BLOG

What to automate first when everything is manual

The audit that finds the right first automation: frequency times pain, minus judgment.

17 November 2026 2 min read #workflow-automation

“When everything is manual, where do you even start?” is the most common automation question we hear. The instinct is to start with whatever is easiest to automate — but easiest to automate is usually easiest for a reason, and it’s rarely where the hours actually go.

The right first automation is the one with the worst ratio of how often you do it to how much it costs you. Frequency times pain. Everything else is noise.

The two-question audit

Take one week and write down every repetitive task. For each, answer:

  1. How many times per week does a human do this? (Count the times, not the minutes.)
  2. What happens when it’s done wrong or late?

Score = frequency × consequence. The top of that list is your first automation, even if it’s the hardest one to build. Especially if it’s the hardest one to build — that difficulty is why it’s still manual, and why it’s still costing you.

What NOT to automate first

Anything that requires judgment. Deciding which email deserves a reply. Choosing what to post. Interpreting why a number moved. Automation is spectacular at doing the same correct thing a thousand times and terrible at deciding what the correct thing is. The test: if a task needs a human to look at the output and decide “is this right?”, automate the steps around it, not the decision.

The sweet spot is everything adjacent to judgment: gathering the data, formatting it, routing it, scheduling it, alerting on it.

The shape of a good first automation

From the audits we run, the first project is usually one of three shapes:

  • A pipeline: data arrives from system A and needs to land in system B, transformed, on a schedule. (Prices, inventory, leads, content.)
  • An alert: a condition worth knowing about immediately — a failed job, an expired listing, a threshold crossed — currently discovered days late.
  • A formatting-and-routing task: the same document gets assembled from the same sources and sent to the same people, every time.

Each of these is a system that runs without asking, fails loudly when it breaks, and reports what it did. That last part matters — a silent automation is a future surprise, and surprises are what you’re trying to eliminate.

The worked example

Offlane is what this looks like at full scale: a gaming site whose data — live matches, patch notes, prices — flows in from 8+ external APIs on a schedule nobody maintains, refreshed by pipelines that run themselves. Nobody there remembers the last time “update the data” was a task on anyone’s list, because it stopped being one.

You don’t need eight pipelines. You need the one with the worst frequency-to-pain ratio. Tell us what it is — the automation service starts with exactly this audit.

Want this handled for you?

Start a project