Nobody Hates a Reminder. Everybody Hates a Nag.

Think about the last automated email that made you roll your eyes. Not a scam. Something legitimate, from a company you actually do business with, reminding you of a thing you already knew about, for the fourth time, in a tone that assumed you were ignoring it.

That feeling is the whole problem with automating follow-up. The reminders are trivial to build. Keeping them from turning people against you is the hard part.

We got pulled into a project recently that was, on paper, one of the simplest things you can ask for: "Can you help us send reminders automatically?" The simplicity was a trap. What follows is how we actually thought it through, the parts we were careful about, and the lesson underneath it that applies to just about any business that has to chase people for something.

The setup

The client was an organization that helps local economic-development offices run their grant programs. If that sounds far from your world, stay with me, because the shape of the problem is everywhere.

Here's how it works. A local business gets public grant money. In exchange, it makes promises: hire a certain number of people, keep them on, and meet the other terms of the agreement. Someone has to make sure those businesses actually hold up their end, collect the paperwork that proves it, and be able to show a board and the public that the money did what it was supposed to do.

One person we worked with was doing all of that out of a spreadsheet. No reminders. No automation. Every follow-up was a manual email she typed herself, every deadline a date she had to remember on her own. Multiply that across dozens of businesses and several deliverables each, and most of her week was spent being a human reminder service.

So yes, the surface request was real. Automate the nudging. But the surface request is never the actual problem.

The real problem wasn't the reminders

The actual problem was relational, and she felt it before she could name it.

These aren't faceless accounts. They're local business owners she sees around town, people she'll need goodwill from for years. The moment an automated system starts firing cold, repetitive "YOU HAVE AN OVERDUE ITEM" emails at them, her relationship with those owners takes the hit, not the software's. The tool becomes the thing people dread, and she becomes the person behind it.

Automating a chase is easy. Automating it without poisoning the relationship is the entire job.

This is the part most people skip. They think of follow-up as a logistics problem, so they reach for a logistics tool, wire up a sequence, and turn it on. Then they're surprised when response rates drop and a few people get genuinely annoyed. The logistics were never the hard part. The hard part was tone, timing, and knowing when a machine should step back and let a human step in.

What we actually did

We slowed down before building anything. A few decisions did most of the work.

We mapped the whole sequence before writing a single automation. The temptation with this kind of project is to bolt on one automation, then another, then a third when you notice a gap. Do that and the logic spider-webs fast: overlapping triggers, duplicate messages, edge cases nobody planned for. We insisted on one complete list of every message, every trigger, and every condition first, then designed the system around that single picture. Build the map, then build the roads. We've written before about how automating the wrong thing just speeds up the mess, and this is the same lesson one step earlier.

We tuned the cadence to be early, not loud. A reminder that lands two weeks before something is due reads as helpful. The same reminder on the due date, then the day after, then two days after, reads as nagging even when the words are identical. The timing carries more of the tone than the wording does. So we set the schedule to nudge well ahead of deadlines and give people room, rather than escalating the second a date passed.

We built in the positive messages too, not just the chases. This one matters more than it sounds. A system that only ever emails you when you're behind trains people to flinch every time your name shows up. So the plan included touchpoints like "thanks, you got everything in on time" and a note on the anniversary of the agreement. When some of your messages are good news, the reminders stop feeling like an accusation.

We set a threshold for when a human takes over. If someone goes quiet past a certain point, the system does not keep hammering them with louder automated emails. It flags the director to make a personal call. The machine handles the boring 90 percent, the polite early nudges that most people just need once. The human handles the 10 percent that actually requires judgment and a relationship. We take this approach a lot. It's the same thinking behind a texting tool we built that deliberately doesn't send the texts.

The honest limits

A couple of things we were careful not to pretend.

First, the bar for the records was "did we collect the thing," not "is this measured to three decimals." It would have been easy to design a system that demanded perfect, validated data on every field and then choked when real paperwork came in messy. The goal was defensible evidence that items were collected and deadlines were tracked, not false precision that nobody could actually maintain. Chasing perfect data would have recreated the exact manual burden we were trying to remove.

Second, we didn't automate the hard conversation, and we wouldn't. When a business is genuinely struggling to meet its commitments, that is not an email sequence. That's a phone call from a real person who can listen and figure out what's going on. Automating that would be automating the wrong thing. There's a whole category of work where the right move is to keep AI out of it, and relationship repair is high on the list.

And we didn't flip it on for everyone at once. It started with one person as a test case, so we could watch how real businesses responded to the real tone before anything got rolled out wider. Reminders are the kind of system that quietly drifts out of tune, which is why set it and forget it is a lie. You have to watch how people actually react and adjust.

The lesson you can take to your own business

You almost certainly have a version of this. Invoices that need chasing. Documents you need back from clients before you can start. Renewal dates. Onboarding steps a new customer has to complete. Contracts out for signature. Anywhere you're currently nudging people by hand, there's a strong pull to "just automate it."

You should. But before you do, make four decisions on purpose instead of by accident:

• Map every message and trigger first. Don't add automations one at a time. Lay out the whole sequence, then build it as one system. Piecemeal is how you end up double-emailing people and confusing yourself.
• Decide the cadence and tone deliberately. Early and calm beats frequent and loud, almost always. The timing sets the mood more than the words do.
• Include the good-news messages. If every automated email you send is a complaint, people learn to dread you. Mix in the thank-yous and the milestones so your name isn't only ever bad news.
• Decide where the human takes over. Name the point where the machine stops nudging and a real person makes contact. Automate the repetitive early steps. Protect the conversations that need a human.

Here's the thing worth sitting with. The reason to automate this was never to send more messages. It was to free up a person who was spending her week as a manual reminder service, so she could spend it on the handful of situations that actually needed her. That's what good automation does. It doesn't replace the person. It clears away the busywork so the person can do the part only a person can do.

The reminders were always the easy part. Building something people don't come to hate is the work. If you get that part right, nobody even notices the system is there. They just notice you're on top of things, and that you're easy to deal with.