LAKEWRIGHT FOR CONSULTANTS

The engagement, compressed

Walk in with a laptop.
Walk out the hero.

Your client's data estate is a decade of file shares, vendor drops, and databases nobody fully remembers. The old playbook maps it in weeks of interviews and sampling. You're about to do it before lunch — on your laptop, with their data never leaving their building.

See the day What you leave behind

01 / THE OLD ENGAGEMENT

Three weeks of asking people what's in their own data.

Discovery interviews. Spreadsheet inventories that are wrong before they're finished. Opening a few dozen sample files by hand and extrapolating. A deliverable built on what people believe is on the share — and an awkward silence when the believe turns out wrong in month two.

the traditional assessment

2–4 weeks of stakeholder interviews
manual sampling: dozens of files out of millions
PII exposure: unknown, assumed, or worse — asserted
deliverable: a slide deck that ages like milk

02 / THE DAY

One engagement day, timestamped.

Everything below is the shipped product doing what it does. No slideware.

on-site · your laptop · their wallsread-only credentials

9:04

You get read-only credentials

A service account on the file server, or a bucket key. That's the entire setup. No agents installed on their systems, no data egress to clear with security — the scan runs on your machine, inside their network.

9:12

Point it at the estate

lakewright scan smb://fileserver01/finance — a native connection, no mount, no admin rights. S3 and cloud buckets work the same way. It reads heads of files, never whole objects, and never writes to the source.

10:30

The inventory exists

Every feed identified by evidence: the NACHA files, the X12 claims, the fixed-width mainframe extracts, the Excel dumps, the scanned PDFs, the email archives with the real data in the attachments. Schemas profiled, duplicates flagged, drift surfaced.

11:45

The moment

You show them the report. There is always something — PII sitting in a share nobody remembered, the same customer data in four inconsistent copies, a "temporary" export from 2019 still syncing. You found it before lunch, read-only, with receipts. This is when you become the person who knows their data better than they do.

13:30

From findings to foundations

The plans the agent proposed are plain YAML. The client's owner reviews and approves the feeds that matter — the human gate is theirs, not yours — and the first governed Iceberg tables land that afternoon.

15:00

Make their AI features work

lakewright semantics --dialect snowflake — generated semantic views over the landed tables, so Cortex Analyst (or Genie, or Fabric Copilot) can answer questions about data that was an unreadable EDI drop this morning. Their platform, suddenly earning its license fee.

16:30

The deliverable writes itself

The audited scan report, the decision log, the approved plans, the landed tables, the semantic views. Not a deck of impressions — a database of findings, every one traceable to file and byte.

03 / WHAT YOU LEAVE BEHIND

The assessment ends. The value doesn't.

The client keeps the engine

The watcher keeps landing new drops under the plans they approved. Your assessment becomes their running infrastructure — and you're the one who installed it.

The audit trail is the report

Every finding traces to a file, a digest, and a recorded decision. When someone challenges a conclusion in the readout, you answer from the log, not from memory.

Follow-on work finds you

The inventory always surfaces more than one engagement's worth of problems. Remediation, migration, governance — scoped from evidence you already gathered.

Your kit fits in a repo

One compose file: Lakewright + its audit database, up on any laptop with Docker in one command. Same kit, every engagement, no client-side install.

The consultant's oldest problem is proving value fast enough to be trusted with the real work. This is proof by lunchtime — governed, audited, and left running when you walk out.

Request early access Solution briefs