Skip to main content
Get a Proposal

Notes on SEO, data collection, and automation

  • Published Oct 16, 2026
  • 7 min read

Repetitive work: build RPA, or another back office?

Daily exports, data entry, reconciliation, and fixed notices are often faster as a script or RPA than as a new back office — when the steps stay the same. When the rules change every week, automation costs more. This note is the test, and what we would hand over.

First ask: are the steps the same every week?

This is for people deciding between another back office and automating the repetitive steps. The test is one question: are the steps the same every week? If yes, a script or RPA is worth it. If the rules change weekly, tidy them first or build a back office with permissions. VAV Agent does both, and the proposal names the steps that should stay manual.

Quote transcription, report exports, and fixed-format notifications often meet the bar for automation: short paths, stable fields, explainable exceptions. Building a new back office in that case duplicates a system that already exists.

If the rules change weekly and exceptions outnumber the happy path, automation welds an unstable process in place. Tidy the rules first, or build a back office with permissions and an audit trail.

Prefer scripts when there's an API or stable files

Scripts read files, call APIs, and write logs; colleagues can take over via the repository. RPA fits when the target system has no API and UI automation is explicitly allowed. The choice belongs in the proposal — not in the tool vendor's hands.

Either way, use a dedicated client-provided account with minimal permissions. Never borrow an employee's personal account, and never leave credentials in personal chats.

Automation replaces operations, not accountability

Price exceptions, external promises, and irreversible customer notifications belong at human confirmation points. A flow without failure alerts and fallback steps shouldn't be wired directly to actions the customer can see.

A proper handover has four things: the flow and its exceptions, a schedulable task, logs and alerts, and the steps to finish an order by hand. Missing any one, it isn't an automation delivery yet.

A test you can apply on the spot

If you can answer 'when it fails, who gets notified, and what three steps finish the order manually', it's worth scripting or automating. If you can't, don't automate yet. We write the not-recommended steps into the proposal, so the scope doesn't become a bot that's forever being repaired.

How we can help with this

Send the steps in a few sentences: which sheet or system it starts in, where the data is pasted, and when a person has to look. We first say whether it should be a script, RPA, or a back office, then deliver a scheduled job, logs, failure alerts, and the steps to finish one order by hand.

Steps whose rules change every week, or whose exceptions outnumber the normal path, go in 'not this time'. We will not weld an unstable process in place just to ship a bot.