Key takeaways
- Days 1 to 30: find what you have. Name owners, build the register, contain GenAI data leakage.
- Days 31 to 60: decide the rules. Policy, risk tiers, governance group, triage of every registered use case.
- Days 61 to 90: wire it in. MOC, purchasing, vendor questionnaire, training, first board report.
- Total effort is roughly 25 to 40 person-days, mostly the governance lead, with a few hours each from ops, OT, quality, EHS, HR and legal.
- Do not wait for perfect. A 70 percent register wired into MOC beats a perfect one nobody updates.
Most mid-market manufacturers do not need an AI governance program. They need a quarter of focused work that leaves behind a few simple, working controls, and then the discipline to keep them running. This page lays out that quarter week by week, with deliverables, effort and who is involved.
It assumes a lean team: an IT manager with a small team, one or two controls engineers who also look after OT security, a quality manager, an EHS manager, HR, and legal support that is part-time or external. It assumes one to three sites. Larger groups can run the plan at a pilot site and then roll it out.
Does this apply to you?
Use this plan if most of these are true:
- You have no AI policy, or only a GenAI paragraph in the IT acceptable use policy.
- Nobody can say how many AI systems are in use across the business.
- Vendors and OEMs are adding AI features to software and machines you already run.
- The board, a customer or an auditor has started asking about AI.
- You are in the EU and want a defensible position on the AI Act's AI literacy duty and on which, if any, high-risk obligations could apply to you.
As of November 2026, after the Digital Omnibus on AI (Regulation (EU) 2026/1744), Annex III high-risk obligations apply from 2 December 2027 and those for AI in products under Annex I legislation from 2 August 2028. That leaves time, but not much for anyone who discovers in 2027 that they deploy a high-risk system without knowing it.
The plan at a glance
Days 1-30 FIND owners, register, GenAI containment
Days 31-60 DECIDE policy, risk tiers, governance group
Days 61-90 WIRE IN MOC, purchasing, vendors, training,
first board report
Days 1 to 30: find what you have
| Week | Milestones | Deliverables | Who |
|---|---|---|---|
| 1 | Executive sponsor named. Governance lead appointed with time allocated. Kick-off note to managers. | One-page mandate | MD/COO, governance lead |
| 1 | Interim GenAI rule published: approved tools (if any) and "never goes in" list | One-page interim rule | Governance lead, IT |
| 2 | Interviews: maintenance, quality, planning, process engineering, IT, HR, purchasing (30 min each) | Interview notes | Governance lead plus function heads |
| 2 | Technical scan: GenAI domains in proxy logs, Copilot and AI licenses, SaaS list, OT devices with vendor cloud links | Scan results | IT, OT engineer |
| 3 | Vendor release notes and contracts searched for AI features in MES, CMMS, QMS, ERP, vision, OEM machines | List of vendor AI features | IT, engineering |
| 3 | First register populated, every row with a proposed owner | Draft register | Governance lead |
| 4 | Owners confirmed. Gaps flagged. Quick readiness self-check. | Register v1, readiness baseline | Governance lead, owners |
Effort for days 1 to 30: about 8 to 12 person-days, most of it the governance lead. Function heads give 1 to 2 hours each.
Use the AI use-case register template and see AI use-case inventory for where AI hides.
Plant-floor reality: The interview with maintenance usually finds the most. Ask what the OEM service engineer changed on the last visit and whether any machine "learns" or "adapts". Then ask the OT engineer which boxes in which cabinets talk to the outside world.
Days 31 to 60: decide the rules
| Week | Milestones | Deliverables | Who |
|---|---|---|---|
| 5 | Governance group formed (6 to 8 members), charter agreed, monthly slot booked | Charter | Sponsor, governance lead, members |
| 5 | Data classification agreed: what counts as restricted (recipes, parameters, CAD, PLC code, network info) | Data class table | Data owners, IT, engineering |
| 6 | AI policy drafted from template, including scope, safety exclusions, risk tiers, approvals | Policy draft v0.9 | Governance lead, IT, OT |
| 6 | Risk tiering scheme agreed | Triage worksheet adapted | Governance group |
| 7 | Triage workshop: every register entry tiered in one or two sessions | Tiered register | Governance group plus owners |
| 7 | High-tier items flagged for full assessment; EU AI Act scoping questions listed | High-tier list, questions for counsel | Governance lead, legal |
| 8 | Policy reviewed by ops, quality, EHS, HR, legal; works council briefed where applicable | Policy v1.0 for sign-off | All functions |
| 8 | GenAI shop-floor rules finalized, including PLC code review rule | GenAI standard | IT, controls lead |
Effort for days 31 to 60: about 10 to 15 person-days in total. The triage workshop is the biggest shared time cost: 6 to 8 people for two 2-hour sessions.
Guidance: AI policy for manufacturers, AI risk assessment, roles and RACI.
Watch out: In Germany and other EU countries with strong employee co-determination, plan the works council conversation into weeks 6 to 8, not after sign-off. If any registered use case monitors or rates workers, it may need an agreement before it can continue. That can take longer than 90 days, so start early and run in parallel.
