What should a growing team automate first?

The temptation is to automate everything

When a team starts thinking about automation, the list of candidate workflows tends to balloon. Inbox triage, status updates, lead routing, invoice approvals, reporting, follow-ups, onboarding emails, expense categorisation, calendar coordination — the list goes on. Everything that feels repetitive looks like it should be automated.

That instinct is mostly right. Most teams under-automate. But the order matters. The wrong workflow automated first creates a brittle system that breaks every two months and a team that loses trust in the idea of automation. The right workflow automated first builds confidence and reveals the second one.

The two questions that filter what's worth automating

We ask the same two questions before recommending anything be automated, in this order.

First: is the team repeating the same judgment in the same shape, multiple times a week? Not the same task — the same judgment. Approving a leave request follows the same logic every time. Sending a follow-up to a stale lead follows the same logic every time. Routing a support ticket to the right team follows the same logic every time. If the decision shape is consistent, the workflow is automatable. If the decision is genuinely different each time, automation is the wrong tool — the team needs better information at the point of decision, not the decision removed.

Second: would this still be a problem if the team grew by 50 percent? Some friction is small at current scale and disappears at no scale. Other friction grows linearly with headcount and becomes catastrophic. Automate the second kind. Leave the first kind alone.

Repeated decisions, not one-off tasks

The single most common mistake we see is teams automating one-off admin tasks — building a small tool to handle a specific quarterly report, or a niche export, or a workflow that runs twice a year. These projects feel like quick wins because the task is bounded and the team is annoyed by it. They are almost never worth the time.

The math is simple. Automation has a fixed setup cost — design, build, test, document, maintain. That cost is amortised over how often the workflow runs. A workflow that runs twenty times a week amortises in weeks. A workflow that runs twice a year amortises in years, if at all. And the twice-a-year workflow is the one most likely to drift before the next run, because nobody maintains a tool that touches their day once a quarter.

The good candidates look the opposite. Workflows that run dozens of times a week. Workflows where the team is spending more time copying, checking, and chasing than thinking. Workflows where the same person ends up doing the same coordination dance with the same other people every Tuesday.

Keep humans in the loop where judgment matters

The best automation is quiet. It removes the boring setup around a decision while keeping the decision itself with a person. A leave-approval workflow that auto-collects context, summarises the request, and presents one screen with an approve/reject toggle is better than a workflow that auto-approves based on rules and creates an incident every time a rule misses an edge case.

The discipline is to separate the parts of a workflow that are mechanical from the parts that require judgment, and only automate the first. Status collection, data fetching, reminders, formatting, routing — mechanical. Approvals, exceptions, escalations, prioritisation calls — judgment. The first set should be invisible. The second set should be presented to a human with enough context to decide quickly.

Five places automation usually wins first

Across the teams we have helped, the workflows that almost always justify automation early look similar.

Lead and inbound routing. The right person seeing the right enquiry within the right time window. Manual routing scales badly and is one of the most common sources of dropped revenue.

Approval flows. Leave, expenses, purchases, content sign-off. The decision stays with a person; the chasing and context-gathering does not.

Status collection for weekly reporting. The hour the team spends every Friday compiling updates from five tools is one of the cleanest automation targets in any organisation.

Customer or client onboarding sequences. Welcome emails, document collection, scheduling, kickoff prep. The cadence is the same for every new account; only the names change.

Renewal and follow-up reminders. The work that gets forgotten under operational pressure and quietly leaks revenue. A boring reminder system protects more pipeline than a flashy CRM upgrade.

Where not to automate

There are two categories where automation usually makes things worse before it makes them better.

Workflows where the rules are not yet stable. If the team is still figuring out how a process should work, automating it locks in the current half-broken version. Let the process settle for a quarter before encoding it.

Workflows that are doing double duty as communication. The Monday status meeting that everyone hates but that also happens to be the only time three teams talk to each other. Removing the meeting via automation also removes the conversation, and the consequences show up later. Replace the meeting with something deliberate, not just with software.

The smallest useful automation

When teams imagine automation, they imagine a platform — a connected set of workflows running across the business. That is the long version, and it is usually the wrong place to start.

The first automation should be one workflow. Picked because it runs often. Picked because the manual version is painful. Picked because the team has done it manually long enough to know what the edge cases look like. Built end-to-end, used for a month, refined based on what actually happens in production. Once that one is calm and trusted, the second one reveals itself.

The teams that succeed with automation almost always look like this in retrospect: one workflow shipped, three months of quiet, another workflow shipped, six months later a small platform of three or four workflows that fit together. The teams that fail almost always look like the opposite — a six-workflow plan announced in month one and partially shipped by month nine, with no single workflow yet reliable enough to trust.

The pace is the same as everything else worth doing. Smaller than the plan. Shipped sooner than the plan. Maintained more than the plan accounted for.