Automation
Automate repeatable work that follows known rules, such as moving data between systems, building reports and processing files.
Someone downloads the same export every morning, reformats it in a spreadsheet and uploads it to another system, or spends half a day assembling a report before each board meeting. Work like this follows rules that can be written down, so software can do it the same way every time.
Common problems
The same transfer every day
Staff copy data between systems on a schedule, and a sick day or holiday means it doesn't happen.
Reports built by hand
Figures are pulled from several systems into a spreadsheet before each meeting, and the method lives in one person's head.
Mistakes from skipped steps
Errors come from a step that was missed or done twice, even though the rule itself is simple.
What we do
Automation works through a system's API, database, files or scheduled jobs. Where the only way in is through the screens, that's robotic process automation, which carries more maintenance risk and is a separate decision.
Write down the rules
Follow the process through a full cycle, including month-end and exception cases, and record the rules it actually uses.
Scheduled and triggered jobs
Build jobs that run on a schedule or when something happens, like a new record, a file arriving or a status change.
Reports from the source
Assemble reports directly from the systems of record, so they're current and repeatable.
Alerts and a manual fallback
Send failures to a named person, and keep a documented way to run the process by hand if the automation is down.
Email delivery in our CRM
Campaign email from our CRM goes through SMTP2GO. Sends are queued in batches and retried with increasing delays if the provider is unavailable. Opens, clicks, bounces and unsubscribes come back as events, are matched to the original send and are recorded once even if the provider sends them twice. An unsubscribe marks the contact so later campaigns skip them.
What you end up with
Work that happens on time
Scheduled work runs on time, including during leave and holidays.
Failures you hear about
A failed run alerts a person the same day, before a missing report or record is noticed elsewhere.
Written rules
A plain description of what the automation does, so you can check it against policy and change it later.
How the work runs
We follow the process through at least one full cycle before building, including whatever happens at month end or census date. The first version runs alongside the manual process until the results match, and only then does the manual step stop.
Questions
Should we use AI for this?
If the work follows rules you can write down, no. Ordinary automation is cheaper, gives the same result every time and is easier to check. A model is worth considering when the input is free text or documents without a fixed layout. See AI Integration.
How is this different from RPA?
This kind of automation works directly with a system through its API, database or files. Robotic process automation drives the system's screens the way a person would. We use RPA where a direct route is unavailable or impractical, because a change to the screens can break it.
Will this replace staff?
It removes the copying and checking. What your team does with the time is your decision.