IntelligenceHub

Your code and your data, finally in one conversation.

IntelligenceHub connects Claude to your Dynamics NAV / Business Central estate — the logic in every object and the live numbers in every ledger, linked by a reference graph between them. Read-only, scoped to each user's own NAV/BC permissions, audited by name.

Logic + data + the graph between them — within your permissions
T18 CU80 T36 CU50123 P42

The problem

Two decades of customization. Zero shared memory.

Every mature NAV/BC estate is an archaeology site: layered modifications, undocumented dependencies, and logic that lives only in the heads of the people who wrote it. The shape of a typical estate:

1000sof objects in a mature corpus
4version layers stacked on standard (base, localization, partner, client)
Allcompanies and countries sharing one customized codebase
?people who can say what changing one field will break

What it is

A direct line between AI and your codebase

IntelligenceHub is a corporate web application with a built-in MCP server that makes your NAV/BC system conversational: permission-scoped, read-only access to both the logic (objects) and the data (SQL tables) — so the AI works with the real code and the real numbers, not a guess.

Your NAV / BC estateC/AL or AL objects, all version layers, all companies
IntelligenceHubParses and indexes the code, guards the data. Read-only by design, permissions derived per user.
Claude / AI assistantConnects via token or Entra SSO — 15 tools spanning code, data and the reference graph

Ask in the web UI or from your own Claude — same permissions, same audit trail either way.

Capabilities

What it does with your estate

Full-corpus browsingTable 18 · Codeunit 80 · Page 42

Search and read any object across thousands — tables, codeunits, pages, reports — without opening a development environment.

Version-layer diffingNAVW1 → NAVBG → partner → client

See exactly what each layer changed against standard, and separate Microsoft's code from localization, partner and client customizations.

Dependency & reference graphswho calls · who reads · field-level

Answer the inverse question your IDE can't: "who uses this object" — down to field level, each reference with its source line and snippet. Trace a posting flow from document to ledger.

Semantic code searchlocal model · your server

Ask by meaning — "credit limit check on order release" — not by keyword. A small local model indexes every object on your own server; source code never leaves the network.

Multi-company awareness

One question unions a logical table across every company you're allowed to see — "total open balance per customer across all companies" is a single call.

Guarded live data & analyticsread-only SQL · row-capped · audited

Data questions become guarded read-only queries against the live database — and answers become charts and dashboards on request. Logic can even be cross-checked against the actual data.

Permission-scoped by designEntra SSO · deny wins · fail closed

Each user's access is derived from the real NAV/BC permission tables, subtractable down to single columns. No HR tab by design: payroll and compensation data is denied for every user, regardless of NAV/BC permissions — and every call is audited by name.

Why code + data

Code alone explains. Data alone reports. Together they answer.

"Why is document 103027 not posting?"

Reads the posting routine, checks the actual record against every condition in it, and points to the exact check that fails — with the source line.

"Give me receivables and payables per company."

Unions the live ledger across every company you can see and returns one table or chart — no export, no report developer.

"Which customers are over their credit limit — and where is it enforced?"

One answer from both sides: the list from live data, plus the exact codeunit lines where the limit is checked.

"Is this custom discount logic actually used?"

Cross-checks logic against reality: finds the code path, then counts the real transactions that ever hit it — so dead customizations can finally be retired.

"How does the code handle partial shipments? Then show me orders stuck half-shipped."

Explains the shipment logic first, then queries the live orders sitting in exactly that state — from "how it works" to "here they are" in one conversation.

Who gains

Value at every level of the team

C-Level

"Which companies have overdue receivables growing month over month?"
  • Decide with real scope numbers, not vendor estimates
  • Dashboards and charts on demand — Power BI-style visuals, built by simply asking Claude
  • Institutional knowledge survives staff turnover
  • Independence — no lock-in to one developer or partner

PMO

"Jobs where actual cost exceeds budget — sorted by overrun."
  • Estimates grounded in actual code scope, not gut feel
  • Risk register fed by real dependency data
  • Budget vs actual, utilisation and WIP straight from the job ledger — no waiting for a report

Consultants

"What breaks if we change this field?"
  • Impact analysis in minutes instead of days
  • Onboard onto an unfamiliar estate in days, not months
  • Customization audits that are actually complete

Accountants

"This balance looks wrong — explain how it's calculated, then verify it against the entries."
  • Trace any number on a report back to its source logic
  • Plain-language answers about posting behavior
  • Month-end issues resolved without waiting for a developer
  • Check the ledgers against the code — find entries that break the posting rules

Sales

"Which customers bought item A but never item B — cross-sell candidates?"
  • Churn-risk lists — customers with no sales in months
  • Quote-to-order conversion and expiring quotes on demand
  • Top customers and items without waiting for a BI request

Developers

"Who calls this function, across all layers?"
  • Ask in plain language, get answers with object references
  • Trace any field or function to its true source layer
  • AI-assisted development grounded in the real code
  • Estate housekeeping — free object IDs, dead-code candidates, modified-standard hotspots

Curated example prompts ship for 12 roles — from Helpdesk and Sales to Audit & Compliance and Onboarding.

In practice

Where it pays for itself

BC migration planning

Map every customization against standard before a C/AL-to-AL or version upgrade — which standard objects were modified, the upgrade hotspots — and know the true scope up front.

Scoping weeks become scoping days
Incident triage

A user reports "Nothing to Post." Find where that exact message is raised, what triggers it, and what the customized layer changed around it — while the ticket is still warm.

Root cause before the escalation call
Customization audit

Produce a complete, layer-by-layer inventory of what was changed, where, and what depends on it.

An audit you can hand to a client
Living documentation

Generate and refresh functional documentation from the code itself, so it never drifts out of date.

Docs that match reality
Ad-hoc reporting

"Show me revenue by company for Q2 as a chart." A Power BI-style dashboard in the time it takes to ask — no report developer, no export, no wait.

A chart in the time it takes to ask
Master data hygiene

Missing posting groups, duplicate customers by VAT number, number series about to run out — the quiet time-bombs, surfaced before they fire.

Problems found before users find them

Next step

Point it at your estate.

IntelligenceHub runs in production today on a large, multi-company NAV estate. A demo takes one hour and uses your own objects, not a canned sample.

Book a demo

Setup is one pasted config block and a Microsoft sign-in — nothing product-specific to install, nothing to leak, access follows your Entra group.