# Dynamics NAV Private AI Layer Blueprint Generator

> Source: https://netray.co/tools/dynamics-nav-private-ai-layer-blueprint-generator
> Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems.
> Section: Free Tools

## Short answer

Describe your NAV environment and get a shareable plan for a private AI assistant: components, security controls, 3-phase rollout and open questions.

## Category

on-prem-ai

## Erp

Microsoft Dynamics NAV

## Audience

- CIO
- IT manager
- NAV developer
- ERP architect

## Tool

**Kind:** template

**Output:** document

**Fields:**

#### Company name

**Key:** company

**Type:** text

**Default:** Harbour Components Inc

#### NAV version

**Key:** version

**Type:** select

**Options:**

- NAV 2013
- NAV 2013 R2
- NAV 2015
- NAV 2016
- NAV 2017
- NAV 2018

**Default:** NAV 2017

#### Database engine

**Key:** db

**Type:** select

**Options:**

- SQL Server 2014
- SQL Server 2016
- SQL Server 2017
- SQL Server 2019 or later

**Default:** SQL Server 2016

#### Hosting

**Key:** hosting

**Type:** select

**Options:**

- On-prem server
- Private cloud
- Air-gapped

**Default:** On-prem server

#### Named users

**Key:** users

**Type:** number

**Default:** 90

#### NAV companies or plants

**Key:** sites

**Type:** number

**Default:** 3

#### Main use cases

**Key:** usecases

**Type:** textarea

**Default:** Item availability and where-used, open purchase lines by vendor, customer balances, lot traceability lookups, search of quality records.

#### Data that must never leave the site

**Key:** neverleave

**Type:** textarea

**Default:** Customer price lists, supplier contracts, payroll, export-controlled drawings.

#### GPU available

**Key:** gpu

**Type:** select

**Options:**

- None yet
- One 24 GB class GPU
- One 48 to 80 GB class GPU
- Two or more GPUs

**Default:** One 24 GB class GPU

#### Target go-live

**Key:** golive

**Type:** text

**Default:** Q1 2027

**Blocks:**

#### NAV Private AI Layer Plan: {{company}}

**Body:** System: Dynamics {{version}} on {{db}}, hosting: {{hosting}}.
Scale: {{users}} named users, {{sites}} NAV companies or plants.
Target go-live: {{golive}}.

#### Context

**Body:** {{usecases}}

Goal: natural-language access to NAV data and documents with all processing inside {{company}} infrastructure.

#### Components

**Table:**

**Columns:**

- Component
- Design
- Notes

**Rows:**

**0:** Data access path

**1:** Read-only SQL replica with per-company views; published pages, queries and codeunits as web services for logic-bound reads

**2:** Company name is part of every NAV table name

**0:** Change detection

**1:** Timestamp columns or SQL change data capture feeding the index

**2:** Avoid full nightly rebuilds

**0:** Retrieval index

**1:** Attachments, comments, item text, external documents linked from records

**2:** Local vector store

**0:** Model serving

**1:** Open-weight model via vLLM or Ollama, GPU: {{gpu}}

**2:** Quantise if memory is tight

**0:** Tool server

**1:** MCP server with named read actions, parameter checks and row limits

**2:** No free-form SQL from the model

**0:** Identity and permissions

**1:** Map Windows user to NAV user; apply permission sets and company access in tool code or via delegation

**2:** Test with a low-privilege user

**0:** Audit

**1:** Local log of question, calls, queries and answer

**2:** Align retention with policy

#### Security controls

**Table:**

**Columns:**

- Requirement
- Control

**Rows:**

**0:** Never leaves site: {{neverleave}}

**1:** No external APIs; inference and index inside the {{hosting}} boundary; egress blocked

**0:** Least privilege

**1:** SELECT-only SQL login on curated views; narrow web service identity

**0:** Calculated fields

**1:** FlowField values recomputed from ledger entries and reconciled to NAV pages

**0:** Write-back

**1:** Drafts only, human approval, posting through codeunit or page web services

**0:** Traceability

**1:** Each answer cites the view or service behind it

#### Three-phase rollout

**Table:**

**Columns:**

- Phase
- Scope
- Exit criteria

**Rows:**

**0:** 1. Read-only questions

**1:** Pilot group, curated views, document search

**2:** Agreed test set answered correctly and permissions verified

**0:** 2. Drafts

**1:** Drafted documents and summaries reviewed by users

**2:** Low rework rate in review

**0:** 3. Approved actions

**1:** Small list of actions through web services

**2:** Audit log reviewed by IT and finance

#### Open questions for IT

**Body:** 1. Can we provide a readable SQL copy and how current must it be?
2. Which pages, queries and codeunits are already published as web services?
3. Which authentication is configured on the service tier?
4. Are security filters used on tables the assistant will read?
5. Which C/AL customisations change standard posting or calculations?
6. What is the plan for moving to Business Central, and should views be designed for repointing?

**Signatures:**

- Prepared by
- IT architect
- Business owner
- Security reviewer

**Notes:** A starting plan. Verify object names, web service availability and OData version for your NAV build and customisations.

## Content

**Intro:** Generate a plan for putting a private AI assistant beside Dynamics NAV. The document covers the SQL and web service access paths, retrieval, model serving, permissions and a phased rollout, ready to share with IT and security.

**Sections:**

#### How to use the generator

**Body:** Choose your NAV version and database engine, then describe the questions users should be able to ask. Record the data that must stay on site. The output includes a components table, controls tied to your inputs, a three-phase rollout and a list of questions for your DBA and NAV developers.

**Bullets:**

- Pick the exact NAV version
- List published web services you already have
- State if multiple companies share a database
- Export to Word for review

#### The design behind the plan

**Body:** NAV stores data in SQL Server with the company name inside every table name, and exposes pages, queries and codeunits as web services. The plan uses SQL on a replica for broad reads and web services where NAV logic and permissions matter, with a tool server in front and an open-weight model on local GPUs.

#### Common mistakes

**Body:** Reading FlowField-style values straight from tables, forgetting the company prefix, running the assistant with a super permission set and writing to tables with SQL. The plan flags each of these and keeps write-back behind approval and supported services.

**Cta:** If you would like help validating this plan against your NAV customisations, Netray can review it with your team.

**Faq:**

#### Does it work for NAV 2009?

**Answer:** The generator targets NAV 2013 to 2018, where web services and SQL Server are standard. NAV 2009 deployments with the classic client need a different integration approach.

#### Is anything uploaded?

**Answer:** No. The document is assembled in your browser from the values you type.

#### Does it assume cloud services?

**Answer:** No. Every component runs inside your network, including the model and the audit log.

#### Can I reuse it for Business Central?

**Answer:** Partly. The structure applies, but Business Central has API v2.0 and different table naming, so use the Business Central generator.

## Search terms

Dynamics NAV AI architecture, NAV private LLM blueprint, NAV OData AI integration plan, NAV 2018 on-prem AI, NAV web services AI design, air-gapped NAV AI, NAV item ledger entry AI

---
Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json