Two Months of Clean CSV Beats Any AI Strategy Deck

The first deliverable in any AI initiative is not a model. It is two months of clean CSV. Without it, every claim about AI return is an opinion, not a figure. The CSV is the baseline that makes the monthly proof possible, and the monthly proof is what turns an AI experiment into a line item your finance lead will renew (skillpilotadvisory.ai/scan, verified 15 Oct 2026).

Clean does not mean perfect. It means the columns are defined, the source is named, the figures are aggregated, and two consecutive months use the same format. A bookkeeper can produce this in a week if you tell them what columns you need.

What clean means, column by column

Column	What it holds	What a clean version looks like
Date	The period the row covers, monthly	Two consecutive months, same format, no gaps
Workload	Hours spent on the task AI is meant to touch	Aggregated, not per-employee. Named source (timesheet, service log, export).
Output	What the task produced (units, reports, cases)	One unit per row. If the unit changes month to month, the baseline is not comparable.
Cost	Loaded cost per hour or per unit	Stated, not estimated. Loaded means salary plus overhead, not just payroll.
Quality signal	A correction rate, error count, or rework indicator	Optional but valuable. Without it, the monthly proof can only judge speed, not accuracy.

How to get it from a bookkeeper in a week

1. Name the one workload AI is meant to touch. Not five. One.
2. Define the columns above. Write them on one page.
3. Ask the bookkeeper for two consecutive months in that format. They already have the raw data in the accounting system.
4. Check the format. Same columns, same units, same period length. If month one is weekly and month two is monthly, the baseline is broken.

Scan takes this CSV and builds the readiness map, the 12-month business case, and the monthly proof engine from it. AI never sets a score, changes a number, or invents a cause (skillpilotadvisory.ai/scan, FAQ, live 15 Oct 2026). The CSV is your data. The output is your judgement.

What this does not cover

Two months of CSV does not prove AI works. It proves you have a baseline to measure against. If the workload changes in month three for reasons unrelated to AI (a new hire, a process change, a market shift), the baseline is no longer comparable. State that when it happens. The monthly proof engine is designed to catch this, because it judges the new figure against your own definition of normal.

Sources and context

• Scan's monthly proof engine takes two consecutive months of CSV and nine questions, names the one line that moved, and writes the sentence to send to your boss (skillpilotadvisory.ai/scan, verified 15 Oct 2026).
• Most AI programmes fail the board test: spend rises while cost, attrition, revenue and cycle time barely move (skillpilotadvisory.ai/about, live 15 Oct 2026). The missing baseline is one of the reasons.
• Malaysia's National Guidelines on AI Governance and Ethics (MOSTI, published Nov 2024) name transparency as an expectation. A clean CSV baseline is how transparency starts.
• Malaysian enterprises often sequence AI top-down before defining the problem, which turns adoption "into an unnecessary expense" (Manminder Kaur Dhillon, CEO, Supernewsroom.ai, World AI Show Malaysia 2026, The Rakyat Post, 14 Sep 2026). The CSV is the bottom-up answer to that.

FAQ

What if we do not have two months of data? Start recording now. The baseline does not have to be historical. It has to be two consecutive months in the same format, and it can start today.
Do we need a data team? No. A bookkeeper can produce the CSV. The constraint is defining the columns, not building a pipeline.
What if the columns change? Start over. A baseline with changing columns is not a baseline. Two months in the same format is the minimum.
Is the CSV shared with anyone? No. Only the signed-in account that saved it. Uploads are aggregated figures, not customer records (skillpilotadvisory.ai/scan, FAQ, live 15 Oct 2026).

Start your data baseline on Scan