Six questions per workflow.
Before we write a single line of code.
An honest scoping document that shows which processes have AI potential — and which don’t.
How it works
- Be concrete. No abstractions — name the actual tools, people, times and amounts your organisation works with.
- Reality over SOP. Describe how it actually goes day-to-day, not how it’s written in the procedure manual.
- Question 6 is the most important. The risk on failure decides whether we can build autonomous, or whether a human must always be in the loop.
- Blank answers are valuable. What you don’t know automatically becomes a topic at the scoping session.
Your organisation
- 01
Who does this work?
Which role(s)? How many people? Full-time or part-time?
- 02
How long does it take each time?
Average time. Give a range if it varies a lot. Be honest — actual time, not SOP time.
- 03
How often does it happen?
Per day / week / month? Seasonal effects?
- 04
Which tools and systems are involved?
All software, APIs, folders, spreadsheets, mailboxes — including the "silent" tools one person maintains.
- 05
What is the error margin and how often is consultation needed?
Not just errors — also interruptions and waiting on colleagues. That’s hidden loss.
- 06
What happens if this goes wrong?
The most important answer. Determines whether we can build autonomous, or whether a human must always be in the loop — and how deep the audit trail needs to be.
↳ Follow-up for IITSFor high-risk processes we build by default: mandatory human-in-the-loop, full audit log per LLM call, and escalation protocol. Not optional — it’s the baseline under the EU AI Act.
- 01
Who does this work?
Which role(s)? How many people? Full-time or part-time?
- 02
How long does it take each time?
Average time. Give a range if it varies a lot. Be honest — actual time, not SOP time.
- 03
How often does it happen?
Per day / week / month? Seasonal effects?
- 04
Which tools and systems are involved?
All software, APIs, folders, spreadsheets, mailboxes — including the "silent" tools one person maintains.
- 05
What is the error margin and how often is consultation needed?
Not just errors — also interruptions and waiting on colleagues. That’s hidden loss.
- 06
What happens if this goes wrong?
The most important answer. Determines whether we can build autonomous, or whether a human must always be in the loop — and how deep the audit trail needs to be.
↳ Follow-up for IITSFor high-risk processes we build by default: mandatory human-in-the-loop, full audit log per LLM call, and escalation protocol. Not optional — it’s the baseline under the EU AI Act.
- 01
Who does this work?
Which role(s)? How many people? Full-time or part-time?
- 02
How long does it take each time?
Average time. Give a range if it varies a lot. Be honest — actual time, not SOP time.
- 03
How often does it happen?
Per day / week / month? Seasonal effects?
- 04
Which tools and systems are involved?
All software, APIs, folders, spreadsheets, mailboxes — including the "silent" tools one person maintains.
- 05
What is the error margin and how often is consultation needed?
Not just errors — also interruptions and waiting on colleagues. That’s hidden loss.
- 06
What happens if this goes wrong?
The most important answer. Determines whether we can build autonomous, or whether a human must always be in the loop — and how deep the audit trail needs to be.
↳ Follow-up for IITSFor high-risk processes we build by default: mandatory human-in-the-loop, full audit log per LLM call, and escalation protocol. Not optional — it’s the baseline under the EU AI Act.
Score matrix per workflow
Filled in by IITS after the scan.
| Workflow | Automatable (%) | Impact (€/year) | Compliance risk | Measurability | Score (/10) |
|---|---|---|---|---|---|
| 01 | |||||
| 02 | |||||
| 03 |
The math behind the impact
Loaded labor cost = gross salary × 1.4 (employer contributions, holiday pay, pension, sick leave). For a senior accountant at € 75,000 gross, we use € 105,000 as the true cost. We divide that by 1,520 productive hours per year.
40% of time saved. One adjacent workflow benefits. Error margin unchanged.
60% saved. Two adjacent workflows improve. Error margin drops 30%.
80% saved. Three or more workflows improve. Error margin drops 60%, lead time halves.