Insights / AI and Automation
InsightMicrosoft 365CopilotAI adoption

Microsoft 365 Copilot can reach into almost everything. It pays off first in fairly ordinary work.

Natalia PletnevaCommercial Director, Inlight IT

Microsoft 365 Copilot can now reach into meetings, email, documents, presentations and organisational information. The range of capability has expanded quickly, but the evidence on where it creates useful value first is becoming more specific.

The strongest early gains are appearing in preparing for and following up meetings, drafting, working through long documents, and finding or synthesising information. The useful question is less about department or seniority, and more about which parts of the working week Copilot can compress.

See the four places to start
Working week / suitable work / time returned
WORKING WEEK PREPAREDRAFTREADFIND Meetings and follow-upactions + context First draftsworth reviewing Reading and synthesisbefore the decision Organisational retrievalworks through access COPILOT BOUNDARY Accuracy-critical workgoverned systems own it
PrepareWhat happens around meetings?
DraftWhat gets to a first version?
ReadWhat has to be understood first?
FindWhat does the business already know?
The evidence

Three controlled trials, one consistent finding

Early Copilot claims often relied heavily on enthusiasm and self-reported productivity. There is now a larger body of evidence to work with, and the more conservative studies are the more useful ones.

01

What did Microsoft's own field experiment find?

A randomised field experiment involving more than 6,000 workers across 56 firms found that people using Microsoft 365 Copilot completed documents 12% faster and spent around 30 minutes less each week reading email.

Measured inDocuments and email
02

What happened in a large controlled public-sector deployment?

HM Revenue & Customs ran a broader controlled deployment involving 3,000 randomly allocated licences. Its more conservative estimate was a saving of around 60 minutes per full-time employee per week. Drafting in Word, Teams collaboration, meeting summaries and searching for documentation were among the most common uses.

Measured inWhole working week
03

How small is the commercial hurdle in practice?

The Australian Treasury trial is useful precisely because its conclusions were more restrained. Its 218 participants found the clearest benefits in basic administrative work: finding and summarising information, generating meeting minutes, knowledge management and drafting. Treasury calculated that an APS6 employee needed to redirect only around 13 minutes per week from lower-value to higher-value work to offset the licence cost used in the trial.

Measured in13 minutes a week
The boundary

Let the governed system produce the number. Use Copilot to explain it.

Governed systems
  • Produce the number
  • Native Excel formulas
  • Best for accuracy and reproducibility
  • Required for financial reporting and compliance
Copilot
  • Investigates and communicates
  • Drafts the commentary around the result
  • Best for analysing data, trends and explanations
  • Always reviewed and verified by a person
Governed system
Number
Copilot
Investigate
Explain
Communicate

Microsoft's own guidance advises against AI-generated output for financial reporting and other high-stakes scenarios.

Where it pays off

Four places worth looking before searching for something more ambitious.

These opportunities appear in finance, HR, marketing, customer support, sales and operations. In SMBs they may converge in one person; in mid-market organisations they tend to concentrate in a particular group of roles. Either way they share three characteristics: they happen repeatedly, the result can be reviewed, and relatively modest time savings accumulate.

Opportunity 01 · The work around meetings
Where it shows upUnderstanding the previous discussion, preparing, capturing decisions, identifying actions, producing follow-up communication, and retrieving the conversation again several weeks later.
Who it helpsAn operations manager turning a weekly project meeting into actions and a management update; an HR manager preparing follow-up communication; customer support turning an escalation into a concise handover; a sales or account manager preparing for and following up a customer conversation.
Why it accumulatesNo single action is transformative. Ten minutes preparing, another ten writing up actions and another ten reconstructing context later becomes a meaningful amount of time across recurring meetings.
Opportunity 02 · Getting to the first useful draft
Where it shows upThe first useful draft rather than the finished output: turning known source material into something worth reviewing.
Who it helpsFinance turning completed analysis into management commentary; HR turning approved policy material into an employee communication; marketing working from a strategy or product brief; operations turning project notes into a status update; customer support turning case history into a first response.
Why it accumulatesThe Microsoft field experiment's 12% improvement in document completion reflects exactly this kind of everyday knowledge work. The person still brings the commercial judgement, context and final approval.
Opportunity 03 · Compressing the reading load
Where it shows upEmail chains accumulate, Teams conversations become hard to reconstruct, supplier proposals arrive in different formats, policies expand, and project history is distributed across meetings and documents.
Who it helpsA CFO comparing narrative and commercial assumptions across supplier proposals; an HR manager summarising consultation material or comparing policy versions; a marketing manager distilling research; a support manager understanding a long-running escalation.
Why it accumulatesThe gain comes before the decision. Copilot helps the person reach the point where judgement can begin, and summaries remain reviewable against the source material.
Opportunity 04 · Finding what the business already knows
Where it shows upQuestions such as: what has happened with this customer over the past three months? What did we agree with this supplier? What have we already written about this issue?
What the paid licence addsMicrosoft 365 Copilot Chat is already available at no additional cost with eligible subscriptions for secure web-grounded chat. A paid licence adds work grounding and deeper access across Teams, Outlook, Word, PowerPoint and Excel, together with organisational context.
The dependencyCopilot works through existing access. If a user has historical access to information they no longer need, Copilot can make it considerably easier to discover. Check the permissions and information sources involved before relying on retrieval broadly.

