Home / A major county in California

1.1 Million Calls, None of Them Leaving the County

One of the largest counties in California, serving over 3 million residents, wanted to analyse its health and human services call recordings to understand caller satisfaction. The recordings were full of PII, and California law meant they could not leave the county's possession to be analysed.

Results

1.1M

Recordings redacted a year

Zero

PII leaving the jurisdiction

Private cloud

The county's own environment

Unblocked

An analytics programme that was not otherwise lawful

The Constraint

The county wanted to use an experience management platform to analyse call quality. The recordings contained names, phone numbers and Social Security numbers.

Under CCPA and CPRA, those recordings could not be shared with third-party providers outside the state. The only route forward was to redact every recording before it left county hands.

That set two hard requirements, and most redaction tools fail at least one of them:

01

Volume.

The county receives 1.1 million call recordings a year. Any per-file, operator-driven workflow was arithmetically impossible.

02

Locality.

The software had to run inside the county's own four walls, so no PII crossed the jurisdiction at any point — including during processing.

A cloud redaction service fails the second requirement by definition, because the content leaves to be processed.

What Was Deployed

Redactor was deployed into the county's own private cloud — the same environment already holding the call recordings. Processing happened where the data already lived, so nothing crossed the jurisdictional boundary.

Bulk audio redaction handled the volume: queue-based processing running unattended against the archive rather than one file at a time, with spoken PII detected and removed automatically.

The redacted recordings could then be shared with the analytics provider without infringing state privacy law, while the originals stayed in county custody.

Why It Worked

The deciding factor was not detection quality. It was where the software could run.

For a jurisdiction-bound workload the question is whether content crosses the boundary at all, and a hosted redaction API answers that question the wrong way. Redactor runs in shared SaaS, dedicated cloud, private cloud, on-premises and fully air-gapped, with the same engine in each.

Same Constraint?

If your content cannot leave your network, tell us where it has to stay and we will tell you which deployment answers it.

Ask for a reference

Every figure on this page comes from the deployment it describes. For a serious evaluation we will arrange a call with an organisation doing something close to what you are doing.