Approval workflows
Multi-step approvals with routing rules, escalation on timeout and an audit trail of who approved what and when.
Most internal processes were never designed. They accumulated. Somebody copies a value from one system into another, somebody chases an approval by email, somebody rebuilds the same report every Monday. We map what actually happens, then rebuild the parts that do not need judgement.
Fig. 01A rebuilt processIllustrative
Typical stack
Automating a broken process makes it fail faster. The first phase is always observation — what triggers this, who touches it, where does it stall, and what happens on the exception path that nobody has written down.
Process discovery. We follow a real instance end to end, including the informal steps that live in somebody's inbox.
Exception mapping. The happy path is the easy part. The value is in handling the cases that currently get escalated.
Human-in-the-loop by default. Anything irreversible — a payment, a deletion, an external message — keeps an approval gate.
Measurable before and after. Cycle time and touch count recorded before we change anything, so the result is checkable.
Multi-step approvals with routing rules, escalation on timeout and an audit trail of who approved what and when.
Values moved between systems by integration rather than by a person retyping them, with validation at the boundary.
Recurring reports assembled, formatted and delivered on a schedule, sourced from the systems of record rather than a spreadsheet.
Account creation, access grants, equipment requests and revocation driven from a single HR event.
Anything the workflow cannot resolve lands in a queue with the context attached, instead of failing silently.
Incoming documents classified, extracted, validated and filed, with only low-confidence cases reaching a person.
Five stages from observation to a workflow that runs unattended. The order matters — measurement comes before automation, not after it.
Every engagement ends with artefacts you own outright and can hand to another team without calling us first.
A process map. The current and target flow documented, including the exception paths that were previously undocumented.
The running workflows. Version-controlled definitions on infrastructure you own, not locked inside a vendor account we control.
An operations runbook. What each workflow does, how to pause it, and what to check when it stops.
Baseline measurements. Cycle time and manual touch count before and after, so the change is arguable with numbers.
In our experience it removes the parts of a role nobody wanted — retyping, chasing, reconciling — rather than the role. We would rather say that plainly than sell it as headcount reduction and be wrong.
Workflow definitions are version-controlled and editable by your team. We hand over the definitions and the documentation; you are not required to come back to us for a routing change.
Eventually, yes — read access during discovery and scoped write access at launch. We work through service accounts with least-privilege scopes rather than a shared human login.
If a process runs a handful of times a month and takes minutes, the automation will not pay for itself. We will say so rather than build it.
Describe how it works today. We will tell you whether it is worth automating and what the honest effort looks like.