---
title: "AI that works on your real records"
description: "1flux is AI ERP software built AI-native: one data model, one permissioned command layer and one audit trail. Today its AI turns a customer's PO into a quote."
url: https://1flux.ai/platform/ai
last_updated: 2026-10-11
---

The AI-native ERP

# AI that works on your real records

1flux is AI ERP software with one data model, one set of permissions and one audit trail for people and AI. Today it reads a customer's PO and drafts the quote.

## In short

1flux is AI ERP software built AI-native. CRM, quotations, purchasing, inventory and accounting share one data model, and actions run through one command layer that checks permissions and keeps an audit trail. That's why AI in 1flux can do real work inside your rules. Today it powers **Import PO with AI**: drop in a customer's PO or RFQ and 1flux drafts the quotation, matched to your catalogue, for your team to review, approve and send.

Import PO with AI: a customer's PO becomes draft quotation lines, matched to your catalogue, for you to review.

View as text

1. A salesperson opens a new draft quotation and chooses Import PO with AI.
2. They drop in the customer's purchase order, PO-4471.pdf.
3. 1flux imports 8 lines with a 94% confidence score: it fills the customer reference (PO-4471), the quote date and the currency (USD), and matches each item code to the catalogue.
4. The salesperson chooses the customer and checks the totals. The source file stays attached to the quotation.

The problem

## Why does bolted-on AI fall short in an ERP?

Bolted-on AI falls short because it sits beside modules that each keep their own data: it sees one part of the business and can run only a handful of scripted actions.

Most ERP software now claims AI. Often it’s a chat window added to an older system, with its own copy of the data and its own permissions to maintain. For a trading business, the useful test is simpler: does the AI work on the real records, inside the real rules, and take work off your team’s desk?

1flux was built the other way round. The AI sits inside the one data model every module shares, so it works on live records and through the same permissions and audit trail as your team.

|                         | Bolted-on AI                                  | AI-native: 1flux                                                                 |
| ----------------------- | --------------------------------------------- | -------------------------------------------------------------------------------- |
| **Where the AI sits**   | A chat window beside separate modules         | Inside the one data model every module shares                                    |
| **The data underneath** | The module it was added to, or a copied index | One connected record of every quote, order, stock movement and ledger line, live |
| **What it can do**      | A handful of scripted actions                 | Real work on real documents: today, a draft quotation from a customer’s PO       |
| **Permissions**         | Set up again for the AI add-on                | The same roles and record rules as your team                                     |
| **Audit**               | Hard to tell what the AI changed              | Every action recorded with its source                                            |

Bolted-on AI sits beside the modules. In 1flux, AI works inside the one data model and the same rules as your team.

Built in, not bolted on

## Four foundations under the AI in 1flux

1flux AI works on four foundations that every module shares. They’re the same foundations your team works on every day.

- **One data model.** CRM, Inventory and Accounting are three apps on one login and one data model. Every document links to the one before it, from quotation to sales order, delivery note, invoice, payment and journal.
- **One command layer.** Actions run through permissioned commands that check who is acting, can be retried without doing the work twice and leave an audit trail. Stock commands already record their source: a person, an import, the system or AI.
- **Permissions on every record.** Roles built from more than 100 permissions, record visibility along reporting lines, limits by legal entity and [row-level security](https://1flux.ai/glossary/row-level-security) in the database, not just on screens. See [access control](https://1flux.ai/platform/access-control).
- **Rules as configuration.** Workflow stages, approvals and required fields are data, not code, so the rules AI meets are the rules you set. See [workflows and approvals](https://1flux.ai/platform/workflows-and-approvals).

That’s what AI-native means in 1flux. Read [why the architecture matters](https://1flux.ai/platform/ai-native-erp) for a trading business.

1. AI that works today

   - Import PO with AI Customer PO or RFQ in, draft quotation out

2. Rules as configuration

   - Workflow engine Stages, approvals and required fields are data, not code

3. Permissions on every record

   - Permissions Roles, record visibility, entity access, row-level security

4. One command layer

   - Commands Permissioned actions that record their source: a person, an import, the system or AI

5. One data model

   - One data model CRM, Inventory and Accounting share linked records, not copies that sync

The AI that works today sits on the same rules, permissions, commands and records as everything else in 1flux.

Import PO with AI

## What does AI ERP software do in 1flux today?

Today, 1flux AI turns a customer’s purchase order or RFQ into a draft quotation. **Import PO with AI** sits in the quotation builder: drop in the document and the draft fills in.

- **Reads PDFs and images:** PDF, PNG, JPEG or WebP files of up to 10 MB, including a scan or a phone photo.
- **Drafts up to 200 lines**, with the customer’s reference, the document date and the currency.
- **Matches item codes** to items in your catalogue, with your unit and tax code.
- **Shows a confidence score** for the whole extraction.
- **Keeps the source file** on the quotation, with an **AI source** badge.

