Automations
The automation problem is a slow leak, not a crisis, which is why it goes unfixed for years. Someone — often several someones — spends hours a week moving data between systems, generating the same reports, sending the same follow-ups, reconciling the same numbers by hand. It never breaks badly enough to demand attention, so it quietly costs the business a salary’s worth of time forever. I’m hired to find that work and make it disappear, built so you own it rather than renting it back per task for the rest of the company’s life.
The problems I solve here
The first is the manual ritual that’s become invisible. A person re-keys orders into the accounting system every morning, or compiles a weekly report from four sources by hand, or copies leads between tools. It’s accepted as just how things are done. The job is to spot which of these repeated tasks are pure mechanical work a machine should own, and to take them off people entirely.
The second is the automation that was bolted together on a connector platform and has quietly become load-bearing and expensive. It started as a quick click-together fix; now core operations run through it, the per-task bill climbs every month, and when it breaks nobody can really see why. The honest answer is usually to rebuild the important flows as owned code and stop paying rent on the company’s own operations.
The third is the process that’s being automated before it’s been fixed. Someone wants to automate a workflow that’s convoluted because it grew by accident. Automating a broken process just makes it produce wrong results faster. The real work there is to simplify the process first, then automate the clean version — and sometimes the simplification alone removes most of the pain.
How I think about automations
The first rule is that you fix the process before you automate it. Automation is a multiplier, and multiplying a mess gives you a bigger mess at speed. A lot of what looks like an automation job is really a simplify-the-workflow job wearing a technical disguise, and the most valuable thing I’ll sometimes tell you is that three of your five steps shouldn’t exist. Solve the real problem — the convoluted process — not the impressive-looking one of automating every step of it.
The second is the build-don’t-rent call, which is sharp here. Connector platforms are fine for low-stakes, low-volume, experimental automations — genuinely useful for proving an idea quickly. They become a trap the moment a flow is core and high-volume, because then you’re paying per task, indefinitely, to run your own operations, with logic locked inside a tool you don’t control and can’t properly debug. My line: prototype on the easy tool if you like, but the moment an automation is load-bearing, own it as code. It becomes a fixed asset instead of a meter that runs forever.
The third is that automation is one of the clearest returns on owning your data and systems. When your systems are yours and your data flows on your terms, automating across them is cheap and open-ended — every manual task becomes a candidate to remove. When your operations are scattered across rented tools that barely talk to each other, every automation is a fight against vendor boundaries. Ownership is what makes automation compound instead of stall.
How I’d actually do it
I’d start by finding the single most expensive manual task — the one eating the most hours or carrying the most risk of human error — and automate exactly that, properly, as owned code. Not a grand “automate the whole business” programme, which tends to stall under its own weight, but one painful task removed cleanly, banking the time saving immediately. Then the next, compounding from a working base. That’s the MVP instinct applied to operations: smallest real win first, value straight away, build from there.
Each automation I’d build to be observable and robust — it logs what it did, it fails loudly rather than silently producing wrong data, and it’s simple enough that someone can understand it without its author present. A quiet automation that corrupts data unnoticed is worse than the manual task it replaced. Done right, each one is a permanent, owned reduction in the running cost of the business.
Where this connects
This is one of the specialisms I’m hired for. Automations act on data once it’s moving, which makes them the natural partner to API integrations — integration moves the data, automation does something useful with it — and they often feed the dashboards that report on the result. A focused Node or Laravel service is usually the right shape to own one. The engagement this most often powers is technical advisory when the question is what’s worth automating, or an MVP development when it’s time to build the owned version.
When you need a team instead
Automating a large operation end to end, maintained as the business changes, is ongoing team work. That’s my agency, Digital Royalty — the crew and process to build and run automation at scale. My lane is the senior call on what to automate, what to fix first, and what to own versus rent, plus the lean first build. I’ll tell you which one you need.
Who this isn’t for
If you want every step of your current process automated exactly as it is, no questions asked, I’m not your fit — I’ll want to fix the process first. I’m worth hiring when you want the manual work gone for good and owned, not just moved into a tool you’ll rent forever.
Talk to me
If your team is spending hours on work a machine should do, or a connector bill is climbing on flows you should own, tell me about your problem and I’ll tell you what’s worth automating — and what to simplify before you do.