback. But be honest with yourself about whether that is what is actually happening, or whether you are just rationalizing a fun project.

The math is simple. The mistakes are predictable. Here are the three that get almost everyone.

The first is the estimation lie. People are wildly optimistic about how much time an automation saves and wildly optimistic about how fast they can build it. Flip both. Assume it saves less than you think and takes longer than you think. If the payback window still looks good under those pessimistic assumptions, you have a real winner. This is the same reason smart investors stress test a deal against the downside case, not the dream case.

The second is ignoring the run cost. A single automation with a 20 dollar monthly subscription is fine. Forty of them is a 800 dollar a month bill you forgot you were paying. I audit my own automation stack every quarter and kill anything that is not clearly earning its subscription. Dead automations are like gym memberships. They quietly drain you while you feel virtuous about having them.

The third is the maintenance blind spot. Automations break. Data sources change formats, apps update their interfaces, a login expires. Every automation you own is a small ongoing liability, not a set it and forget it asset. The more you build, the more surface area you have for things to silently fail. This is why I would rather have five automations I trust than thirty I have to babysit. Simplicity is not the enemy of leverage. Complexity is.

Now, there is a real shift happening that changes some of this math, and I would be doing you a disservice not to mention it.

Building automations used to be the expensive part. You needed to know how the tools worked, how to structure logic, how to handle the weird edge cases. That build cost was the whole reason so many automations had long payback windows. The building took forever.

That cost is falling fast. I now use AI to draft the logic, write the connective code, and troubleshoot the errors on almost everything I build. I keep a stack of models open in Galaxy.ai so I can bounce a problem between a few of them without paying for four separate subscriptions, which matters when you are the kind of person who wants Claude for reasoning through a workflow and something else for a quick code snippet. When your build cost drops from six hours to two, a whole tier of automations that used to be tier three suddenly become tier one.

So if you looked at automation a year ago and decided it was not worth it, run the numbers again. The payback windows have quietly gotten a lot shorter, because the most expensive input, your build time, just got cheaper. That is the part most people have not caught up to yet. The opportunity is not that automation exists. The opportunity is that it just got dramatically cheaper to build while most people are still using last year's cost assumptions.

Here is what I want you to actually do, because reading about a system and running a system are different sports.

Pull up a list of every repetitive task in your business. Every single one that follows the same steps more than a couple times a week. For each one, run the four numbers. Build cost, run cost, time saved, value of that time. Get your payback window for each.

Sort them into the three tiers. Then build exactly one tier one automation this week. Just one. Get it live. Watch it work. Feel the difference between building for fun and building for return.

Then do it again next week. Within a couple months you will have a handful of automations, each one chosen because the math said yes, each one quietly paying you back every week. That is a wealth system. Not a pile of clever toys. A stack of small machines, each one earning.

The guy who automates everything loses. The guy who automates the right things, the ones the math actually justifies, wins slowly and then all at once. Be the second guy.

Build the system. Skip the guesswork.

Alex Rivera, Wealth Architect at The Wealth Grid