AI hub

35 AI assistants at work, humans keep approval power

Where all enterprise AI is managed: assistants, pipelines, models, credits and run logs.

Screenshot 1 of 8

Sound familiar?

The problems this app removes

Capabilities

Catalogued against screens that exist in the product, grouped by the work they do — not a flat marketing list.

35 domain assistantsAI
Collections, accounting, lead screening, purchasing, inventory planning, HR, payroll, FP&A and more.
Switch assistants on or off
Pick the assistants each area needs and turn them on or off in one action.
Schedules and run conditionsAdvanced
Run assistants on a schedule or whenever data meets the conditions you configure.
Suggest-only by defaultAI
Every assistant starts at notify-and-suggest level and never writes data on its own.
Record-level assistantsAI
Run assistants on an invoice, customer or contract detail page to analyse that exact record.

Multi-step playbooksAdvanced
Chain several assistant steps into one complete automated playbook.
Scheduled or from templatesAdvanced
Playbooks run on a recurring schedule or start quickly from ready-made templates.
Triggers and quotas
Admins configure triggers, frequency limits and AI usage quotas in one place.

AI suggestion queueAI
Every action AI wants to take waits in one queue for an authorised person to approve.
Batch decisions
Approve or dismiss suggestions one by one or many at once to save time.
Mute suggestion types
Stop receiving unhelpful suggestion types so the queue stays short and focused.
AI activity feed
A timeline of everything AI has analysed, suggested and executed.

Quality monitoringAdvanced
Track error patterns and user feedback to see which assistants need tuning.
Cost per assistantAdvanced
See AI cost broken down by assistant, with trends over time and forecasts.
Per-user quotas
Cap AI usage per user to keep the monthly budget under control.
User feedback
Users rate AI results, helping admins improve each assistant's configuration over time.

Ask your business dataAI
Ask in natural language, including by voice, over data within your own permissions.
Writes need confirmationAI
Every write action from chat shows a confirmation card before anything happens.
Connect Claude via MCPAdvanced
Issue tokens so Claude and other MCP clients read XBuddy data under the grantor's permissions.
Sensitive writes await approvalAdvanced
Writes from external AI are classified; sensitive ones only create suggestions awaiting approval.

Real screens

On mobile

In practice

The IT director's first question: what can the assistant read.

1.The assistant reads as the identity of the PERSON asking, never as a service account
2.Four axes apply together: capability, module, data scope and field-level access
3.Every write is a proposal, waiting for someone with authority to approve it

Nobody reads more than they are cleared for, including when they ask through AI.

FAQ

Can AI change the books by itself?

No. Actions needing approval go to the suggestion queue; only authorised approvers execute them. Assistants default to notify-and-suggest.

Can external AI like Claude see everything?

No. Access uses MCP tokens limited to the grantor's permissions, with logging and rate limits.

How is AI cost charged?

Through your plan's credits and quotas; AI Hub shows cost per assistant plus a forecast.

Do assistants understand Vietnamese business rules?

Yes. They work in Vietnamese on XBuddy data that follows Vietnamese accounting, payroll and tax standards.

A thirty-minute demo focused on the process you are actually stuck on.