[How Import PO with AI works](https://1flux.ai/platform/ai/import-po)

Item codes on the customer's PO are matched to your catalogue, and the draft lines arrive for you to review.

View as text

1. A customer sends purchase order PO-4471 as a PDF. A salesperson chooses Import PO with AI in a new quotation and uploads it.
2. While the button reads “Reading PO…”, 1flux extracts the customer’s reference (PO-4471), the document date (8 October 2026), the currency (USD) and the lines.
3. Each line’s item code is compared with your catalogue. Exact matches, such as 0101-0012, become catalogue lines with your unit and tax code. HX-2210 has no exact match, so it arrives as a custom line without a tax code.
4. The draft shows “Imported 8 lines from PO-4471.pdf” with a 94% confidence score. The source file is attached with an AI source badge. The salesperson chooses the customer and reviews every line before saving.

Control

## How does 1flux keep AI inside your rules?

1flux puts AI inside the same permissions, drafts and approvals as your team, and keeps the source of its work on record.

- ### Same permission as a person

  Import PO with AI needs the same permission as creating a quotation by hand.

- ### Draft first

  AI fills a draft quotation. A person reviews it, chooses the customer and saves.

- ### Your approvals apply

  If an approval rule matches, the quotation can’t be sent until it’s approved.

- ### Source on record

  The customer’s file stays attached with an **AI source** badge, next to a confidence score.

- ### People send

  A person sends every quotation, by email, secure link or WhatsApp.

- ### Documents can’t give orders

  Instructions written inside an uploaded file are treated as data, not followed.

[How 1flux keeps AI inside your rules](https://1flux.ai/platform/ai/control)

FAQ

## Questions, answered

Still deciding? [Talk to sales](https://1flux.ai/contact)

### What makes 1flux AI ERP software, not an ERP with AI added?

1flux was built on one data model and one permissioned command layer, so its AI works on the same live records, approvals and audit trail as your team. An ERP with AI added usually connects a separate tool to modules that each keep their own data. In 1flux, that architecture powers Import PO with AI today. Read more in [AI-native vs bolted-on ERP](https://1flux.ai/platform/ai-native-erp).

### What does AI do in 1flux today?

1flux uses AI to turn a customer's purchase order or RFQ into a draft quotation. **Import PO with AI** reads a PDF, PNG, JPEG or WebP file of up to 10 MB, fills the customer reference, date and currency, drafts up to 200 lines and matches item codes to your catalogue. A confidence score shows how sure the extraction is, and a person reviews every line before saving.

### Does a person check AI work before it reaches a customer?

Yes. Import PO with AI only fills a draft quotation. Someone on your team reviews the lines, chooses the customer and saves. The quotation then follows your normal process: if an approval rule matches, for example on the total amount, it can't be sent until it's approved. Sending by email, secure link or WhatsApp is always a deliberate step by a person.

### Who can use AI in 1flux?

1flux lets anyone whose role can create quotations use Import PO with AI. It sits in the quotation builder next to **Browse catalog** and **Add item**, so it needs the same permission as building a quotation by hand. The draft it creates is an ordinary quotation, with the same approval rules, locked PDF and numbered revisions as any other.

### Which AI model does 1flux use?

1flux uses large language models from established providers through a model gateway, so it can use the model best suited to reading business documents without changing how the feature works for you. Whatever the model, the same guardrails apply: a fixed output structure, documents treated as data rather than instructions, a confidence score and a person reviewing every draft.

### What are the foundations of AI in 1flux?

1flux uses AI-native to mean the AI works on the same structured, connected and permissioned records your team uses, rather than being bolted onto spreadsheets and disconnected apps. Every document links to the one before it, actions run through one command layer that records their source, and workflows are configuration, not code. Today that foundation powers Import PO with AI.

Related

## Keep exploring

### [Import PO with AI](https://1flux.ai/platform/ai/import-po)

Drop in a customer's PO or RFQ and get a draft quotation, matched to your catalogue, with a confidence score.

### [Control and trust](https://1flux.ai/platform/ai/control)

How AI in 1flux stays inside the same permissions, drafts, approvals and records as your team.

### [AI-native ERP](https://1flux.ai/platform/ai-native-erp)

What AI-native means, how it differs from AI bolted on, and the questions to ask any ERP vendor about its AI.

### [Quotations](https://1flux.ai/products/quotations)

Quotes from your catalogue or a customer's PO, approved, locked and sent.

### [Platform](https://1flux.ai/platform)

One data model, one command layer and one set of rules under CRM, Inventory and Accounting, with AI on top.

Last updated 11 October 2026