Data Analyst Agent: People Analytics in Natural Language
Ask a question in natural language and get people analytics answers in seconds. The Data Analyst Agent (also known as Gi Admin) is part of Gi's team of four AI Agents working 24/7. It does not replace your analytics team. It frees them up for the work that matters. Your organisation's data stays in your organisation, and personal data never reaches the model.
What it does
The Data Analyst Agent is a backoffice assistant for administrators, managers and customer success teams. You ask questions in plain English or Portuguese about your people, teams and activity, and it replies with structured insights, tables, trends and Excel exports drawn from your own GFoundry data.
Key capabilities
Ask in natural language: no SQL, no filters, no dashboards to configure
Reviews, IDPs, training, recognition: covers the whole people-data surface
Exports to Excel: ready-to-share files generated server-side
Personal data never reaches the model: pseudonymisation by design
Use cases
Ad-hoc analysis, reports for meetings, Excel exports. Examples of who uses it:
HR managers and team leaders: team composition, performance gaps, training needs, well-being signals
People analytics and CSMs: trends, ad-hoc reports, at-risk groups
Executives: engagement, recognition culture, talent profiles without writing a query
Examples of what to ask
The examples below are illustrative, not exhaustive. Phrasing is flexible, the agent understands intent.
Performance and evaluation cycles
"Who were the top performers in the last cycle based on their final scores?"
"Which teams have the lowest average goal completion rates?"
"Who has not had an evaluation in the last 6 months?"
Individual development and growth
"List all employees with active Individual Development Plans and their priorities."
"Who has not engaged with their development plan in the last quarter?"
People profiles and team composition
"What is the archetype breakdown of my Sales team?"
"Show me the people profiles across all departments."
"Which teams are over-indexed on Overstretched Achievers?"
Churn signals and disconnection risk
"Who are the high-risk disconnected users this month?"
"Which employees have been inactive on the platform in the last 30 days?"
"Show teams with low well-being signals."
Learning and training
"Compliance training completions by team?"
"Which training programs have the highest completion rates?"
"Who enrolled in the leadership program but did not finish it?"
Recognition and culture
"What badges are given most often in the organisation?"
"Which teams are most active in peer recognition?"
"Are there isolated groups with little recognition activity?"
Organisational directory
"Find all employees in Finance with a Manager role."
"Who works in the Lisbon office?"
"List all team leaders by department."
How it relates to the rest of the Gi team
Audience: the Data Analyst Agent is for admins and managers; Gi (the conversational assistant) is for every employee.
Data: the Data Analyst Agent answers from your live people and activity data; Gi answers from published content and uploaded documents.
Output: the Data Analyst Agent returns analytics, tables, and Excel exports; Gi returns content guidance and policy answers.
Data security and privacy
What the AI provider sees
When you ask the Data Analyst Agent a question, the service constructs a prompt to send to an external AI provider (OpenAI API, with a contractual no-training clause). Before the prompt leaves your environment, every employee reference is replaced by a pseudonymous identifier. The AI provider receives those identifiers together with the structured attributes relevant to the question, for example performance scores, completion rates, or archetype labels. It does not receive employee names, email addresses, manager names, or free-text notes.
Pseudonymisation is not anonymisation. The mapping from identifier to person stays inside your environment, which means GDPR continues to apply. Your DPO can exercise data subject rights (access, erasure, restriction) on the underlying records through the standard backoffice tools.
Where the data lives
All employee data used by the Data Analyst Agent stays in the European Union. The application, the database, and the analytics layer are all hosted in EU regions.
Strict tenant isolation
Every query is scoped to your own organisation before any value is sent to the AI provider. A request from one organisation cannot return rows belonging to another.
Read-only by design
The Data Analyst Agent can answer questions about your data. It cannot modify records, trigger workflows, or push data anywhere outside your organisation.
Excel exports
Excel exports are generated inside your environment, scoped to your organisation, and never sent to the AI provider. Names appear in the file because the file is meant for you. Exports are stored encrypted and removed after a short retention window configured for your tenant.
Controls available to your DPO
Data Processing Agreement: a DPA covering the Data Analyst Agent, including the list of sub-processors, is available on request.
No-training contract: the AI provider is contractually prevented from using your prompts or responses to train its models.
Conversation retention: prompts, responses, and exports follow the retention windows configured for your tenant.
Sub-processor changes: GFoundry notifies tenants in advance of any change to the list of AI sub-processors.
What it does not do
It does not change data. Read-only by design.
It does not cross organisations. Each admin only sees their own organisation's data.
It does not give prescriptions. The agent surfaces facts and patterns; people decisions stay with the people who own them.
It does not expose names to the model. Real names appear only in the answer or export presented to you.
Tips for better answers
Be specific. "Top 10 performers in Sales last cycle" beats "show me performance".
Mention the segment. Add the team, department, time window, or role to scope the query.
Ask for an export. If you expect more than a few dozen rows, ask for an Excel.
Follow up naturally. The agent keeps context, so "and the Marketing team?" works after a previous question.