TOOL / Free self-assessment
AI governance readiness assessment for manufacturers
28 questions across seven dimensions, from ownership and use-case inventory to OT security and vendor AI. You get a 0-100 index, a maturity stage and a 30/60/90-day plan built from your own answers.
ABOUT
About this assessment
AI is already in your plants: in vision inspection on the line, predictive maintenance on a vendor's cloud, the MES vendor's new copilot, and in ChatGPT tabs on the engineering PCs. This assessment checks whether you govern it on purpose. Each question describes five levels in observable terms, so pick the one that matches what a visitor would actually see on site today, not the policy you plan to write. If you do not know, choose "Not sure": it scores as level 1 and is flagged as something to find out.
- An AI governance index from 0 to 100 and a maturity stage
- Scores for 7 dimensions against a realistic mid-market target
- Your three weakest areas with specific next actions for the answers you chose
- A 30/60/90-day plan you can print or paste into a management review
MODEL
What the assessment covers
DIM-01
Strategy & ownership
Who owns AI risk, how AI links to business goals, and whether governance has budget and leadership attention.
4 questions · target 3.5/5
DIM-02
Use-case inventory & risk
Knowing where AI runs, including inside vendor equipment and SaaS, and classifying each use case by risk before it goes live.
4 questions · target 3.5/5
DIM-03
Policy & acceptable use
A short, usable AI policy and generative AI rules that people on the shop floor and in the office actually know.
4 questions · target 3.5/5
DIM-04
Data governance
Classification, ownership, quality and traceability of the data AI systems learn from and act on.
4 questions · target 3/5
DIM-05
OT/IT security for AI
How AI components fit your zones and conduits, remote access, and safeguards before AI can influence control systems.
4 questions · target 3.5/5
DIM-06
Vendor & third-party AI
Finding AI in what you buy, checking vendor security and data terms, and staying informed when vendors change models.
4 questions · target 3/5
DIM-07
People, skills & oversight
AI literacy by role, defined human oversight, worker involvement and a way to report AI problems.
4 questions · target 3.5/5
SCALE
Maturity stages
0–20
1. Ad hoc
AI is used without deliberate governance. Risk depends on individual judgment. Focus on ownership, an inventory and interim generative AI rules.
21–40
2. Emerging
Some pieces exist, usually a draft policy or a partial list. Make them official, connect them to MOC and procurement, and close the biggest gaps.
41–60
3. Defined
Core governance is written down and mostly used. The next step is consistency across plants and evidence that controls work.
61–80
4. Managed
Governance is integrated into existing processes and measured. Focus on calibration, vendor assurance and closed-loop safeguards.
81–100
5. Optimized
Governance is routine, measured and improving. Keep it lean and watch for drift as tools, regulations and plants change.
EXHIBIT 1
Sample result
Fictional 3-plant automotive parts supplier, 600 staff
Fictional company. Illustrative scores only.
All 28 questions
Strategy & ownership
- Who is accountable for AI risk across IT, OT and the products you sell? (ISO/IEC 42001 clause 5 (Leadership))
- How are AI initiatives tied to business and operations goals?
- Is there budget and time for AI governance, not only for AI projects?
- What does leadership see about AI use and risk?
Use-case inventory & risk
- Do you have an inventory of AI use cases, including AI embedded in vendor equipment, SaaS and generative AI tools? (NIST AI RMF MAP function)
- How do you classify the risk of each AI use case? (EU AI Act Art. 6 / Annex III; NIST AI RMF)
- Is there an approval gate before an AI system goes live on a line or in a business process?
- How do you monitor AI systems after go-live? (NIST AI RMF MEASURE / MANAGE)
Policy & acceptable use
- Do you have an approved AI policy? (ISO/IEC 42001 clause 5.2)
- What are the rules for public generative AI tools (chatbots, code assistants, image generators)?
- Is AI linked into existing processes such as MOC, quality management and procurement?
- How are policy exceptions and breaches handled?
Data governance
- Is operational data (recipes, process parameters, drawings, quality records) classified?
- Is there a named owner for data used to train or feed AI systems?
- How good is the data AI systems rely on (completeness, accuracy, timestamps, context)?
- Can you trace what an AI system recommended or decided, and on which data? (EU AI Act Art. 12 (record-keeping, high-risk))
OT/IT security for AI
- Where do AI components (edge boxes, cloud connectors, inference servers) sit in your OT network? (IEC 62443-3-2 (zones and conduits))
- How is remote access for AI vendors and cloud services controlled? (IEC 62443-3-3 FR1, FR5)
- Do your risk assessments consider AI-specific threats such as data poisoning, prompt injection or model tampering? (NIST AI RMF; IEC 62443-3-2)
- Can AI change setpoints or write to control systems, and what safeguards apply? (IEC 62443-3-3 FR3 (System integrity))
Vendor & third-party AI
- Do you know which of your suppliers' products and services use AI?
- Do you run security and data checks on AI vendors before buying? (IEC 62443-2-4; IEC 62443-4-1)
- Do contracts cover AI-specific terms (use of your data for training, IP, liability, regulatory roles)?
- How do you stay informed when a vendor changes or retrains a model you rely on?
People, skills & oversight
- Do staff who use or oversee AI have role-appropriate AI literacy? (EU AI Act Art. 4 (AI literacy))
- Is human oversight defined for each AI use case? (EU AI Act Art. 14 / Art. 26)
- Are workers and their representatives involved when AI affects their work? (EU AI Act Art. 26(7) (high-risk, workplace))
- Is there a way to report and learn from AI errors or near misses?
Related guides and templates
FAQ
Frequently asked questions
How is the AI governance readiness score calculated?
Each answer is a maturity level from 1 to 5. Answers are averaged per dimension, converted to a 0-100 scale with (average - 1) / 4 x 100, and combined using the dimension weights. "Not sure" counts as level 1 and is flagged, because something nobody can confirm is not yet governed. The methodology page explains the bands and targets.
Is my data sent anywhere?
No. The assessment runs entirely in your browser. Answers are stored only in your browser's local storage so you can pause and resume, and nothing is sent to a server. Use "Start over" to clear them. Printing or copying the summary is the only way the results leave your device, and you control both.
Who should fill in the assessment?
Ideally the person who owns AI or digital topics, with input from one OT engineer and one person from quality or EHS. Doing it together in a 30-minute meeting gives a more honest answer than one person guessing. Mid-market manufacturers often find the biggest gaps in vendor AI and OT safeguards, which IT alone may not see.
What score should a mid-market manufacturer aim for?
The targets shown are around level 3 to 3.5 per dimension: governance that is defined, written down and used, without heavy bureaucracy. Plants with AI influencing safety functions, quality release or worker evaluation should aim higher in the related dimensions. Use the target markers as a starting point, not a compliance threshold.