UOB Engineers Apply Microsoft 365 Copilot to FMEA, DMAIC and SPC

On 11 September 2026, 24 UOB professionals from operations, maintenance and support gathered at IES in Jurong East for GenAI for Professional Engineers. Led by Dr Alfred, the class focused on a practical question: how can Microsoft 365 Copilot support established engineering methods without replacing professional judgment?

The day connected generative AI with three disciplines familiar to process and reliability teams: Failure Mode and Effects Analysis (FMEA), the DMAIC improvement cycle, and Statistical Process Control (SPC). Rather than treating Copilot as a general-purpose chatbot, participants explored how it can help structure information, surface questions, summarise evidence and accelerate first drafts that engineers then verify.

Why Microsoft 365 Copilot matters to operations, maintenance and support

These three functions see different parts of the same operating system. Operations staff understand the normal workflow and the points where variation becomes visible. Maintenance teams hold the history of equipment failures, inspections and corrective work. Support staff see recurring incidents, user symptoms and service patterns.

When their knowledge is captured in separate reports and conversations, improvement work starts slowly. Microsoft 365 Copilot can help teams turn approved documents, meeting notes and structured data into a shared first draft. The value is not that AI knows the process better than the people running it. The value is that it can reduce the mechanical work between evidence and discussion, giving the team more time to challenge assumptions and decide what action is justified.

Applying Microsoft 365 Copilot to FMEA

FMEA asks a team to identify how a process or asset might fail, examine the effects and causes, review existing controls, and prioritise appropriate action. It is collaborative work grounded in domain expertise. Copilot can assist with the preparation and documentation around that work.

Starting from approved process descriptions, maintenance histories or incident summaries, a team can ask Copilot to:

• Draft a structured list of process steps and candidate failure modes for expert review.
• Group recurring symptoms, causes and existing controls found across source documents.
• Identify missing evidence and generate questions for the operations or maintenance specialist.
• Convert workshop notes into a consistent FMEA table with owners and follow-up actions.
• Summarise what changed between two versions of an analysis.

The important boundary is clear: Copilot can propose and organise; the engineering team must validate. Severity, occurrence, detection ratings and risk decisions require evidence, context and accountable human sign-off. Teams that want a complete method for identifying failure modes, analysing causes and building control plans can continue with WSQ Continuous Process Improvement with FMEA.

Using Copilot across the DMAIC improvement cycle

DMAIC provides the improvement backbone: Define, Measure, Analyse, Improve and Control. The class examined how Copilot can support each phase while preserving the evidence trail.

This approach makes Copilot a consistent assistant around the DMAIC method, not a substitute for it. Readers who want to deepen the full improvement cycle can explore the Six Sigma course pathway, including training at foundational and practitioner levels.

Applying Microsoft 365 Copilot to SPC

SPC distinguishes ordinary process variation from signals that may require investigation. Copilot can make the workflow easier to communicate: explain a chart in plain language, draft a data-collection checklist, document an out-of-control review, or turn a meeting into a follow-up plan.

It should not decide that a process is stable simply because a chart looks familiar. The team still needs valid data, correct subgrouping, appropriate chart selection and verified control limits. It must also distinguish control limits, which describe process behaviour, from specification limits, which describe customer or engineering requirements.

For maintenance and support teams, this distinction is especially useful. A single alarming event may be a special cause. A slow rise in repeat incidents may reflect a systematic shift. Copilot can help assemble the narrative and supporting evidence, while qualified people determine the correct response. The CASL Statistical Process Control (SPC) in Manufacturing course develops the underlying skills in variation, control charts, capability indices and out-of-control follow-up.

A practical prompt pattern for engineering work

The most useful prompts in this context do not ask for a final answer. They establish role, evidence, output structure and a verification gate. A reusable pattern is:

1. State the task: identify the process-improvement decision or document required.
2. Bound the evidence: tell Copilot which approved files, tables or meeting notes it may use.
3. Define the structure: request the columns, headings, assumptions and evidence references needed.
4. Expose uncertainty: require missing information, conflicts and low-confidence items to be listed explicitly.
5. Keep a human gate: label the result as a draft for engineering review and name the required reviewer.

For example: Using only the approved maintenance summary and process map, draft candidate failure modes by process step. For each item, show the source evidence, missing information and questions for the maintenance lead. Do not assign risk ratings. Output a review table.

That last instruction matters. It prevents a polished AI response from looking more authoritative than the evidence supports.

Responsible use in a regulated enterprise

Banking and engineering teams work with sensitive operational information. Any use of generative AI must follow the organisation’s approved tools, access controls, data-classification rules and review procedures.

• Use only approved enterprise accounts and authorised data sources.
• Do not place confidential, personal or security-sensitive information into an unapproved AI service.
• Minimise or redact data that is not necessary for the task.
• Ask Copilot to cite the supplied evidence and expose assumptions.
• Validate calculations, ratings, control limits and recommendations independently.
• Keep accountable human approval for engineering and operational decisions.

The strongest outcome is not an AI-generated document. It is a faster, more transparent review process in which experts can see the evidence, challenge the draft and own the final decision.

Continue building the capability

The UOB class at IES showed how generative AI becomes useful when it is anchored to real professional methods. Teams can begin with the productivity layer through WSQ Enhance Work Productivity with Microsoft 365 Copilot, then deepen the engineering disciplines through the FMEA, Six Sigma and SPC learning pathways.

For organisations, these subjects can also be combined into role-relevant corporate training for operations, maintenance, engineering and support teams.

Frequently asked questions