Available Copilot reports can show adoption and usage trends, depending on products, licenses, permissions, configuration, data availability and privacy thresholds. But a high usage rate cannot tell you whether that work is reliable.
That distinction is becoming a serious management question. As organisations prepare wider Microsoft 365 Copilot deployments and increasingly capable AI agents, the rollout conversation often centres on licenses, access, governance and usage. All necessary. None sufficient.
A team can log in every day and still mishandle confidential data, misread an AI-generated chart, or accept an incorrect answer because it sounds convincing. Another team can use AI less often but apply it exceptionally well to high-value work. Neither story is visible from a usage count alone.
The operational question for HR, L&D and transformation leaders planning their next rollout wave is simple: Are we measuring tool adoption—or are we measuring whether people can use AI well enough for the work we expect them to do?
Why this matters now: the shift from copilots to agents
Microsoft's 2026 Work Trend Index, published in May, examined AI-related workplace behaviour using anonymised productivity signals and a survey of 20,000 workers using AI across ten countries, fielded by Edelman Data x Intelligence on behalf of Microsoft. Those 20,000 survey participants were not Microsoft employees as a group; these are data from Microsoft-commissioned research, not Selectic assessment results. Survey-based readiness findings are self-reported, not validated individual skill scores. The report highlights the changing division of work between people and AI agents. Source: Microsoft Work Trend Index 2026.
McKinsey's July 2026 research makes the implementation challenge tangible: almost 90% of the organisations or units described in individual leaders' accounts remained in the first two stages of AI transformation—enablement and automation—rather than reinvention. This classification is based on those individual accounts, not representative audits or a measure of market prevalence. Source: McKinsey, From adoption to impact.
The implication is not that every company needs more training. It is that training, workflow design and governance should be informed by evidence about actual workforce capability.
The measurement gap: adoption is not readiness
Microsoft already provides useful adoption analytics. Depending on products, licenses, permissions, configuration, data availability and privacy thresholds, its Copilot reporting can show enabled users, active users and use by application; Copilot Analytics also offers reporting on usage, organisational impact and agent activity. These are valuable operational signals, not substitutes for assessing a person's ability to apply judgement in a realistic task. Source: Microsoft Learn, Copilot controls measurement and reporting.
Think of a complete measurement system as three layers:
| Layer | What it answers | Practical evidence |
|---|---|---|
| Adoption | Who has access, and who is actually using AI? | License activation, active usage and feature adoption |
| Capability | Can employees use AI effectively and spot problems? | Structured skill assessment and realistic tasks |
| Experience | Do people trust, understand and feel supported by the rollout? | Usage, confidence, expectations and concerns collected through surveys |
A rollout team needs all three. More prompts submitted is not, by itself, proof of stronger skills, better decisions or business value.
This is where the decision shifts. If a function has high usage but employees cannot consistently judge the quality of AI-assisted work, the next investment is not necessarily another batch of licenses. If quality is strong but adoption remains low, the obstacle may be a lack of relevant workflows, clarity or support.
Three questions that decide whether an AI rollout is ready to scale
A usage dashboard is a starting point, not a verdict. Before extending Copilot or AI agents to more teams, a rollout owner should be able to answer three business questions.
1. Are people getting reliable work done with AI?
A license activation proves access; a usage count proves activity. Neither shows whether an AI-assisted output is accurate enough to inform a customer decision, a financial report, or an internal workflow. Agree on the kinds of tasks that matter for each role and what acceptable quality looks like. Do not use more prompts as a proxy for stronger capability.
2. Do employees know where human judgement is non-negotiable?
AI can accelerate drafting, research and analysis. It can also produce persuasive mistakes. HR, IT and business teams need clarity on review responsibilities, escalation routes and sensitive-data boundaries. The goal is not to ban useful tools; it is to know where the organisation needs safeguards and support.
3. Is the workforce prepared to adopt the change—not just technically enabled?
Two departments can have the same adoption figures but different barriers: uncertainty about permitted uses, low trust in outputs, or difficulty fitting AI into actual workflows. A rollout plan should distinguish practical capability, declared use and employee perceptions rather than treating confidence as an objective skills score.
These questions describe decisions HR should make, not Selectic's internal assessment design. Selectic uses structured AI Readiness measurement to give organisations evidence on practical capabilities and employee experience, with the assessment methodology and scoring discussed in an appropriate evaluation.
