30 minutes, free: book your first call

BPMN process automation Switzerland

BPMN process automation in Switzerland: approvals, case handling and task flows

You have an approval chain, a case file or a document run that today lives in email, a spreadsheet and one person’s head. Before we model anything, we read how it actually runs. What comes out of that is put in writing.

Describe your process

Which processes suit BPMN automation?

What BPMN is, and what it explicitly cannot do, are questions about the notation. This page shows what happens to your process, in what order, and what is left in your hands at the end.

Approval workflows

A request travels through three stages by email. Whose turn it is right now is recorded in no system. If someone is on holiday the case sits there, and nobody notices.

Document and case processing

A case consists of a folder, a spreadsheet and a mailbox. How far along it is can only be answered by someone going to look. What it is waiting on, nobody knows.

Operational task orchestration

Several systems have to do something in a particular order. Today a script or a person holds that order together. When it jams halfway, what has already run is unclear.

Modernising existing workflows

The flow was automated long ago, but the rules sit in the code of an application nobody enjoys touching. A change means searching, guessing, testing. The flow is written down nowhere.

Why it still runs in email anyway

A diagram takes an afternoon to draw. The first reason lies in what comes after it: between a diagram that documents the flow and one that executes it lies everything else.

The second reason is more mundane. The process everyone describes differs from the one that actually runs. One person knows the exceptions, and they are written down nowhere. A model built on the description goes wrong at exactly those points. You find out in production.

Four steps, and the first two produce no code

  1. 1. We read the process as it runs

    We read the mailbox, the spreadsheet, the exports from whatever holds it today, and the code that touches it, and we talk to the person who knows the exceptions. Without read access we do not start.

  2. 2. We write down what we found

    The process as it actually runs, including the places where two departments describe the same step differently and both are certain. It also lists the steps we would delete so they never get automated. Which version stands is your decision.

  3. 3. We model what an engine can execute

    An engine executes only part of what can be drawn. Only constructs an engine actually executes go into the model. Which technical details a model needs for a particular engine, and what a tool like the Camunda Modeler does for you there, are set out in Camunda Modeler: what the tool does.

  4. 4. You get a running process you can operate yourself

    Model, integrations and test cases sit in your repositories. Where a case stands and what it is waiting on is queryable, without anyone searching a mailbox. You should be able to run this without us.

The effort sits in one and two. Modelling comes afterwards, which is why it is third.

What is left in your hands

The process recordHow the process actually runs today, in writing, including the places where two departments describe the same step differently, and the steps we would delete so they never get automated.
The executable modelThe flow as a model an engine executes. A file in your repository, versioned like code and openable in any BPMN modeler.
The integrationsThe code the process uses to talk to your systems, so that nobody types anything in a second time.
The test casesYour real cases, including the ones that go wrong. They run on demand and without us.
The operational viewWhich case is where, since when, and what it is waiting on. Queryable, without anyone searching a mailbox.

We do not sell you a licence, and what we recommend hangs on no reseller agreement.

Often your existing supplier is the right answer

If the process happens entirely inside one system and that system ships a workflow module, its vendor is closer to it than we are. If we see that, we will tell you.

Three signs point the other way, and you can check all three yourself:

  • •The process crosses system boundaries: ERP, document store, line-of-business app, mailbox.
  • •The flow cannot be read out of the system. It sits in configuration, rules and forms, and nobody can print it.
  • •The flow is part of your supplier’s product. Then all you own is the right to use it.

If it turns out that what you are missing is an application, that belongs under custom software development.

Five things determine the effort, and you can see four of them yourself

There is no number on this page, because any number without the process would be a guess. Here is the arithmetic behind it, the same arithmetic we use when we quote.

  • •How many variants the process really has, counting only the exceptions to the main path.
  • •How many systems it talks to, and whether each of them has a usable interface.
  • •Whether the flow is decided, or still being negotiated between two departments.
  • •Whether test cases exist, or have to be created first.
  • •What happens to the cases already running today: carry them over, or let them run out.

You can estimate the first four before any conversation. We help with the fifth.

When you are in the wrong place

This fits you if a named process is holding people up, more than one system is involved, and somebody at your end is allowed to decide how the flow should work. It does not fit when:

  • •The flow is still being negotiated between two departments. An engine would freeze the argument as it stands.
  • •You cannot give us access to the systems the process talks to. A flow can be modelled from a description. Integrating it takes access, and a model that docks onto nothing is a picture.
  • •You need a binding number without anyone having spoken to you about the process first. We give everyone a binding number only after the call.
  • •It is a single form, with no cases, no deadlines, no handovers. A form tool is enough for that, and we will tell you so openly.
  • •You are looking for someone to confirm a tooling decision that has already been made.
  • •There is nobody who is allowed to decide how the flow should work. Without that person the process record is a collection of opinions.

How it starts

  1. 1. You describe the process

    A paragraph is enough: who triggers it, who signs off, which systems it touches, and how you notice when it jams. A screenshot or an existing diagram helps if you have one to hand.

  2. 2. We come back within 48 hours

    With a practical first step, and with the questions we have to ask before more can be said. We name no number at this stage, because so far we know the process only from your description.

  3. 3. A 30-minute call, free of charge

    What we need to know for an assessment we ask during that call and record there, so you are left with no homework.

  4. 4. By the end of that call your path is named

    You get one of three answers. Either: this process belongs on an engine. Or: it needs a decision inside your organisation first, because two departments describe it differently. Or: it is too small for one, and a form tool will do. Which of the three is yours, we tell you while you are still on the phone.

  5. 5. The first step is quoted afterwards, with a fixed scope

    The first step is the process record. Its scope is settled before the reading starts, which is why it can be quoted. For everything after it, a number produced before reading would be a guess, and we do not hand you a guess.

Three questions before the first call

Do we have to choose a tool first?
No. What the engine has to be able to do follows from what your systems speak and who runs the process afterwards.
If you already have one in house, the first question is whether it is enough. We talk about replacing it only once the answer is no.
We already have a diagram. Is that a basis to build on?
It is a good basis for the conversation. A diagram describes the process as everyone describes it, and the process that actually runs can differ from it. As a build specification it is therefore rarely enough.
What separates a drawn model from an executable one is the subject of our article on process modelling with BPMN. So we read the flow as it actually happens first, and compare it with the diagram.
What if you are not reachable in two years?
A fair question to put to any supplier. The answer is what sits on your side.
The flow is written down as a model a subject-matter expert can read. The integrations are code in your estate. The test cases state what the process has to be able to do, and they run without us. Whoever takes over reads that and carries on.

Projects by our founder

Each project names the client, what was built and where it stands today.

All projects

Have a process that should stop living in email?

Send us the approval, task flow or document run that slows people down. A paragraph is enough. We come back within 48 hours with a practical first step; the call after that takes 30 minutes and costs nothing.

If the honest answer is that you do not need process automation for this, you hear it in that same call.

Is this process running on Camunda 7 today? Then the question upstream of it is where that estate is going: Camunda 7 migration. What your estate decides about it is set out in Camunda 7 Migration: What Your Existing System Decides.