Skip to main content

User Extraction

Also available in workflows

This specialist runs in chat and as a node you can add to a workflow for repeatable, scheduled runs. See the Agent Catalog for how the two relate.

Overview​

People data arrives as resumes, CVs, and employee records — useful, but not in a form you can load into a system. User Extraction reads a document you attach and turns it into two clean, structured tables.

The first table is a profile summary, one row per person, with contact and background details. The second is a work-history table, one row per role, linked back to each person by email so the two stay connected.

Ellie extracts only what the document states and leaves fields blank when a detail is missing. Because this involves personal information, it handles the data carefully and stays compliance-minded — you remain responsible for storing and using the extracted data in line with your organization's privacy policies.

What you'll need​

How to ask​

Attach the document, then ask Ellie to extract the profiles.

Extract a single resume:

Extract the profile and full work history from the attached resume.

Process a batch of records:

Pull every person in the attached employee export into a profile table
and a separate work-history table.

What you get​

Two downloadable spreadsheets:

  • User Summary — one row per person, with name, contact details, job title, department, company, education, skills, certifications, languages, and related background fields.
  • Work Experience — one row per role, with the person's email, job title, company, start and end dates, responsibilities, achievements, and skills used.

The shared email column links each work-history row back to the right person. Blank fields mean the detail wasn't in the document.

Tips​

  • Clear sections and headings in the source produce cleaner, better-linked results.
  • Consistent email addresses keep each person's roles tied to the right profile.
  • Nothing is invented — a blank field means the information wasn't present.
  • Scanned images need to be converted to text first; text-based documents work best.
  • Review the profiles before importing them, and handle the personal data according to your privacy obligations.