The 30-day Copilot readiness blueprint
This is a planning template, not a claim that every organisation can complete a full transformation in 30 days.
| When | Decision | What to do | Output |
|---|---|---|---|
| Days 1–5 | Where is AI being introduced? | Select one business unit, role family and priority workflows. Define acceptable use and success criteria with HR, IT and business owners. | Agreed rollout scope and role-specific risks |
| Days 6–12 | What is our baseline? | Establish a role-relevant capability baseline; collect declared usage and concerns; review available Copilot adoption data. | Baseline by relevant cohort and skill area |
| Days 13–18 | Where do we intervene? | Separate access problems from skill gaps, workflow friction and low confidence. Prioritise two or three high-impact needs. | Segmented support and learning plan |
| Days 19–25 | Does the support work? | Deliver targeted enablement; run a small practical exercise or pilot with clear quality criteria. | Learning and workflow feedback |
| Days 26–30 | Are we ready to expand? | Compare available before/after evidence, review privacy and quality issues, and decide what to scale, adjust or stop. | Management decision and next measurement point |
You do not need every employee to become an AI expert. You need to know which skills matter for each role, what support is proportionate and how to determine whether people are improving.
Selectic · Blueprint
Is your Copilot rollout ready to scale?
Get the 30-Day AI Rollout Readiness Blueprint: a practical planning sheet for HR, IT and business leads.
What about the EU AI Act?
AI literacy is also a governance concern. Article 4 of the EU AI Act requires providers and deployers to take measures supporting AI literacy, taking account of people's knowledge, experience, training and the context in which AI is used. The current provision does not require every individual to reach a prescribed proficiency score. It also does not mandate a particular assessment product or a universal pre/post test. Source: European Commission AI literacy Q&A.
For an organisation, documented and proportionate capability-building can help inform its approach. But an assessment alone does not establish compliance, and deployment obligations depend on the use case and applicable rules. Involve legal, privacy and employee representatives where appropriate.
The board question that changes the rollout
Imagine two departments with similar Copilot activation rates. One reports confidence but struggles to validate outputs. The other reports more reservations yet performs well on practical checks. This is an illustrative scenario, not a Selectic client result.
A usage-only view might treat the departments as equally ready. A capability-and-experience view leads to different interventions: output verification for the first team; communication, support and trust-building for the second.
That's what turns an AI adoption programme into a workforce decision rather than a software deployment report.
Before your next rollout wave: what to measure, what to decide
Before the next expansion wave, answer four questions:
- Which roles and workflows are most exposed to AI-related errors or low adoption?
- What would reliable, safe AI-assisted work look like in those roles?
- What do employees actually know, use and feel—not just say they are confident about?
- What evidence would justify expansion, additional training or a pause?
Request Selectic's 30-Day AI Rollout Readiness Blueprint. It gives your team a concise planning grid for the pilot, with stakeholder decisions and evidence to review before scaling. The blueprint is a decision aid, not an assessment or a promise of AI Act compliance.
Selectic · Blueprint
Is your Copilot rollout ready to scale?
Preparing the next Copilot or AI-agent rollout wave? Request the blueprint—or talk to Selectic about an evidence-based AI readiness baseline for a defined workforce cohort.
Selectic's AI Readiness Assessment helps organisations measure work-relevant AI capabilities through a structured assessment and understand declared use and perceptions. This provides a baseline HR and L&D can use to prioritise support and evaluate improvement over time.
Frequently asked questions
Is Microsoft Copilot adoption the same as AI readiness?
No. Adoption metrics describe access and usage. Readiness includes capability, judgement and employee experience. The measures complement one another.
What should HR measure before a Copilot rollout?
At minimum: the roles and workflows affected, the quality of AI-assisted work, boundaries for safe use, workforce capability, and employee experience. The detailed measurement framework depends on the use case.
How often should AI readiness be measured?
Start with a baseline before a meaningful rollout wave, reassess after targeted learning or a material workflow change, then select a cadence proportional to business risk. There is no universal interval.
Does the EU AI Act require a workforce AI literacy test?
No specific universal test or proficiency threshold is mandated by Article 4. Organisations must take proportionate measures to support AI literacy, considering the people and context involved.
Can a readiness assessment prove Copilot ROI?
Not on its own. It can measure a capability baseline and development. ROI requires a separate connection to workflow outcomes, quality, productivity or another relevant business measure.
Sources:
- Microsoft, 2026 Work Trend Index: https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-the-opportunity-for-every-organization
- McKinsey, From adoption to impact (8 July 2026): https://www.mckinsey.com/capabilities/people-and-organization/our-insights/from-adoption-to-impact-three-horizons-of-ai-transformation
- Microsoft Learn, Copilot controls measurement and reporting: https://learn.microsoft.com/en-us/microsoft-365/copilot/copilot-controls/measurement-reporting
- European Commission, AI literacy – Questions & Answers: https://digital-strategy.ec.europa.eu/en/faqs/ai-literacy-questions-answers
This article is not legal advice.