Process automation means seeing how work really moves, cutting the wasted trips, and then linking the steps so each one starts the next, while people take the odd cases and keep watch.
Follow one order
In our story, Hana’s toy workshop is your business, and the toy orders are your customers’ requests. Zippy running up and down the stairs is your team carrying work from desk to desk. The glitter trail is the record your apps keep. The marble run is the linked-up process, and the signs at the forks are your business rules. The wind-up crabs are software helpers that press buttons on screens. Duchess the octopus is an AI helper that sorts odd orders onto the routes Hana built (the story calls them paths). The jack-in-the-box is the call for a person.
Stuck: a slip left on the stairs
Roll a marble down this route
An illustration of one made-up order, not a measurement.
The route, step by step
Watch · 2 min 22 sec
Zippy the mouse runs every toy order up and down the stairs, and one slip gets left on stair nine. A short story about a marble run where each step starts the next, and the odd orders still reach a person.
The idea in plain words
Each one in everyday words first, then the name experts use for it.
Stop carrying the work. Let it roll.
01
Follow one real request from start to finish. The trail shows back-and-forth loops the plan on the wall leaves out.
Experts call this: process mining (reading the trail your apps keep), process discovery (finding the real route)
Worth knowing: Trails show only what your software records; we also ask the people who do the work. Put a mess on autopilot, and it repeats the mess, so we fix the route first.
02
Once the loops are snipped, the steps are linked, people included, so work passes itself along and you can see where each request is. If a step sticks, it calls a person.
Experts call this: workflow automation, process orchestration (one system passing work between steps and people)
Worth knowing: Someone on your team still owns the run and takes what sticks.
03
House rules sit on a sign anyone can read and change, without rebuilding the run. Little software helpers press the same buttons on screens, the same way.
Experts call this: business rules, decision tables, robotic process automation (RPA)
Worth knowing: A new sign gets a quick check before the rush, and a helper needs looking after when a screen moves.
04
For orders no sign covers, an AI helper can pick a route, but only from routes you built, and it notes why. If no route fits, a person takes it.
Experts call this: exception handling, agentic automation (AI helpers that choose a next step), human in the loop (a person decides when the helper can’t)
Worth knowing: The sorter can be wrong, which is why it keeps to your routes and leaves a note.
See the difference
One order, five steps. Pick how it moves, make something stick, and drop the marble.
Where is the order?
Nobody knows without asking
Zippy carries it · A slip falls on the stairs
The slip lies on stair nine. The paint bench has nothing to start on, and the customer taps on the window.
| Make something stick | Zippy carries it | Each step starts the next |
|---|---|---|
| Nothing sticks | Zippy runs the order up, down and back again. It gets there, after a lot of stairs. | Each step starts the next, and the board shows where the order is. |
| The paint runs out | Zippy waits at the paint bench, and the next orders pile up on the stairs. Nobody upstairs knows. | The paint step tries again. Still stuck, so the jack-in-the-box pops up, a person tops up the paint, and the board shows which step is waiting. |
| Someone moves a button | Zippy has to run upstairs to ask where the paint button went, and the order waits. | The crab helpers press where the button used to be and get lost. The paint step doesn’t finish, so the jack-in-the-box pops up, and the person keeping watch shows them the new spot. |
| A slip falls on the stairs | The slip lies on stair nine. The paint bench has nothing to start on, and the customer taps on the window. | There are no stairs to drop it on: the slip rides the run with the order, and the board shows where it is. |
Worth knowing: A linked run still needs an owner who watches it and fixes what it can’t.
Worth knowing: Click-through helpers get lost when a screen changes, so someone has to look after them.
Who routes it?
Pick an order. Ordinary ones follow the signs. Odd ones go to Duchess, the story’s AI sorting helper, who picks only from Hana’s paths (the routes she built). If no path fits, a person takes it.
The sign at the fork
Duchess, the octopus in our story, stands in for an AI helper: a sorting step that picks only from routes you set, and notes why.
