The Three-Task Rule: When Repetitive Work Becomes a Bug
TL;DR: If you've done it twice, you'll do it again. The third time is your signal to automate.
A Little Background
I lead partnerships at Gadget.
Gadget, like many early start tech startups, enables you to wear many hats.
Partnerships is a very blended role that combines everything from the entire commercial GTM funnel (prospecting, outbound, sales, account management) and so much more (primary acquisition channel is events: so there's all the research, coordination, travelling to and from, post-event sales followup, etc).
Anything that takes more than 1 hour to do manually and recurs more than once per week is a great candidate for automation. Just a few of these simple tasks add up; what could you do with an extra 40 hrs per year (very low, very conservative estimate)?
So I've found that the seemingly insignificant task that make up your process, are often invisible. It's the stuff you do that prospect and customers (sometime even coworkers) never see.
But if you sit with them, intentionally design them to be part of a system, connect it all together, they often lead to higher ROI.
Everyone wants to ship the next feature. Nobody wants to fix their own broken process.
The Rule
I don't automate on the first time. Not the second.
The third time I catch myself doing the same manual task, I stop. That's the signal. Two is a coincidence. Three is a pattern—and patterns should be code.
The three-task rule sounds arbitrary until you realize how much repetitive work accumulates. That 45-minute data cleanup you "just need to get done real quick"—done 12 times this quarter. That's nine hours. Nine hours you could have spent on literally anything else.
What I Built (After Waiting Too Long)
"Automation" has existed long before AI came along.
The shift is that now, very often instead of needing a specialized expert, you can "vibe code" custom tools that solve your specific problems.
These aren't full products with beautiful UIs. There are quick but effective solutions to workflows that are currently manual, repetitive, or frustrating.
I've built a few:
- Directory scrapers: Input URL, get structured data, export as CSV, feed into Clay. What previously was a manual sales prospecting task now takes 2 minutes and costs much less than hiring someone or doing it yourself.
- Custom Sales & Partner management tools: Connect to CRM, visualize geographic location/density. This informs ad spend, physical events/travel and more.
- Compliance helpers: I've had to complete at least one long form questionnaire/portal per month for a prospective deals. Having a tool that can quickly pull from official documentation + give you statements from world-class startups as a drafted starting point is a huge time saver.
- SaaS replacements: some paid tools can even be replaced with your own custom versions. Need full control over a survey tool? Skip the $200/yr price tag and create something in an afternoon.
- Burst: A lightweight email marketing and automation tool. Used for new sign-up personalized outreach campaigns and churn automation. Not because other tools are bad is bad—because I wanted something that worked exactly how I work and didn't require some extra module for $12k/year.
The pattern: I knew I should automate. I didn't. The friction became familiar, which made it invisible.
The Trap
Internal tools scale you.
When you're early, the constraint isn't customers. It's your time and attention.
There's a few other advantages:
- Momentum - even for small tasks, or partial replacements, you get a "first draft" starting point done for you which can accelerate the rest of the process (even is manual)
- Extensible - you have full control and can always add to your tools. What often starts with a simple feature can evolve overtime into something more proprietary and reusable (in different situations).
We skip this work because it feels indirect. Building for customers shows immediate output. Building for yourself shows nothing—until suddenly you have hours back.
The three-task rule forces the decision early, before the pain becomes invisible. Two is coincidence. Three is a bug. Fix it.
Most people automate too late. By the time they build the tool, they've already paid the cost a dozen times.
Count to three. Then build.
