Mobile app development for businesses

Starting with “we want to be on the App Store” is starting with the shop window. The useful question is what happens if that person can't log the work on the spot: it gets lost, it's written on paper, or someone calls the office. If the answer is one of those three, there's an app. If the answer is “they enter it later on a computer”, a website that works well on a phone is usually enough.

The task is usually small and repeated. I arrived, I saw this, I couldn't do that, here's a photo, I'm alerting the right person. Eight well-chosen fields are worth more than a clone of the ERP. The ERP is still the system of record. The app is how the field stops lying to it.

Publishing in the public stores or distributing the app only inside the company is a question of audience. A staff technician doesn't need a Google Play listing. A customer who downloads it does. We put it in the scope because it changes accounts, reviews and timelines. The developer account, if needed, is opened in the company's name.

The field doesn't have the office network. The app saves the report, sends it when it can, and says so on screen. A report marked as sent that never left is worse than a pending one: the office counts the job as done and the client waits for a visit nobody recorded.

Camera, location and notifications are requested only if the task uses them, at the moment they're needed, with a plain sentence explaining why. Asking for them at launch “just in case” is how people deny permission and the app stops being useful. Every permission without that sentence gets removed.

An alert has to be worth the interruption. “There's a new job in your area” is. “What's new in the app” isn't. The alert text, who receives it and when it stops making sense are part of the product. They're not a campaign bolted on at the end.

The app isn't finished when the report is saved in someone's pocket. It's finished when the office sees the right status: waiting to send, received, flagged, closed. That word has to match the one the desktop app or the portal already uses. If mobile invents another, the next day there are two truths.

The photo, the time and who was there travel with the report if the scope says so. The rest of the phone doesn't. The app isn't general access to every client “since the sales rep carries it anyway”. It shows what that person would see in the portal, not the margin and not another team's files.

Whoever confirms at the office has to be able to doubt. If the report arrived incomplete, it looks incomplete. Hiding the gap so the dashboard turns green is how the field stops trusting the tool and goes back to the group chat.

If the fleet is three-year-old Android phones, that's where it's tested. If management uses iPhone and the field uses Android, both go into the plan and we check which screen breaks on the small one, in direct sunlight with one hand busy. The latest model in the ad isn't the benchmark.

App stores change their rules. If the job includes publishing, ongoing support for the listing and releases is written down. If not, the company knows the binary is in its hands and the listing won't maintain itself. Distributing through the company's own device management is an option, chosen at the start.

AI, if it comes in, comes in as a named task: dictating the report, suggesting the part, classifying the photo. It isn't an open chat with every client's data in someone's pocket. What the model can read is written alongside the rest of the scope, on the same list security reviews.