Automating Admin Tasks Without Surrendering Control

I show you how I replaced a patchwork of SaaS tools with a custom workflow that automates the admin grind, yet leaves every decision in my hands.

When I first walked into a 200‑employee manufacturing outfit, I found a wall of spreadsheets, email approvals, and a dozen SaaS subscriptions that never spoke to each other. The CFO complained about recurring fees, the ops team cursed the endless data entry, and I could see the same admin loops repeating every week. My instinct was to throw more software at the problem, but the reality was that each tool was a half‑baked template that the business had to bend, not the other way around. The first rule I set for myself was simple: any automation must sit inside a system we own, not inside a vendor’s UI that we can’t see.

Map the real work, not the software

I started by shadowing the people who lived the admin grind: the office manager filing purchase orders, the HR assistant reconciling time‑cards, the accountant pulling month‑end reports. I recorded each step, the decision points, and the data sources they touched. The map revealed three patterns: most steps were rule‑based, the exceptions were few, and the hand‑offs were where errors slipped in. I stopped asking "What does the software do?" and asked "What does the work actually require?" That shift let me strip away the noise of SaaS feature lists and focus on the core logic that drives the business.

With a clean process diagram in hand, I could see where a single script could replace a manual copy‑paste, and where a tiny database could replace a shared folder. The map also exposed hidden dependencies – for example, the procurement team relied on a legacy ERP export that never updated in real time. Those dependencies became the anchors for the new system: if I could feed them directly into a custom API, the rest of the workflow would flow automatically.

Replace the rented SaaS with a lean owned stack

The next step was to retire the SaaS cobweb. In the rebuild I led, we eliminated 21 separate subscriptions and built a single internal portal that handled purchase requests, approvals, and reporting. The portal runs on a modest container cluster on AWS, and the code lives in a private Git repo that the team can audit any day. Because we own the stack, we can add a field, change a validation rule, or expose a new endpoint without waiting for a vendor roadmap.

Ownership also means cost control. The $0 ongoing SaaS cost we achieved didn’t come from a miracle discount; it came from eliminating recurring licenses and moving the compute bill to a predictable, usage‑based model. The team now sees the exact price of each transaction, which forces us to keep the automation lean.