The busywork audit: how custom AI workflows give teams their week back

Here's an exercise we run with every client, and you can do it right now: list the tasks your team does every single week that (a) happen on a screen, (b) follow roughly the same steps each time, and (c) nobody would miss doing.

Copying leads from forms into the CRM. Assembling the Monday report from four dashboards. Renaming and filing creative assets. Summarising call notes. Chasing invoice approvals. Reformatting the same content for five platforms. Sound familiar? That list is your busywork inventory — and it's almost always bigger and more expensive than anyone expects.

Busywork is a tax on your best people

The cruel joke of busywork is that it usually lands on your most capable people, because they're the ones trusted to get it right. Your senior marketer builds the weekly report. Your best ops person reconciles the spreadsheets. Hours of judgment-capable brainpower spent on tasks that require none.

When we audit a team's week, we routinely find 8–15 hours per person of automatable work. Across a ten-person team, that's more than a full-time employee's worth of hours — currently being spent on copy-paste.

What a custom workflow actually looks like

Forget the sci-fi version. A custom AI workflow is usually a chain of small, boring, reliable steps: watch for a trigger (new form entry, new file, incoming email), read and understand the content, transform it (summarise, extract, categorise, draft), then push the result where it belongs (CRM, Slack, spreadsheet, inbox) — with a human checkpoint wherever judgment matters.

Some real examples from our projects: a lead-routing flow that reads enquiries, scores intent, drafts a personalised reply, and files everything in the CRM before the founder has finished breakfast. A reporting pipeline that pulls from ad platforms and analytics, writes the narrative summary, and posts it to Slack every Monday at 9. A content engine that turns one long-form piece into platform-native drafts for five channels, each awaiting a human yes.

The part nobody tells you: where automation is a bad idea

We turn down automation requests regularly, and it's worth explaining why. Don't automate a process that's still changing weekly — you'll automate the wrong thing, twice. Don't automate judgment calls with real consequences (pricing exceptions, sensitive customer replies) beyond a draft-for-review. And don't automate a broken process; automation makes processes faster, including bad ones. Fix first, then automate.

How we run it

Every engagement starts with the audit: we map the week, find the busywork, and rank it by hours saved versus build effort. Then we ship the top of the list in small pieces — a working automation every week or two, not a six-month platform project. Your team learns to trust each piece before the next one arrives, and by the end they're spotting automation candidates themselves. That's the real win: not the workflows we build, but the team that starts thinking in workflows.

The goal isn't replacing people. It's returning their hours to work that actually needs a human — the thinking, the taste, the relationships.

Want to know what your busywork inventory looks like? Send us a message — the audit conversation is free, and it's usually eye-opening.