Skip to main content

Process analysis before automation. Cut first, then build.

Most automation projects don't fail on the technology but because a cumbersome workflow just runs cumbersome faster. That is why it goes in four steps: report, measure, cut, automate. Software only comes in the last one.

13 questions · about 5 minutes · short result without email

Fix workflows, don't reinvent them. The starting point is what already exists.

  • Cut steps, rebuild logic, close media breaks, connect tools: on the workflow that already exists.

  • Not included: rethinking your business model on a greenfield.

A small team stands at a wall with a printed process map and sticky notes; a woman puts a red dot on one step.

Staged scene

Tuesday, 8:15 am

Before anyone talks about software, the workflow is up on the wall.

Every red dot is a place where the team waits, chases or types twice. That is where the method starts.

Four steps. In this order.

Start with the tool and you automate the detours along with it. That is why software only comes in step four, after cutting.

1/4Report

The people who work with it every day know where it gets stuck.

So the analysis starts with the team, not with the tool. Every disruption is noted at the moment it happens: what, where, how many minutes.

  • Note it in the moment, not from memory
  • Make people and handovers visible
  • Capture the workflow as it really runs
  1. Request
  2. Review2 reports
  3. Re-enter data5 reports
  4. Chase up6 reports
  5. Approval2 reports
  6. Transfer4 reports
  7. Done1 report
Time in the workflow20 reports

Example, made up: one workflow, one week.

2/4Measure

Opinions turn into a number.

After a few weeks it is clear which causes come up how often and what they cost in hours and euros. From then on the discussion is about numbers, not gut feeling.

  • Sort reports by cause
  • Project times honestly
  • Name waiting times, loops and error sources
  1. Request
  2. Review30 min
  3. Re-enter data95 min
  4. Chase up120 min
  5. Approval25 min
  6. Transfer70 min
  7. Done15 min
Time in the workflow5.9 hper week

Example, made up: one workflow, one week.

3/4Cut

What nobody needs, nobody has to automate.

Before a single line of software is written, whatever is superfluous goes. A removed step, a required field or a settled responsibility often solves more than any tool, and costs no licence.

The step most people skip.

  • Cut unnecessary steps
  • Parallel instead of sequential
  • Unify rules and formats
  1. Request
  2. Review30 min
  3. Re-enter datacut
  4. Chase upcut
  5. Approval25 min
  6. Transfercut
  7. Done15 min
Time in the workflow1.2 hper week

Example, made up: one workflow, one week.

4/4Automate

And the rest runs by itself.

Only now does software come in: for what is left after cutting and follows fixed rules. Where judgment is needed, a person still decides. Afterwards it gets re-measured at the same point.

  • Rule-based steps and data transfers
  • First what brings a lot for little effort
  • Re-measure after a few months
  1. Requestruns by itself
  2. Reviewruns by itself
  3. Re-enter datacut
  4. Chase upcut
  5. Approvala person decides
  6. Transfercut
  7. Doneruns by itself
stays with people0.4 hper week

Example, made up: one workflow, one week.

Three steps go before any tool enters the picture. What gets automated is the leaner workflow, not the old detour.

Six marks of a good candidate

Is a workflow even suitable? These six questions give the answer before you spend money on tools.

Frequency
How often does the workflow run?
Daily beats monthly
Volume
How many cases per run?
More cases, more leverage
Time effort
How long does one run take by hand?
Minutes and hours add up
Error-proneness
Where do errors creep in?
Manual entries and transfers
Proportion
Does what goes away justify the effort?
Not everything possible is worth it
Fixed rules
Does the workflow run by the same rules every time?
Without rules there is nothing software could repeat

The self-check asks exactly this, for one workflow, in about five minutes. You see the short result right away, without an email address.

Start the self-check

What gets automated first

After cutting there is usually more than one candidate left. Impact against effort decides what comes first.

  • Start here

    High impact, low effort

    Begin here · Pays off quickly · Builds momentum

  • Larger project

    High impact, high effort

    Plan carefully · Longer haul · Highest value

  • Can wait

    Low impact, low effort

    Onto the list · Nice, but not urgent · If time allows

  • Better leave it

    Low impact, high effort

    Not worth it · Ties up people and money · Rather cut or simplify

And which kind of workflow first?

Not every workflow deserves attention first. This split shows where automation pays off most.

  1. 1/3

    Core business

    Directly value-creating: this is where the company makes its money.

    Top priority: biggest lever, highest visibility

    Typical examples: Sales & distribution, Production & service delivery, Customer service & support, Product development

  2. 2/3

    Supporting workflows

    Keep the place running without generating revenue themselves.

    Often the best entry point: lots of routine, quickly noticeable

    Typical examples: HR & recruiting, Accounting & finance, IT support, Facility management

  3. 3/3

    Steering workflows

    Strategically and legally important, but rarely pure routine.

    Automate selectively, with judgment

    Typical examples: Strategy & planning, Quality management, Compliance & governance, Change management

Maturity: is the workflow even there yet?

Automation needs a stable workflow. From level 3 it pays off; below that, tidying up brings more.

  1. 1

    Different every time

    No fixed rules, everyone does it their own way.

    Not there yet

  2. 2

    Written down

    The workflow is described but not always followed.

    Prepare first

  3. automate from here3

    Consistent

    Runs by the same rules every time.

    Ready to automate

  4. 4

    Measured

    Time and errors are counted.

    Very well suited

  5. 5

    Continuously improved

    Re-measured and adjusted regularly.

    Ready for extensive automation

Rule of thumb: automate from level 3. Below that, rebuilding pays off more.

Two colleagues try out a new workflow on one laptop; she ticks off a printed checklist.

Staged scene

Thursday, 4:00 pm

One workflow, one department, two weeks. Then the rest.

Try it small, on real cases, with the people who run the workflow afterwards. What does not hold there is not rolled out.

What works and what goes wrong

What makes automation projects succeed and what makes them fail in droves.

This works

  • Rebuild the workflow first, then automate
  • Start with simple, frequent workflows
  • Involve the people who run the workflow daily
  • Write down how it really runs today
  • Define how you'll measure success up front
  • Try it small before you roll out

This goes wrong

  • Automate a cumbersome workflow one to one
  • Jump straight into the tool without analysis
  • Try to tackle everything at once
  • Ignore scaling and maintenance
  • Present the team with a done deal
  • Consider privacy and compliance too late

Which workflow annoys you the most?

Place it in the self-check: 13 questions, about five minutes, the short result instantly and without an email address. If you would rather talk about it straight away, that works too.