Days 61 to 90: wire it in
| Week | Milestones | Deliverables | Who |
|---|---|---|---|
| 9 | Policy signed by MD/executive team and published | Policy v1.0 | Sponsor |
| 9 | MOC form updated with AI question; MOC reviewers briefed | Updated MOC form | Engineering, EHS |
| 10 | Purchase requisition and RFQ updated with AI tick box; buyers briefed | Updated purchasing forms | Purchasing |
| 10 | Vendor questionnaire sent to top 3 to 5 AI-enabled vendors | Vendor responses tracked | IT, OT, purchasing |
| 11 | Training: 20 min for operators and technicians in shift meetings; 45 min for engineers and managers; one session for the executive team | Training records | HR, governance lead |
| 11 | FAT/SAT checklists updated with AI features, data flows and remote access | Updated checklists | Engineering |
| 12 | First full assessment completed for the highest-risk use case | Assessment record | Owner, IT, OT, quality |
| 12 | Gaps in OT security around AI devices logged (zones, remote access) | Findings list | OT engineer |
| 13 | First quarterly board or owner report: register by tier, value, risks, plan for next quarter | One-page report | Sponsor, governance lead |
| 13 | Retrospective and plan for the next 6 months | Roadmap | Governance group |
Effort for days 61 to 90: about 8 to 13 person-days, plus training time for staff.
Board reporting: see board questions on AI and OT risk and the board briefing toolkit.
Total effort and budget
| Role | Approximate time over 90 days |
|---|---|
| Governance lead | 18 to 28 days (20 to 40 percent of a role) |
| Executive sponsor | 1 to 2 days |
| IT manager | 3 to 5 days |
| OT / controls engineer | 3 to 5 days |
| Quality, EHS, HR, purchasing | 1 to 2 days each |
| Legal (internal or external) | 1 to 3 days |
| Use-case owners | 2 to 4 hours each |
Roughly 25 to 40 person-days in total. External spend is optional: a legal review of the policy and AI Act scoping, and an OT security review of AI devices and vendor access, are the two items most worth paying for if you lack in-house expertise.
What good looks like on day 90
- A register with every known AI use case, an owner and a tier.
- A signed AI policy, a one-page operator version and GenAI rules.
- A governance group that has met at least twice and taken real decisions.
- MOC and purchasing forms that catch new AI.
- Top vendors questioned about data use, model changes and remote access.
- Staff trained, with records.
- A board or owner report that management can repeat every quarter.
After day 90
The next six months usually cover: full assessments for remaining high-tier use cases; a deeper look at OT security around AI devices; data readiness for the next wave of use cases; contract updates at vendor renewal; and, if customers or strategy require it, a gap assessment against ISO/IEC 42001.
Common mistakes
- Starting with the policy. Without a register, the policy is written in the dark.
- Spending weeks on tooling. A spreadsheet is fine for the first year.
- Governance group too big or too junior.
- Skipping OT. The plan then covers office AI and misses the line.
- No time allocation. "Do it on top of your job" fails by week 4.
Monday-morning checklist
- Name the sponsor and the governance lead, and agree their time in writing.
- Publish the interim GenAI rule this week.
- Book the seven interviews for week 2.
- Ask IT for 90 days of GenAI domain traffic and a list of AI licenses.
- Put a date in the diary for the triage workshop in week 7 and the board report in week 13.
This page is general guidance, not legal advice. Confirm regulatory obligations with your counsel.
Sources and further reading
- NIST AI Risk Management Framework
- ISO/IEC 42001:2023 Artificial intelligence management system (ISO)
- Regulation (EU) 2024/1689 (Artificial Intelligence Act), EUR-Lex
- Regulation (EU) 2026/1744 (Digital Omnibus on AI), EUR-Lex
This guide is general information for manufacturers, not legal advice. Confirm regulatory obligations with qualified counsel.
FAQ
Frequently asked questions
How long does it take to set up AI governance in a manufacturing company?
A working baseline (register, policy, risk tiers, governance group, purchasing and MOC hooks, first board report) is realistic in about 90 days for a single-site or small multi-site manufacturer. Maturing it, for example toward ISO/IEC 42001 certification or full high-risk compliance where relevant, typically takes a further 6 to 18 months.
Who should lead an AI governance program in a mid-market manufacturer?
A named governance lead, often the IT manager, an operations excellence or CI lead, or a digital manufacturing lead, with 20 to 40 percent of their time for the first 90 days. An executive sponsor, usually the COO or managing director, should own the outcome and chair the governance group.
Do we need consultants to build AI governance?
Not necessarily. Templates and a structured plan get most mid-market companies to a baseline. Outside help is most useful for a short legal review of the policy and EU AI Act scoping, an OT security review of AI devices and vendor connections, and facilitating the first risk triage workshop if the team has not done one before.
What is the first thing to do in an AI governance program?
Name an owner and start a register. Ask maintenance, quality, planning, process engineering, IT, HR and purchasing what they use that predicts, classifies, recommends or optimizes. In parallel, publish a short rule on what data must never go into public GenAI tools. Those two steps reduce the most risk for the least effort.
Related guides
AI Governance
How to build an AI use-case inventory in a manufacturing plant
An AI use-case inventory is a register of every AI system you build, buy or have embedded in equipment, with an owner, a risk tier and enough detail to act on.
Guide · 7 min read
AI Governance
How to write an AI policy for a manufacturing company
A manufacturing AI policy has to cover more than office chatbots.
Guide · 8 min read
AI Governance
AI risk assessment for manufacturers: a practical tiering scheme
Tier each industrial AI use case by what it can do to the process, people and product: does it advise, act in closed loop or touch safety; does it use worker data; does it release product…
Guide · 7 min read
AI Governance
AI governance roles and RACI for mid-market manufacturers
Mid-market manufacturers need a small AI governance group drawn from roles they already have, a clear RACI across the AI lifecycle, and a firm split between what plant managers own (use…
Guide · 7 min read