How It Works
A column-by-column AI Readiness and Data Governance Assessment in nine steps.
Step 01
Upload Your Dataset
Upload your data as a CSV or Excel file, or connect directly to a MySQL database. No account needed for file uploads. Your file is read in memory to extract column statistics and never written to disk or retained after your assessment completes. Additional database types (e.g. Postgres) are on the roadmap.
Step 02
Statistical Sampling
We calculate a statistically valid sample scaled to your dataset size, ensuring proportional representation across data categories with outlier values always included.
Step 03
PII Scanning & Masking
Before any data is sent to AI, we automatically scan every column for personally identifiable information — including email addresses, phone numbers, Social Security numbers, and credit card numbers. Detected values are masked with placeholder tokens (e.g. [PII:EMAIL], [PII:SSN]). AI never sees your original sensitive values. This step runs on our servers before AI analysis begins.
Step 04
Deep Dive Analysis
For every column, we examine the actual data values — not just the column name — to determine true semantic meaning. This proprietary analysis resolves ambiguity where possible, producing a richer, more accurate Dataset Briefing than column names alone can provide.
Step 05
Column-by-Column Analysis
Every column is assessed for data quality, naming clarity, PII/HIPAA risk, and how likely AI is to misinterpret it. Results arrive in under a minute.
Step 06
Your AI Data Maturity Score
A score from 0–100 tells you exactly where your dataset stands — from Not AI Ready to AI Ready — with specific findings for every column.
Step 07
Your ADM Data Dictionary
Every assessment automatically generates a complete ADM Data Dictionary — column names, detected types, quality scores, null rates, ambiguity flags, PII detection, and semantic descriptions. Export to Excel for team sharing, onboarding, and data governance documentation.
Step 08
Your Dataset Briefing
Generate a draft column definitions document for your AI tool — describing what each column contains and how AI should interpret it. Review and edit before use.
Step 09
Your Starter Prompts
Get 5 starter prompts built from your specific dataset. Each one identifies a question your data can answer and provides the exact prompt to ask it in any AI tool — ChatGPT, Copilot, Databricks Genie, or others.
Why nulls matter more for AI than traditional reporting
In traditional reporting, analysts know to handle nulls explicitly — they write WHERE clauses and aggregate functions with null awareness. AI analytics tools like Databricks Genie and Microsoft Copilot generate queries automatically. When they encounter a column with 30% null values, they may silently drop those rows or return an incomplete result — with no visible warning.
The AI Data Maturity context block fixes this. For columns with significant nulls, it tells the AI exactly what null means in that column and how to handle it — so queries are written correctly the first time.
How We Sample Your Data
Rather than scanning your entire dataset, ADM analyzes a statistically representative sample — large enough to identify data quality patterns with confidence, small enough to deliver results in seconds.
We use a three-part methodology:
Stratified random sampling
Rows are drawn proportionally across data categories, ensuring no segment is over or underrepresented.
Dynamic sample sizing
Sample size is calculated using the Cochran formula at 95% confidence and ±5% margin of error, scaled to your dataset size.
Outlier inclusion
Minimum and maximum values from numeric columns are always included, ensuring edge cases are never missed.
Sampling methodology developed in consultation with Bruce Ratner, PhD, Predictive Analytics Consultant.
How We Handle Your Data
Column statistics go to AI — not your data
We send column names, data types, null rates, and sample values to OpenAI for analysis. Before transmission, we automatically scan your data for PII using two masking layers: a regex layer for structural patterns (email, SSN, credit card, phone) and Google Cloud Sensitive Data Protection (DLP) for semantic PII — including person names, street addresses, and dates of birth. Detected values are replaced with placeholder tokens (e.g. [PII:EMAIL], [DLP:PERSON_NAME]). OpenAI never sees your original PII values. This masking runs on our servers before AI analysis begins.
Assessment accuracy
Results are directionally accurate for identifying data quality and readiness issues. The AI Data Maturity Score is a practical assessment guide — not a formal statistical audit. For datasets where precision matters, we recommend reviewing flagged columns manually.
Free to start — no account, no credit card, no catch, for files up to 1,000 rows / 20 columns. Larger files or database connections use a low-cost credit system.