# Route Nord — Discovery Worksheet

Use this worksheet to identify one operational problem worth solving. Use general or anonymized information until approved data-handling arrangements are in place.

## Business context

Business: [BUSINESS NAME]
Sector and jurisdiction: [SECTOR / LOCATION]
Decision-maker: [NAME / ROLE]
Workflow owner: [NAME / ROLE]
Team members involved: [ROLES]

What does the business want to improve? [OUTCOME]

## Walk through a recent example

Describe a recent inquiry or task from arrival to completion. What triggered it? Where was the information recorded? Who handled each step? Where did it wait? What was repeated or copied manually?

Current trigger: [EVENT]
Input source: [SOURCE]
Current steps: [SEQUENCE]
Systems involved: [TOOLS / PLANS]
Final output: [OUTPUT]
Known failures or workarounds: [DETAILS]

## Establish a baseline

Typical weekly volume: [COUNT / DATE RANGE]
Time spent per item: [MINUTES / HOW MEASURED]
Current response time: [TIME / HOW MEASURED]
Rework or missed follow-ups: [COUNT / DEFINITION]
Data completeness: [OBSERVATION]

If these are estimates, label them. Do not present estimates as measured performance.

## Decide whether AI is appropriate

Could a standard form, rule, reminder, or clearer process solve this without a model?

Where is language interpretation, drafting, or summarization actually useful?

Which decisions must remain with people? [DECISIONS]

What happens if the draft is wrong? [IMPACT]

Which data is confidential or regulated? [CATEGORIES — NO ACTUAL SENSITIVE DATA]

What permissions, consent, professional review, and vendor terms apply? [OWNER TO VERIFY]

## Select the first pilot

One workflow: [NAME]
One problem: [BOTTLENECK]
One measurable improvement: [METRIC]
Baseline and desired threshold: [VALUES TO AGREE]
Named owner: [ROLE]
Budget and operating-cost limit: [AMOUNT / CURRENCY]

Proceed only when scope, data access, review responsibilities, and success measures are agreed. A discovery conversation is not authorization to deploy or process confidential data.