---
title: "AI Rating in manufacturing: high automation, unusable data"
url: https://zerofive.ai/en/blog/strategy/ai-rating-manufacturing-data-and-product
canonical: https://zerofive.ai/en/blog/strategy/ai-rating-manufacturing-data-and-product
language: en
published: 2026-08-26
updated: 2026-09-23
author: "ZeroFive.AI"
tags: AI Rating, manufacturing, industry, data quality, Annex I
abstract: "Sensors everywhere and unusable data. The plant-product separation gate and why 2028 already shapes today's designs."
---

# AI Rating in manufacturing: high automation, unusable data

**In short.** In manufacturing the AI maturity profile carries a weakness that surprises technologically advanced companies: low Readiness through data quality, not infrastructure. A plant full of sensors can hold unusable data, because it was collected for process control rather than analysis. The recurring gate, though, is another one, and concerns the missing separation between plant systems and product-bound systems.

Manufacturers come to the rating expecting a high score on the technical dimension, because they have invested in automation for years. The result surprises them, because the model measures the governability of AI systems rather than the level of automation.

## The four dimensions in manufacturing

**Readiness** measures preparation, data, skills, infrastructure. It is the dimension with the widest gap between perception and reality. The data exists in enormous quantities, and is often unusable for AI because it was collected with process control logic, at inadequate granularity and retention.

**Delivery** measures the ability to reach production and maintain. In a factory it is limited by the fact that systems arrive inside machines, and the software lifecycle follows the machine's, which runs in years.

**Risk** measures governance, compliance and ethics. It is exposed on two separate fronts: systems that measure people, almost never classified, and product-bound systems, where the distant deadline has produced deferral.

**Confidence** measures management commitment and user trust. In manufacturing it is often high in management and low on the shop floor, and that gap matters, because a system operators don't trust gets worked around.

## The gate: plant and product mixed together

The recurring constraint is the absence of a formal separation between systems that stay inside the company and those that end up in products sold.

While that separation is missing, the inventory can't determine the role, and without the role the obligations can't be assigned. A company can have orderly governance over plant systems and no awareness that on one product line it is already a provider under the regulation.

The finding isn't about missing controls. It's about the impossibility of knowing which controls are needed, which is a different and prior problem.

## When the data seems to be there and isn't enough

In manufacturing assessments the Readiness dimension falls on one specific point. A company states it holds ten years of maintenance history, and verification reveals that failure causes are free-text fields filled in differently from shift to shift, or that process data is retained in aggregate form because the control system never historised it.

The problem concerns structure more than volume, and technology doesn't solve it. It has to be tackled first, because it conditions every subsequent project.

## What a class means in practice

| Class | Typical situation in manufacturing | What's reasonable to do |
|---|---|---|
| D | AI systems inside machines, no mapping, unstructured data | Map the systems and verify the quality of historical data |
| C | Inventory in place, no separation of product and plant | Separate the two paths and assign the roles |
| B | Separation done, plant systems governed | Start the conformity path for products embedding AI |
| A | Mature system verified across all dimensions | Reserved for verified assessments, not self-assessments |

The jump from C to B is the one manufacturers defer most often, because the Annex I deadline is distant. It's also the one shaping products already in design today.

## Why measure before designing the next product

The useful moment for an assessment in manufacturing isn't the start of the financial year, it's the start of the development cycle for a new product line.

An AI system embedded in a product subject to harmonisation legislation requires technical documentation, risk management and conformity assessment, and all of that is designed alongside the product. Adding it afterwards means reopening architectural choices already made, at a cost nobody planned for.

A company starting development today on a product aimed at 2028 is already deciding, without knowing it, what conformity will cost.

## Where to start

The prerequisite is the [AI system inventory](/en/blog/compliance/ai-system-inventory-iso-42001), built separating plant systems from product-bound ones from the very first entry.

The full model, with the four dimensions and the gate logic, is on the [AI Rating](/en/services/ai-rating) page. The sector's regulatory picture is on the [AI governance for manufacturing](/en/sectors/manufacturing) page.

For an assessment of your company's position you can [book a meeting](https://calendly.com/fabiolalli/zerofive) or [start the self-assessment](/en/start-ai-rating).