The financial hurdle is smaller than it looks. Microsoft currently lists Microsoft 365 Copilot Business at approximately AU$31.40 per user per month ex GST on an annual commitment before its current promotional discount, and the enterprise add-on at AU$44.90. Set against Treasury's 13-minute threshold, the constraint is rarely the price. It is having enough suitable work to clear the hurdle consistently.

By function

Job title is a weak proxy. The shape of the working week is not.

A 100-person organisation does not necessarily have one Copilot decision. One employee may easily find an hour of suitable Copilot-assisted work each week; another may struggle to find ten minutes. Each function has useful early work, and a boundary worth naming before adoption scales.

Finance and people
01

Finance

Management commentary, meeting preparation, document comparison and summarising commercial material. The boundary: accuracy-critical calculations and financial reporting need governed numbers.

02

HR

Policy drafts, employee communications, meeting summaries and manager briefings. The boundary: employment and other consequential people decisions require much stronger controls.

Customer-facing
03

Marketing

Drafting from approved material, research synthesis, presentation preparation and internal information retrieval. The boundary: general writing alone may already be adequately served by Copilot Chat.

04

Customer support

Case summaries, escalation handovers, response drafts and finding technical context. The boundary: routing, SLA actions and repeatable workflow belong in the service platform or automation.

05

Sales and account management

Customer preparation, meeting follow-up, proposals and account context. The boundary: occasional copywriting alone may not justify the paid capability.

Operating the business
06

Operations

Meeting actions, procedures, project updates, supplier comparison and information retrieval. The boundary: rules-based operational processes usually suit deterministic automation.

07

SMB owner or manager

A combination of all of the above. The breadth can create value — an owner may save time across customer correspondence, supplier analysis, meetings, HR communications and marketing — but the outputs still need competent review.

Choosing the right tool

Some Copilot opportunities belong somewhere else.

Another useful part of choosing where to start is recognising the shape of the problem. If the work is drafting, summarising, preparing, retrieving or synthesising, Microsoft 365 Copilot is a natural candidate. If the process is "when this happens, always do that", Power Automate is generally the stronger starting point. If an AI capability needs to pursue a multi-step goal across systems, that begins to move towards Copilot Studio and agents. If the work centres on structured data capture or transactions, a business application, Power Apps or an integration may be more appropriate.

Choosing the right category can create more value than improving the prompt used with the wrong tool.

Four problem shapes, four tools
01Microsoft 365 Copilot
02Power Automate
03Copilot Studio agents
04Business applications
The Inlight IT view

The early business case for Microsoft 365 Copilot is more practical than transformational.

The strongest evidence is around work businesses already do every day: preparing, reading, drafting, following up and finding information. Those activities can look mundane individually, but they absorb a large amount of knowledge-worker time. SMBs may have a particular advantage, because broader roles give one person several opportunities to use the capability throughout the day. Mid-market organisations can achieve the same effect by identifying the roles where those activities are concentrated. The next layer is more specialised: organisational retrieval brings information governance into the equation, deterministic workflow belongs in automation, multi-step autonomous work moves towards agents, and accuracy-critical decisions still depend on governed systems and human accountability.

The most useful first Copilot deployment is likely to be found in the working week you already have.

01

Start with the recurring work

02

Match the tool to the problem shape

03

Check access before retrieval scales

04

Keep governed systems accountable for the number

Delivery

Where this is delivered.

Copilot adoption is delivered through the same practical work as any other capability: identifying the use cases the working week actually contains, checking what the information environment will support, and putting deterministic work where it belongs.

AI Applications

Identify approved use cases, data requirements and controls before adoption scales.

Best whenTeams can see Copilot could help, but the use case, data boundary and approval model are not yet clear.
Explore AI Applications
Automation

Put deterministic, rules-based process where it belongs rather than asking Copilot to carry it.

Best whenThe work is "when this happens, always do that" — a repeatable process rather than knowledge work.
Explore Automation
Microsoft Power Platform

Structured data capture, transactions and applications built on the Microsoft platform already in place.

Best whenThe work centres on capturing structured data or running transactions, not on drafting and synthesis.
Explore Power Platform
AI Applications

Find the Copilot use cases your working week already contains.

Identify practical AI use cases, data requirements and controls before adoption scales.

Explore AI Applications