Scope Creep Is a Feature

I set out to build a simple tool. Upload a CSV, get a score that tells you if your data is ready for AI. That was the whole spec.

Three weeks later it does considerably more than that.

What it became

AI Data Maturity now generates a complete data dictionary — a column-by-column profile of every field in your dataset, including detected types, quality scores, null rates, ambiguity flags, PII detection, and semantic descriptions of what each column actually contains. It exports to Excel. Your data governance team can use it tomorrow.

It detects PII using two layers — a regex pass that catches emails, SSNs, phones, and credit cards by pattern, and Google Cloud Sensitive Data Protection, which catches names, addresses, dates of birth, and PII embedded in free text fields like notes columns. Both layers run before any data reaches AI. Your original values never leave the masking pipeline.

It profiles column quality, flags ambiguous column names by examining actual data values rather than just the name itself, identifies the business questions your specific dataset can answer, and generates starter prompts tailored to your schema — ready to paste into ChatGPT, Copilot, or Databricks Genie.

And it fills a governance gap that enterprise tools were never designed to close.

The gap

Enterprise data governance platforms are built for data warehouses and core systems. They do important work. They don’t cover the CSV export from last Tuesday, the vendor file that arrived this morning, or the staging table your analysts are already querying with Copilot. Those datasets — the ones moving between teams, being shared with partners, getting connected to AI tools without documentation — almost never have governance coverage.

ADM extends governance to exactly those datasets. No IT ticket. No onboarding. Upload and get a complete profile in under two minutes.

How it got here

None of the features above were in the original spec. Each one arrived because a conversation surfaced a problem worth solving.

A data governance professional looked at the starter prompts and said most analysts don’t know what questions their data can answer. The prompts aren’t just convenient — they solve a discovery problem.

A PhD statistician responded to a LinkedIn message in hours and handed over a three-part sampling methodology — Cochran formula, stratified random sampling, outlier inclusion — that made the tool statistically defensible for enterprise use. One message. One afternoon.

The realization that names, addresses, and dates of birth were going to AI unmasked led to Google Cloud DLP integration. Which led to the discovery that ADM wasn’t just an AI readiness tool — it was a data governance tool. Which changed the positioning. Which opened new markets.

In traditional development, this is called scope creep. It’s treated as a failure of planning.

When development is faster, following the value wherever it leads isn’t a problem. It’s the strategy.

The contradiction it resolves

Traditional data governance asks: do we have the right policies, platforms, and processes to govern our data at scale?

That’s the right question for a data warehouse. It’s the wrong question for the file your analyst just received from a vendor and is about to connect to Copilot.

ADM asks a different question: does this specific dataset have what AI needs to interpret it correctly, handle it safely, and return reliable results?

One question takes months. The other takes two minutes.

The enterprise question matters. Answer it on your timeline. But the dataset your team is connecting to AI today can’t wait for the governance committee.

What curiosity builds

Every feature in ADM arrived because someone asked a question that deserved an answer. The sampling methodology because defensibility matters for enterprise buyers. The DLP integration because the right answer to “how do you handle PII?” isn’t “we mask emails and SSNs.” The data dictionary because a governance professional looked at the output and said this is exactly what her team needs for the files nobody documents.

Curiosity is our product roadmap. Scope creep is how we follow it.