What to automate first, what it actually costs, and where a person still belongs
When people hear “automation,” they usually imagine something dramatic: an advanced robot, an AI that thinks for you, something that quietly takes over someone’s job.
It’s actually a lot simpler than that. It’s just a system that repeatedly does something we’ve already decided on. You watch someone do the same work every week, notice which parts of it never actually change, and hand those off so they run on their own. What’s left is the work that actually needs a person.
That’s my job at Solve Education! Over the past year I’ve built systems for grant applications, recruitment, payroll, marketing stories, and partner communications. A few worked right away. A few cost me more time than they saved before they paid off. Here’s what I’ve learned, in case you’re thinking about the same thing for your own team.
Start with the work that repeats
Almost anything can be automated eventually. But the cost isn’t the same everywhere, and the first one you build should be the one that pays you back fastest.
The cheapest wins are the repetitive ones. Clear inputs, rules someone could write down, a predictable result, happening every week or every month the same way.
Our payroll was exactly that. Every month someone opened the payroll sheet, worked out each person’s totals, filled in a template, and emailed the payslips one by one. None of it needed a decision. It just needed care, and hours of it.
When we automated it, the decision that mattered wasn’t technical. It was that every number gets pulled from the payroll sheet exactly as it appears, with no recalculation anywhere. One source of truth. If something’s wrong, there’s exactly one place to fix it.
The painful, complicated work is worth automating too. We’ve done plenty of it. Just go in knowing it’s a bigger build, and don’t make it your first one.
It doesn’t have to be all or nothing
This is the part I think most people get wrong, and our grant system is the clearest example of it.
Finding funding used to mean someone searching Google, scrolling social media, and working through grant aggregators. The problem was never a shortage of results. It was the sheer number of them, and the hours spent sorting the handful that fit us from the hundreds that didn’t.
So the system does that part. It searches continuously, checks each opportunity against our eligibility rules, and pulls the application questions for anything that passes. Anything that doesn’t fit gets recorded as a no, which means nobody ever evaluates that grant twice. That list of rejections turned out to be as valuable as the list of matches.
Then it drafts. It writes a first version using our past successful applications as reference, along with the guidelines we’ve written for it. And then it stops, and a person takes over.
The draft is a starting point, not a finished one. Our grant writer improves and polishes it, because judging whether a draft is actually good still needs a human. What she learns goes back into the knowledge base, so the next draft starts a little closer.
That’s the part worth copying. It handles the parts that don’t need her, so she can spend her time on the parts that do: judgement, relationships, and the parts of a proposal that decide whether we win it. That’s as much as 60 percent of the time that work used to take, given back to her.
It gets cheaper the more you use it
This is the part that trips most people up, and it’s honestly worth planning for from day one.
Building the system almost always takes longer than doing the task by hand would have. You don’t get paid back on the first run. You get paid back the more times you run it. So the question isn’t “will this save time this month.” It’s “how many more times are we going to do this.”
The same logic applies once you’re building more than one automation. Two systems that share nothing cost you full price twice, because you’re solving the same underlying problems separately instead of once. If you already know more and more automations are coming, it’s worth spending time early on to plan something they can all sit on, even though that makes the initial work slower.
Why this matters for our work
Every hour a machine can take on is an hour your team gets back for the work only they can do. For us, that’s easy to trace. Our funders give us money, but what that money actually buys is people’s time. Give our grant writer back a few hours a week, and she wins more funding. Win more funding, and we run more programs. Run more programs, and more people gain the skills that lead to real work.
That chain looks different at a company, but the shape is the same. Give an engineer back a few hours, and a feature ships sooner. Give a salesperson back a few hours, and they close another deal. Either way, the time saved only matters because of where it goes next.
Automation is just the cheapest way we’ve found to make sure it goes somewhere that matters.
“Automation doesn’t replace people. It gives them their time back.”
Thinking about this for your own organisation?
If your team is losing hours every month to work that never changes, we’d be glad to help you figure out what’s worth automating, what isn’t, and where to start. You can reach us at solveeducation.org.