Assess one process in the Automation Scan
Answer seven weighted questions about process fit, impact, data and ownership. The scan stays local in your browser.
Start the Automation ScanDecision guide · Automation
Which processes should you automate first?
Choose the process with demonstrable value, stable rules, usable data and an owner who can manage exceptions and change — not simply the loudest operational frustration.
In short
A strong first candidate combines business value, process/technical feasibility and operational control. Frequency alone is not enough. If rules are unclear, source data is weak or nobody owns the process, standardise or observe it before automating.
This helps you decide
Determine whether a process is sufficiently bounded to justify serious Automation exploration.
Scope boundary
Logical next stepCompare RPA, workflow, integration and AIUse this for preselection. Process observation, architecture, privacy, security and the business case require separate assessment.
Source status
What is sourced fact and what is our analysis?
The selection model and score bands on this page are Boermans Digital decision frameworks. They are intended for preliminary assessment and are not presented as a scientific or official standard.
Process
Eight selection questions
- 01
Objective
Which concrete outcome should improve?
- 02
Frequency
Does the process recur often enough to justify change?
- 03
Rules
Are decisions explicit, explainable and relatively stable?
- 04
Exceptions
Are deviations known, and is it clear when a person takes over?
- 05
Data
Are input and output available, reliable and permitted?
- 06
Impact
Is the value visible in quality, time, risk or capacity?
- 07
Ownership
Can one owner set priorities and acceptance?
- 08
Operation
Is there capacity for monitoring, change, incidents and continuity?
Signals
Red flags
- Exceptions are unknown.
- Every employee follows a different method.
- Source data is structurally incomplete.
- Nobody can establish acceptance criteria.
- The business case counts build time but ignores operation, licences or change.
Boundary
A practical shortlist
Create a longlist first, score only with evidence and then choose one bounded process for deeper assessment. Record uncertainty and missing information with every score.
Boundary
Boundary
Boermans selection model
Score eight criteria, but never use the score without evidence
This is a practical preselection model by Boermans Digital, not a scientific standard or a business case. Score each dimension 0, 1 or 2 and record the evidence behind the score.
| Criterion | 0 points | 1 point | 2 points |
|---|---|---|---|
| Objective | No measurable outcome | Problem clear, metric not yet | Outcome and metric explicit |
| Frequency | Rare/ad hoc | Regular but low volume | Frequent and structurally recurring |
| Rules | Decisions vary by person | Main logic can be explained | Rules and acceptance criteria are stable |
| Exceptions | Unknown/unbounded | Known but poorly routed | Recognisable with clear human fallback |
| Data | Incomplete, unreliable or not permitted | Usable after remediation | Available, reliable and permitted |
| Value/impact | Not evidenced | Likely benefit | Time, quality, risk or capacity measurable |
| Ownership | Nobody decides | Shared/implicit | One accountable owner |
| Operations | No monitoring/change process | Capacity still to organise | Monitoring, incidents and change assigned |
Indicative interpretation
0–7: improve or observe the process first. 8–12: suitable for focused discovery. 13–16: strong candidate for technical feasibility and business-case work. A high score never removes privacy, security or compliance obligations.
Knockout checks
Four questions can stop a high-scoring candidate
These are not extra points. An unanswered question can mean the process is not ready to build.
- Is someone authorised to define the process and its acceptance criteria?
- Can the required data be used for this purpose technically and organisationally?
- Is it clear when the automation must stop and hand over to a person?
- Is someone accountable for monitoring, incidents, change and eventual retirement?
Examples
Processes with similar volume can still require completely different choices
These are hypothetical examples that illustrate the method; they are not client cases or performance claims.
| Process | Why it may / may not fit | First route to investigate | Main boundary |
|---|---|---|---|
| Send ERP order status to customers | High repetition, clear source data, clear outcome | API or workflow | Data quality and error-state exceptions |
| Register and route supplier invoices | High volume; rules and approvals can often be made explicit | Workflow + API; AI may extract data | Human review for exceptions |
| Compile a weekly management report | Recurring, but source definitions and ownership determine success | API/integration + workflow | One agreed KPI/source definition |
| One-off complex customer negotiation | Low volume and high contextual judgement | Do not automate end-to-end | AI may assist, not own accountability |
Frequently asked questions
Frequently asked questions about process selection
Short answers to questions that commonly arise before an automation project.
Which process should you automate first?
Prefer a process with demonstrable value, repeated volume, relatively stable rules, usable data and clear ownership. Do not automatically start with the biggest or most irritating process.
Can a high-volume process still be a bad automation candidate?
Yes. Volume helps the business case, but unknown exceptions, unreliable data or absent ownership can simply make errors spread faster.
Should you automate a broken process or improve it first?
Usually improve or standardise it first. Automation is stronger when the desired workflow, exceptions and acceptance criteria are sufficiently explicit.
Does AI make more processes automatable?
AI can help classify, extract or summarise unstructured information, but it also adds uncertainty and evaluation requirements. Human control remains important where decisions have meaningful impact.