Automation

Automation means having software or machines carry out tasks people would otherwise do by hand. For federal records, the National Archives (NARA) framed the goal in its 2014 Automated Electronic Records Management Report as reducing "the reliance on all individual agency staff members to capture and manage records appropriately." The driver was volume: "manual processes will not scale up to manage the sheer volume of email, social media, and other electronic records being created."

Levels of automation

NARA's report lays out approaches from "No automation: Manual management of electronic records" up to what it called "The most advanced type of automation":

1. No automation. People handle each item by hand.
2. Rule-based automation, using "automated business rules that act on metadata, user roles, or another feature of records." Rules can also sort records "based on keyword or regular expression searches of the content of records rather than metadata."
3. Business process and workflow automation. See Workflow Automation.
4. Modular re-usable records management tools, shared across many systems.
5. Autocategorization, in which software learns categories from examples. See Machine Learning.

Rules: predictable, but blunt

NARA found that rule-based results "are predictable and consistent and require minimal work on the part of end users." It also warned that "Approaches simple enough for easy implementation may lead to over-retention of low value records, leading to higher storage costs and increased litigation risk, or failure to capture permanent records that occur in unexpected places." A rule does exactly what it says, including on items its author did not foresee.

Automation is not judgment

NARA was direct: "Automation is a tool, not a replacement for a professional records and information management program," and "Automated tools are not a replacement for sound records management practice."

NIST names the human side of the risk. Its Generative AI Profile (NIST AI 600-1, 2024) describes "automation bias, or excessive deference to automated systems," and warns that it "can exacerbate other risks of GAI, such as risks of confabulation." See Large Language Model.

Why it matters for land and mining claim records

Automation handles volume: pulling records, sorting them, comparing one date against another. The decision about what a record means, whether a filing is timely, or whether two names are the same party stays with a person who reads the documents. Automation bias is the risk of skipping that step because the output looks finished.