> ## Documentation Index > Fetch the complete documentation index at: https://docs.parley.so/llms.txt > Use this file to discover all available pages before exploring further. # Overview > Putting the Parley Agent to work on bulk record work that needs to be repeatable and reviewable. The Parley Agent manages Projects, Accounts, Contacts, their links, files, notes, resource fields, and reusable Skills. Alongside the drafting it does on a single matter, it can work in bulk — in a controlled, repeatable way. This section covers that second mode: work too large or too repetitive to do one record at a time. Importing a historical client list, normalizing records after a migration, applying the same Project structure across many matters. ## Two kinds of work Producing a work product — a draft, a form, a brief, a packet — is matter-level work. See [Projects](/projects/overview). Getting the records and files into the right shape so that work comes out well is what this section is about. The Agent can prepare and organize the underlying records and source files, and list work product already generated for a Project. ## When to work in bulk Most firms use the Agent conversationally for ordinary, one-off work. Switch to a bulk run when you need a process that is repeatable, reviewable, and driven by a file you control: * Importing Accounts and Contacts from a clean CSV * Creating or updating a defined set of resource-field values * Linking existing records at scale * Organizing a file tree * Publishing or maintaining a reusable Skill For a spreadsheet migration, use the dedicated import workflow rather than creating records one row at a time. The importer validates a prepared CSV, processes it in one long-lived run, retains a progress log, and supports safe recovery after an interruption. See [Importing data](/parley-agent/importing-data). ## The batch pattern Bulk imports use a **match-or-create** model, and understanding it is the difference between an import you can trust and one you have to unpick. You select a stable natural key in the source data — an Account name, or another genuinely unique field — and the importer resolves each row against the records visible to you in Parley. A matching record is retained. A missing record is created. For Accounts, Contacts, and Projects, nonblank source values can also update the matched record when they differ from what Parley currently stores. **Blank cells are skipped**; they do not clear existing values. Links are resolved separately from the matching key. Import Accounts first, then import Contacts and link each Contact to the appropriate Account. This is what makes an import safe to rerun — as long as the source data, matching key, and progress log are preserved. A failed or interrupted batch can be corrected and resumed without blindly recreating records that were already created or matched. ## Where to go next Prepare CSVs, dry run, import Accounts then Contacts, and recover from an interrupted run. Records, fields, notes, file trees, Skills, and read-only lookups. The practices that keep a bulk operation reviewable and reversible.