Copilot at Melbourne Water.

A Microsoft 365 Copilot proof of concept: metrics defined before licences, with personas, use cases and hands-on workshops.

Role: AI Lead UX Designer (discovery, personas & workshops), via Exco Partners
Duration: Sep – Oct 2025 · 12-week PoC design
Sector: Water utility, enterprise AI

The Objectives

Three Questions the PoC Had to Answer.

Before anyone talked licences, the proof of concept was framed as three answerable questions, so "did Copilot work?" would have an evidence-backed answer.

Speed & quality

Does Copilot help us deliver outcomes faster, better, to a higher quality?

Reliability & friction reduction

Does Copilot help us find information faster and more reliably, and reduce frustration?

Satisfaction & improvement

Does Copilot improve our job satisfaction?

Journey to Copilot success (enhancing AI understanding and usage): Copilot M365 objectives for Melbourne Water — The objectives, as presented to Melbourne Water stakeholders.

The Approach · 12 Weeks, 3 Phases

Discover, Synthesise and Test, With ROI and Business Cases Built In.

Eight objectives spanned the engagement, from stakeholder interviews and metric baselines through to documented ROI, findings to inform future deployments, and a business case for organisation-wide rollout.

Phase 1 · weeks 1–4

Discover + assess

Stakeholder interviews, validation of existing artefacts, and assessment of current patterns and gaps. Output: project charter and discovery findings report.

Phase 2 · weeks 5–8

Synthesise + create

The CX digital framework across people, process and technology, drafted as a blueprint and sense-checked in a mini showcase with key stakeholders.

Phase 3 · weeks 9–12

Test + refine

Stakeholder showcases, framework iteration, and the finalised, signed-off CX digital framework and components.

Objectives 5 to 8: assess metrics, document ROI, inform future deployments and build business cases, alongside the three-phase approach — Objectives five to eight: metrics, ROI, learnings and the business case for further deployment.

Discovery · Deliverables Defined by Metrics

Every Scenario Got a Success Metric Before It Got a Licence.

Four working scenarios (meetings, content creation, data analytics and proposal development), each mapped to roles, applications, "success is when" criteria and a Microsoft ROI estimate. Personas, customised use cases and hands-on workshops closed the gap between licence and behaviour.

1,500+

Staff in scope for the meetings and content-creation scenarios.

−32 min

Per meeting: the ROI target for documenting minutes, actions and agendas.

Scenarios with defined roles, applications and "success is when" criteria.

Discovery, with deliverables defined by metrics: four Copilot scenarios mapped to roles, user numbers, applications, success criteria and ROI — The scenario–metric matrix, with personas, use cases and workshops flagged as the deliverables that make it real.

Enablement · Journey to Copilot Success

Teaching the Prompt as Well as the Product.

Seventy-odd staff went through the training: eight sessions on responsible AI and prompting, then one-on-ones. Technical literacy across that group went up 80%. The personas were behavioural, built around how someone's week ran and where generative AI could sit inside it, because adoption is a behaviour problem first. The wider rollout had 1,500+ staff in scope with a documentation target of 32 minutes saved per meeting.

+80%

Technical literacy, trained cohort.

70+

Staff trained across eight sessions.

Journey to Copilot success: how do I write a good prompt? Training material shown against the Copilot chat interface — Prompt craft, taught inside the tool people already had open.

An AI proof of concept earns its rollout when the metrics are set first, not counted afterwards.