Who routes it
Duchess, an AI sorting helper
| Pick an order | Who routes it | Where it goes | Why | After the sign changes |
|---|---|---|---|---|
| A glass teapot | The sign | Bubble wrap | The sign says: fragile goes in bubble wrap. | Duchess, an AI sorting helper. The wooden-toy path, with no bubble wrap. The fragile row is gone and no row covers glass now, so it slips through to Duchess. She picks the wooden-toy path, and her note says: no glass row, nearest match was wooden. The note shows why, but the teapot goes unwrapped. That’s why a sign change gets a quick test. |
| A wooden car | The sign | Small box | The sign says: wooden goes in a small box. | The sign. Small box. The sign says: wooden goes in a small box. |
| A barking kite | Duchess, an AI sorting helper | The soft-toy path | Her note says: it barks like the toy dogs, so it goes their way. She can be wrong, so the note shows why. | Duchess, an AI sorting helper. The soft-toy path. Her note says: it barks like the toy dogs, so it goes their way. She can be wrong, so the note shows why. |
| A life-size unicorn | A person: Hana | Hana’s bench | No row and no path fits a life-size unicorn, so the jack-in-the-box pops up and Hana builds it herself. | A person: Hana. Hana’s bench. No row and no path fits a life-size unicorn, so the jack-in-the-box pops up and Hana builds it herself. |
The sign at the fork: Fragile → bubble wrap, Wooden → small box. After the sign changes: Metal → tin box, Wooden → small box. Before the rush, test the new sign with one marble. The test marble was a glass teapot, and it slipped past the sign. So Hana puts the fragile row back before the rush. The run didn’t change, only the sign. Fragile → bubble wrap, Metal → tin box, Wooden → small box.
Worth knowing: Rule changes still need a second pair of eyes and a quick test.
Worth knowing: A sorting step can be wrong, so it picks only from routes you set, notes why, and sends anything that fits no route to a person.
Sounds familiar?
If any of these ring true, your work probably takes more stair trips than it needs to.
Seven plain questions. Answer what you can, and we’ll point to where we’d look first.
Answer any question and the places we’d look first appear here.
Answer “No” or “Not sure” to any question to email the list.
Nothing is stored or sent unless you choose to email it.
What we help with
Eight pieces of work. Start with one request, or link them into one route.
Where work really goes, and what to fix first.
A map of the real route, its loops and waits, and which to fix first.
We follow a real request through your apps and talk to the people who handle it, using the trail your software keeps where there is one, and watching desk-level clicks only with consent.
Experts call this: process discovery (finding the real route), process mining (reading the trail your apps keep), task mining (recording on-screen clicks, with consent)
A simpler route that is worth linking up.
Cut needless steps, back-and-forth, and passing work between people for no reason, and agree who owns each step and what “done” means.
Experts call this: process redesign
Also in this area
Work that passes itself along, people included.
Work stops being carried from desk to desk, and you can see where each request is.
Link the steps across your apps and people, so work passes itself along, waits when it should, tries a stuck step again and calls a person when it stays stuck.
Experts call this: workflow orchestration, BPMN (process diagrams a computer can run), durable execution (work that picks up where it stopped after a crash)
Also in this area
Rules people can read, helpers on the boring buttons.
One rule, applied the same way, that is easier to change.
Take routing rules out of people’s heads and buried code, and write them as tables the business can read, test and change.
Experts call this: business rules, decision tables, DMN (a standard format for decision tables)
Fewer details typed twice, and helpers someone looks after.
Software helpers for the same clicks on the same screens, mainly for older systems with no other way in, plus a check-up of helpers that keep breaking.
Experts call this: robotic process automation (RPA), bot maintenance (keeping the helpers working)
Less typing by hand, with a person on the doubtful ones.
Pulling the right details out of documents into the right fields, with a person checking the ones the helper is unsure about.
