Ask-AI on any screen
You type what you want done and the agent does it. It doesn't open the form for you: it calls the action, runs it and hands back the result with a link to the record it touched.
The platform
One platform, not a bundle of apps.
Data, permissions, automation and AI agents are the same across every module. Turning one on extends what the system can do; it doesn't add another app to maintain.
The iERP platform is the shared foundation of its 18 modules: one database per workspace, a single permission model, an automation engine and AI agents able to run real actions against that same data. Every module you enable adds its screens and its actions to that single system.
No lock-in. You can export the whole workspace whenever you want.
Pillar 1 · Agentic software
AI agents that execute, not suggest
An agent gets a goal, a set of tools and a set of permissions. From there it works on its own: it queries, decides, asks for confirmation before writing and leaves a trace of every step.
The same conversation, by voice
A real-time voice agent that can drive the interface while it talks to you. It confirms out loud before any write, in English and in Spanish.
Published actions
Every module publishes its operations as MCP tools: list, create, approve, invoice, notify. That catalogue is exactly what an agent can ever do — no more, no less.
379
Trace and cost per conversation
Each conversation records which tools were called, under which permissions, with what result and at what cost. You can review it, evaluate it and cap the spend.
An agent never holds more permissions than the person using it: authorisation is checked twice, when the tool is chosen and again when it runs.
Pillar 2 · Automation
Any repeated process, running by itself
If you can describe the process — when it starts, what it looks at, what it decides, who it notifies — it can be automated. It doesn't have to be an "ERP" process.
Flow canvas
You place steps, conditions and waits on a canvas. Each step is a real module action: the very same one a person or an agent would run.
Triggers
A status change, a date, a submitted form, an incoming webhook or a time of day. The flow starts on its own, with nobody having to remember it.
People inside the flow
A step can be a question to someone. The flow waits for the answer, records it and carries on down the matching branch.
Where the saving is
In the time spent today copying data from one screen to another, chasing approvals and notifying people by hand. The flow does it the same minute the event happens, and always the same way.
End-to-end example
A deal is won
- CRM marks the opportunity as won.
- The customer and its project are created, with an owner assigned.
- The first invoice is issued in the matching series.
- The team and the customer are notified, each in their own language.
- It all lands in the audit log, with who or what did it.
All five steps are real actions from CRM, Projects, Finance and the notification engine. The flow chains them; you only decide the order and the conditions.
Pillar 3 · One single system
One system, not six tools wired together
What elsewhere are integrations, here is the same database. Nothing to sync, and no piece of data existing twice with two different values.
Self-service workspaces
You create the workspace, pick the modules and you're in. Each one has its own data, users and billing, and you switch between them without signing in again.
⌘K everywhere
A single search box to jump to any screen, find a record or fire an action without leaving the keyboard.
Notifications that actually arrive
In the app, by email and as a daily digest. Each person decides what they want and through which channel.
Built-in time tracking
Clock-in, absences and the working calendar in the same system: what a person clocks is what their manager sees and what counts towards the project.
English and Spanish, in full
Interface, notifications, generated documents and the agents themselves. Language belongs to the person, not to the company.
Light and dark
Both themes are looked after screen by screen. It's chosen per person and remembered between sessions.
Security and compliance
Control over who sees what, and proof of what happened
Concrete mechanisms you can check, not badges or certifications we don't hold.
Per-workspace encryption
Every workspace has its own data key. Sensitive fields are stored encrypted, and one workspace's key does not open another's data.
Permissions by role, module and section
Access is granted through specific capabilities, not by "being an admin". AI agents and downloads of documents holding personal data go through those very same gates.
Action audit log
Who did what and when, including actions executed by AI and downloads of documents holding personal data, with the stated reason.
Chained invoice register
Every issued invoice is sealed with a SHA-256 hash chained to the previous one: the series cannot be altered unnoticed. VeriFactu ready; certificate signing comes later.
Export and archiving
You can download a full export of the workspace whenever you want. When you archive it, the data is kept for 30 days before being deleted for good.
Frequently asked
What people ask before trying it
Three straight answers about what this is and what it takes to start.
The agent performs the action instead of telling you where the button is. A bolted-on chatbot answers from the manual; an iERP agent holds tools over the workspace data, acts with the permissions of whoever is talking to it and leaves every write it makes in the audit trail.
Try it on a process of your own
Look at the plans and get started, or ask for a demo built around a real process from your company.
No lock-in. You can export the whole workspace whenever you want.