SERVICES
│ Ph.1 Current State (Data)
Hypothesize an AS-IS business workflow from CSV data
Starting from nothing but CSV data and forms, use Claude Code to hypothesize an AS-IS business workflow. A hands-on course aimed at locking down the AI-adoption scope and sharing assumptions before discovery.
─ ARTICLE COVER
Consulting Skills ／ Ph.1 Current State (Data)
§I ─ POSITION
Ph.1 "Current State (Data)." Before you go into discovery, experience how far you can hypothesize operations from data alone. A skill for "preparing without stealing the field's time."
§II ─ TAKEAWAYS
Work backward from the column structure and volume trends of order data to the flow of operations
Automate the CSV-analysis → workflow-hypothesis path for data using Claude Code
Walk into discovery with pinpoint "points you want to confirm"
§III ─ EXCERPT
¶ 01
Gaps in sequential order numbers mean a missing manual entry or a separate input channel. A large gap between due dates and actual dates means there is an adjustment process. A skew in the owner column points to authority and person-dependence. Data can be read as a mirror of operations.
¶ 02
An "anomaly," "skew," or "inconsistency" in the data points to a specific operational pattern.
| What you observe in the data | Operational hypothesis |
|---|---|
| A large gap in order numbers | A separate input channel exists (FAX → manual Excel entry, etc.) |
| A cluster where due dates and actual dates differ widely | An adjustment process / handling of exceptions exists |
| The owner column skews toward one person | Person-dependence / concentrated authority is likely |
| Input concentrated in a specific time slot | Last-minute evening entry / batch processing right before month-end close |
| Rows where the amount is "-" or zero | Return processing / cancellation work exists |
| Duplicate primary keys | A re-issuance / correction pattern |
¶ 03
Running this in order lets you draft an AS-IS business workflow in 30 minutes.
¶ 04
Once you build hypotheses from data, the confidence will inevitably vary. Annotating with three levels — "certain from this column name" (○), "inferred from this distribution" (△), and "not in the data, but common in practice, so tentatively assumed" (×) — automatically decides what you need to ask in Ph.2 discovery.
¶ 05
The AS-IS workflow you build in Ph.1 is not a "correct answer" but a "starting point for discussion." Pick 3–5 ×-confidence (tentative) points and, at the very start of discovery, present them: "Is this understanding correct?" This alone turns discovery from "a place where the field explains everything from scratch" into "a place where you confirm an existing hypothesis."
¶ 06
Handing CSV analysis entirely to the AI won't produce good results. Giving instructions in the following order makes it easier for the AI to reach operational hypotheses.
¶ 07
If you skip Ph.1 and go straight to discovery, you'll fire off beginner questions in front of the field and lose their trust in an instant. Show up to discovery after building 70% of the operational hypothesis from data, and the field's time is done in 30 minutes. Show up without checking the data, and the same discovery takes two hours — and you'll still need follow-ups. Preparing in Ph.1 ultimately minimizes total effort.
§∞ ─ DOWNLOAD
─ EXERCISE
Using Azika Foods' order data CSV (covering multiple years) as the material, you sketch out the AS-IS business workflow. It comes with company-background slides, announcement slides, and a flow guide, so it also works as a template for running an internal study session.
Lecture slides
Every slide of the course
Cheat sheet
Quick reference for the field
Exercises
With answers and grading
Advanced / test
Hands-on plus rubric
INCLUDED PDFS ─ 6files
§IV ─ RELATED