Automation has a bad reputation, and it is earned. Too often it starts with a consultant's process chart, ends with a system nobody understands, and leaves the team in between with more work than before. The problem is not the technology. The problem is that what gets automated is what was devised, instead of what was proven.

There is a different order, and it follows the growth path. First see what happens, then follow up with an action list, and only then automate what has already repeated itself dozens of times.

Automate what the team already does

The best candidate for automation is not the biggest process, but the dullest action. The one that returns every week, is handled the same way every time and teaches nobody anything by doing it by hand.

Those candidates do not fall from the sky. They are in the action list. Whoever has followed up signals for a month can point out exactly which actions repeat, how often, and how they are solved each time. That is a stronger foundation than any workshop or process map: it is your own work, recorded.

The first workflow is therefore modest. One action the team already knows, now performed by the brain. The team recognises the logic, because it is their own logic. That is the difference with classic automation: the team does not have to learn to trust what is foreign, but to learn to let go of what is familiar.

If this, then the brain does

Where an alert says "this is here", a workflow says "this is done with it". The form is the same as with the signals: a rule in plain language, with a threshold and a consequence. Only the consequence is no longer a message, but an action.

The action repeatsVisible on the action listThe workflow takes overWrites back, creates, callsApproval where it mattersThe owner approves or declinesEvery run has a traceFrom start to finishIf this, then the brain does

An example: "When a payment arrives that matches an outstanding invoice, write it off and let the credit manager know." Or: "When stock drops below the threshold and there is a regular supplier, prepare an order." The rule is readable, the execution is traceable, and every step is in the trail.

Approval where it matters

Not everything may be done by the brain on its own, and it does not have to. The question is not "automatic or manual", but: where is the risk?

A write-off of fifty euros with an exact match can simply go through. A credit note of twenty thousand euros cannot. That is why every workflow has a threshold below which the brain acts independently and above which it first asks for approval. The approval returns in the channel where the team works, with the context included: what does the brain want to do, why, and on the basis of which rule. One tap on approve, and the trail records who allowed it.

That is how autonomy grows with trust. In the first weeks the brain asks for a lot of permission. What is repeatedly approved without adjustment can have its threshold raised. What is rejected once leads to a sharper rule. The brain learns from the team, and the team sees that.

Automation at Datahub follows the same principle as the rest of the route: you pay for results, not for promises. A workflow that completes successfully counts. A workflow that gets stuck, is rejected or delivers nothing, does not count.

A failed runStops cleanlyThe owner gets a signalIs never billedA successful runDoes the workTraceable from start to finishCounts towards the pricePaid on success

That sounds like a commercial detail, but it is a design choice. Whoever pays per attempt gets workflows that mainly try often. Whoever pays per success gets workflows that work, and a supplier who has an interest in them working. It keeps automation small and honest: better three workflows that do something good every day than thirty that demand attention.

What comes after

Whoever has a team that steers on signals, follows up actions and automates proven work, has the first two stages behind them. Then the question shifts from the one team to the organisation: do the teams do the same thing, do they mean the same thing, and can work be steered across teams?

Those are the stages connect and steer: recording shared concepts, giving key figures an owner and a target, and managing chains of work across departments as a whole. The route towards that is described in the pillar article on the growth path. The order stays the same: every step small, every step proven.

Frequently asked questions

Has the manual work been running long enough? Discover where your organisation stands on the route

Take the self-scan · Plan an introduction

Read more