Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.xpander.ai/llms.txt

Use this file to discover all available pages before exploring further.

Every employee in your company uses a different set of tools, works on different projects, and has different questions. A Personal AI Agent sits in Slack, Teams, or voice and handles all of that for them. It reasons across your connected systems, delegates to Specialized Agents when it needs governed access, and learns each person’s preferences over time. IT deploys it once to a workspace, and every employee gets a personalized agent automatically.

​
How it works

An employee asks a question in Slack. The agent figures out which systems to query, delegates to the right specialized agent, and returns a coherent answer, all in one thread.
Personal AI Agent delegation flow
Personal Agent ≠ Specialized Agent. The personal agent handles conversation and reasoning. Specialized agents handle governed access to specific systems. Delegation happens automatically, the user doesn’t choose which agent to talk to.

​
Example

Here’s what this looks like in practice. A compliance deadline is approaching, and the agent proactively flags it: Aria (Personal AI Agent) - 9:02 AM
Morning, Mike. Your compliance audit deadline is tomorrow. I noticed the PII classification for the Redshift warehouse hasn’t been filed yet. Want me to run it now?
Mike - 9:03 AM
Yes please, go ahead.
Aria - 9:03 AM
Delegating to the Data Agent. Scanning 3 schemas across your Redshift warehouse…
Aria - 9:06 AM
Done. 47 tables scanned, PII found in 12 (23 columns flagged). Report added to your audit folder.
The personal agent identified a deadline, delegated a scan to a specialized agent, and reported back, all in a Slack thread. Mike didn’t need to open Redshift, write a query, or remember the audit deadline himself.

​
Memory

Without memory, every conversation starts from scratch. The agent wouldn’t know your projects, your preferences, or what you discussed yesterday. Personal AI Agents solve this by maintaining three levels of memory:
LevelWhat it storesExample
SessionCurrent conversation context”We were discussing the Acme deal”
User memoriesIndividual preferences across sessions”Mike prefers Slack over email”, “Works on enterprise pipeline”
Agent memoriesOrg-wide knowledge shared across users”Fiscal year starts in April”, “Escalate billing to #finance”
After a few weeks of use, the agent knows each person’s projects, preferences, and patterns. It stops asking clarifying questions it already knows the answer to.

​
Access control

AI with unchecked access to enterprise systems is a liability. Personal AI Agents separate what the agent can reach from what individual employees can delegate, so IT stays in control while employees still get things done.
ControlWho sets itWhat it governs
System accessIT / AdminWhich enterprise systems the agent can reach
OperationsIT / AdminWhich actions are allowed per system
Model selectionIT / AdminWhich AI models the agent can use
DelegationEmployeeTasks delegated using their own permissions
Each specialized agent the personal agent delegates to operates within its own trust boundary, scoped to only approved operations, with full audit trails.

​
Model selection

Personal AI Agents are powered by OpenClaw, xpander’s multi-model engine. OpenClaw selects the best model for each task automatically across providers (OpenAI, Anthropic, Google, and others), optimizing for cost and quality. For air-gapped deployments, it routes to local models instead. You can also pin a specific model per agent if you prefer.

​
What’s next

Specialized Agents

The governed agents that Personal AI Agents delegate to.

Quickstart

Build and deploy your first agent in 5 minutes.