Skip to content
Case 02·Manufacturing·2026

3D Coupe already quotes 4 times faster. To plan its production with AI, it is putting its ERP in order

3D Coupe’s ERP sorts orders by entry date. The shop plans by free machine, material received, thickness, colour and urgency. For AI to close that gap, the ERP first had to tell the truth.

3D Coupe already quotes 4 times faster. To plan its production with AI, it is putting its ERP in order
Key points
  • 2,902 ERP records extracted and analyzed
  • 496 material and tool records audited one by one
  • 410 defects documented, two guides delivered
  • 4 planning streams scoped with the shop

Where it started

After phase 1, where estimating a part went from 58 to 15 minutes, 3D Coupe came back for a broader mandate: planning production with AI. Four CNC machines, three identical and one shorter. Parts that only run on certain machines. Material sometimes not yet received. Orders that block others.

The rules for sorting all that out lived in two people’s heads. And two findings changed the order of the work: the ERP had been receiving data for years without a shared entry rule, and the company’s email, at a web host, could not serve as a base for AI tools.

Case study · 3D Coupe, phase 2
First the ERP and the email. Then, planning with AI.
01

Reference data audit and naming guides

496 records audited, 410 defects documented
Delivered
02

Email migration to Microsoft 365

Mail has been landing in Microsoft 365 since October 16
Delivered
03

Renaming the records in the ERP

In a single day, after a confirmed backup
Ready
04

Connecting AI to Microsoft 365

Dedicated account, excluded mailboxes, notice to the team
After the PIA
05

Production planning

Four parts scoped with the shop, to be built on the ERP’s API
To build
Status as of October 26, 2026
Where each workstream stands, as of October 26, 2026.
Case study · 3D Coupe, phase 2
The gap planning has to close.
What the ERP knows

It sorts orders by entry date.

That’s it.
What the shop knows
Which machine is free, out of four CNCs, three of them identical
Which parts only run on certain machines
Whether the material has arrived or not
Thickness and colour, to group the setups
Urgency, and the orders that block other orders
Rules that lived in two heads, now on paper
What the ERP knows, what the shop knows: the gap planning has to close.

The challenge

Prepare the ground for AI-assisted planning without rushing the tool: reliable records in the ERP, secure email AI can connect to, and rules written with the people who plan today.

What we built

We started with the data. The 2,902 records of the ERP reference data were extracted, and the 496 material and tool records went through one by one. Result: 410 documented defects, such as the same material in several copies, thicknesses written three ways or tools with no machine attached. The team received two naming guides, one for materials and one for tools, so entry stays clean after we leave.

In parallel, email left its web host for Microsoft 365. Before copying anything, each mailbox was inventoried folder by folder, which caught working folders stored in the wrong place. Two-factor sign-in for everyone, computers attached to the company account, phishing protection. Even the ERP, which sends orders and invoices by email, was reconnected to Microsoft 365.

Before connecting AI to the Microsoft 365 account, a privacy impact assessment was written, as Quebec’s Law 25 requires. An account reserved for AI, with no admin rights, and some mailboxes excluded from the start. The connection will come after signature and notice to the team, not before.

Planning itself was scoped with the shop: each machine’s queue, shipment tracking, picking preparation and checking programs before they go to the machine. A meeting with the ERP vendor also confirmed what its API can and cannot do. Better to know now.

The order of the work is deliberate. Planning built on reference data where the same material exists in three copies produces a wrong plan, with confidence. We put the data in order first, then connect AI.

How it unfolded

1

The data

2,902 records extracted, 496 audited, 410 defects documented, two guides delivered.

2

Email

Moved to Microsoft 365, secured, the ERP reconnected to it.

3

Compliance

Law 25 assessment before connecting AI to company mailboxes.

4

Planning

Scoped with the shop, to be built on the ERP’s API once records are fixed.

The heart of a shop is its ERP, and the heart of the ERP is its data. A base put back in order stays with the company for everything that comes next.
Raphaël Dionne, founder of Avenor Innovations
Raphaël Dionne
Avenor Innovations

The results

The ERP is known down to the last defect: 496 records audited, 410 defects documented, two naming guides in hand. The renaming is ready and will happen in a single day, after a full backup.

The company’s email has been on Microsoft 365 since October 16, 2026, secured, with the ERP reconnected to it. That project is complete.

The priority rules that lived in two heads are now written down. Planning will be built on them, not on an assumption.

Each delivered step is already useful without waiting for the next: clean records improve quotes and material planning, and secure email protects the company, whether AI comes or not.

“The result was good enough that we’re going ahead with Avenor on a much broader mandate: production planning, offcuts, ERP integration.”

Martin Picard, co-owner and operations director, 3D Coupe (translated from French)

What the company gets out of it

Data you can rely on

The defects are known, and the guides keep them from coming back. That is the condition for everything that follows.

Reference data audited and documented

Email ready for AI

Microsoft 365, two-factor sign-in, attached computers: AI can connect without opening a door.

Migration complete

Law 25 handled before, not after

The assessment is done before any data leaves for an AI service.

Compliance by design

Shop rules on paper

What lived in two heads is written down. The day someone is away, planning continues.

Four streams scoped

The thinking behind this mandate is laid out here: ERP data: the cleanup to do before an AI project in a plant.

410

defects found in 496 ERP records, before writing a single line of planning

2,902
ERP records extracted and analyzed
496
material and tool records audited one by one
410
defects documented, two guides delivered
4
planning streams scoped with the shop