Tools and automation

Almost everything can be automated. The question is what pays off first.

Two questions tell you where to start: how often does the task come round, and what does an error represent? They identify in minutes the automations that pay off straight away, and the order in which to build them.

Questions to know where to start: 2
Lifespan of a flow: 3–5 yr
Named person per alert: 1
Expert consultants: 11

2
Questions settle where to start: 3–5 yr
How long a flow actually lives: 1
Named person receives every alert: +500
Connectors available

Our method

Straighten the process, then automate it

A request to automate rarely arrives alone. It arrives because a task is painful, and a task is often painful because the process producing it is awkward: three approvals of which two are formalities, a retyping step that exists only because two departments refuse to use the same tool, a control added after a 2019 incident nobody can explain.

A process rethought before being automated goes considerably further for the same budget: you automate the short path, not the historical one. That is where half the gain sits, and it costs nothing.

So we always start by looking at the process rather than the tool. Sometimes a step disappears entirely, and the request for automation disappears with it. That is a good outcome, even though it does not bill.

Straighten the process, then automate it. Never the other way round.
The rule we apply

The trade-off

Where to start so it pays off soonest

Two questions are enough to decide

How often does the task come back, and what does a mistake cost? Two of the four answers consist in leaving the task manual: they are the ones you hear least often.

Should a task be automated? A call based on how often it recurs and what a mistake costs.
	Rare task	Frequent task
Harmless mistake	Keep it manual Rare, harmless A task done three times a year, where a mistake is fixed in five minutes. A flow costs more to write, document and maintain than it will ever save. We say so, even when the request comes from you.	Start here Frequent, harmless The best ratio between gain and risk. Reminders, notifications, copying between two tools, reports sent every Monday. The gain adds up every week and a failure only costs one cycle.
Costly mistake	Automate the check Rare, costly A period close, a filing, a cutover. Automating the act itself carries a disproportionate risk for a marginal gain. What pays off here is control: automate the checklist, not the decision.	Automate, and monitor Frequent, costly Invoicing, reference data sync, feeding an ERP. The gain is real, and so is the risk of a failure. A flow of this kind ships with its monitoring: an alert on failure, a log you can read, and someone named to receive it.

This call is made before the quote, not after. We sometimes end a workshop by saying there is nothing to automate: it is quicker for everyone, and it is what makes the next yes credible.

Lifespan

What makes a flow last

The quote for an automation prices its construction. Its real life starts afterwards, and nobody prices that.

Interfaces change

A connector evolves, an authentication is hardened, a field is renamed. The flow stops or, worse, keeps running and produces wrong results. Over three to five years, this happens.

People leave

The flow alerts an address that no longer exists, or runs under the account of an employee who has been disabled. We use service accounts and distribution lists, never personal ones.

The organisation moves

A branch opens, a company is acquired, an approval route changes. A flow that hard-codes a list of departments quietly becomes wrong. Parameters are read from a table, not from the code.

Nobody does it by hand any more

This is the real danger. Six months after automation, the manual skill is gone. A flow that fails silently is worse than no flow at all, hence supervision, systematic on any sensitive process.

The sequence

How we proceed

We look at the process, not the request
An hour with the people who do the task. Often two steps fall away before a tool is even mentioned.
We place the task on the matrix
Frequency and cost of an error. The verdict comes there, before the quote, and sometimes it is to do nothing.
We choose the tool, not the reverse
Power Automate to link Microsoft services, development when logic or volume justify it. We build both, with no interest in pushing one over the other.
We deliver supervision with the flow
Alert, log, named recipient, written manual fallback. The difference between an asset and a liability.
We train someone on your side
At least one person must be able to read and change the flow. Flows live in your environment, legibly named and documented.

Going further

The pages that continue this one

Frequently asked

What people ask before automating

When does an automation pay for itself?

Count the build time, then add maintenance: a flow lives three to five years and will survive at least one API change, one reorganisation and one departure. A weekly task taking fifteen minutes justifies the investment easily. A quarterly task taking thirty minutes almost never does. We do that arithmetic before quoting, not after.

How do you make sure a flow keeps running?

It is settled at design time. Every flow we deliver on a sensitive process ships with its supervision: an alert on failure, a readable log, and above all a named person who receives the alert. A flow that fails silently is worse than no flow at all, because nobody does the task by hand any more.

Should we rework the process before automating it?

Yes, and that is where most of the gain sits. A process automated as-is keeps its detours; the same process rethought first goes twice as far for the same budget. We start with an hour alongside the people who do the task, and a step often disappears on its own, which makes the project simpler still.

Power Automate or custom development?

Power Automate for anything linking Microsoft services: approvals, notifications, file drops, list updates. Development when the logic gets complex, when volume is high, or when cost per run starts to matter. We do both and we have no interest in pushing one over the other.

Who takes over if you are no longer involved?

Your teams, provided it was planned, and it is. Flows are legibly named, documented, and live in your environment rather than ours. We train at least one person to read and modify them. A flow that only the provider understands is a liability, not an asset.

What we deliver

A flow, and its supervision

A flow has four steps, trigger, checks, action and trace, plus a supervision layer that warns a named person as soon as one of them stops.

A flow is delivered with its supervision

The four steps of the flow, and the layer that raises the alarm when one of them stops.

A flow that fails silently is worse than no flow at all: six months after the automation, nobody knows how to do the task by hand any more. The alert, the log and the named recipient ship with it, rather than being added after the first incident.

Tell us which task weighs on you.

Eleven expert consultantsSaint-Priest, near Lyon, France

Two hours with a consultant, free of charge. We look at the process as it actually runs, place the task on the matrix, and tell you what we would automate, including when the answer is nothing.

Request a scoping workshop

Scoping workshop · 2 hours · free · near Lyon or remote