Experts call this: intelligent document processing
Also in this area
AI only where it helps, and a person watching.
AI used where it helps, with people deciding the rest.
AI helpers that sort unusual requests onto routes you built, note why, keep a record, and pass anything that fits no route, or anything you choose, to a person.
Experts call this: agentic automation (AI helpers that choose a next step), human in the loop (a person decides when the helper can’t), guardrails (limits on what the helper may do), MCP (a common plug between AI helpers and tools)
You can see whether it is getting smoother, in your own numbers.
One view of where each request is and where it waits, a few plain measures agreed with you, and a few people in your own team, trained by us, who look after the helpers and pick the next loop to snip.
Experts call this: process monitoring, automation centre of excellence (a small in-house team that owns the helpers)
Also in this area
For your technical team: we can work with common tools such as Camunda, Temporal, Apache Airflow, n8n, Power Automate, UiPath, Automation Anywhere, Celonis and the workflow features inside your business software.
Ways to start
A walk-through of one real request, start to finish, and a map of the loops we find.
Fix and link one painful spot where work passes between people, with one team.
Look at software helpers that keep breaking, and decide with you whether to fix, replace or retire each one.
How we work
Five steps in plain words, from following one request to keeping watch.
See
We follow one real request through your apps, using the trail your software keeps, and mark the loops and waits we find.
Ask
Phone calls and paper leave no trail, so we sit with the people who do the work and add the steps only they know about.
Cut
We agree with you which loops, and which needless passing of work between people, to snip first, so the route we link doesn’t carry the worst loops with it.
Link
We link the steps for one team and one kind of request, with rules on readable signs, helpers on the boring buttons and a person on the odd cases. Experts call this a pilot (a first small try).
Watch
We set up a plain view of where work waits and hand over the know-how, so your team can look after the helpers, test rule changes and pick the next loop to snip.
Where it fits
Illustrative examples, not customer stories: the kind of work this approach suits.
Good to know
Plain answers to what owners and tech leads ask first. Something else on your mind?
Ask us directlySeeing how work really moves, cutting the wasted trips, and linking the steps so each one starts the next. People take the odd cases and keep watch.
No. RPA is one kind of helper: software that clicks through screens the way a person would. Process automation is the whole route: finding it, fixing it, linking it, and picking the right helper for each step, including a person.
That isn’t the aim. It takes the boring, repeated button-pressing; people keep the odd cases, the judgement calls and the job of keeping watch.
Not to start. A first fix can be removing a step rather than adding software, and we begin with the tools you already have. Start with the simplest thing that works.
Only on the odd cases, where fixed steps fall short. They choose only from routes you set, note why they chose, and send anything that fits no route to a person. They can still be wrong, which is why the note and the person are there.
The run is built to notice: it can try the step again, and if it is still stuck, it calls a person and shows where the work is waiting. Helpers that click through screens also need someone looking after them.
We agree a few plain measures with you at the start, such as how long a request waits and how often it goes back for fixes. The numbers are yours, so you can judge it for yourself.
It depends on the work. We start by following one request, then agree the next step together.
Heard it before?
Answer: Linked up as it is, a messy route just repeats its mess. Follow the route first, snip the loops, then link the steps.
Answer: The trail can show loops nobody planned. It shows only what your apps record, so we also ask the people who do the work.
Answer: It presses only what it was shown. Move a button, and it gets lost, so someone has to look after it.
Answer: Start with the simplest thing that works. Fixed steps suit work that follows the same rules. An AI helper takes only the odd cases, on routes you set, notes why, and passes anything that fits no route to a person.
Answer: The helpers here live inside your software. Nothing walks around, apart from the wind-up crabs in our story.
Answer: Small teams have loops too. Start with one request.
Take the overview or the playbook to share with your team, or tell us about one request your team handles and where it seems to loop. We’ll set up a short call to pick one request to follow, and suggest where to start.
Or write to contact@algoshred.com
