What leaders and teams rolling out Microsoft Copilot report, with every finding tied to its evidence.
Read the findingsPractitioner sentiment, leaders and teams
−3 Mixed from 40 sources that take a side
How this figure is madeFrom 38 authors in 32 discussions. Run a pilot on your team's recurring tasks to measure how often answers are used without rework.
Jan 2025 to Oct 2026.
108 sources from practitioners, by month and sentiment.
IT leaders, departments, and organization-wide evaluations split close to evenly. Every one of their sources that discusses admin controls or oversharing raises it as a problem, and license cost follows. Copilot staying inside existing Microsoft 365 data terms and its meeting recaps draw the most praise, 5 positive accounts each. Of the 11 that say how licenses were assigned, 8 gave them to selected roles first.
Positive, mixed and negative sources
The score counts them as 12 positive, 13 mixed and 13 negative votes, one per author. See How the numbers work.
See the evidenceBelow: 3 praise themes, 4 concerns, 3 close to agreement and 1 disagreement.
Counted for Leaders and teams rolling it out, from 57 sources. Change group
Every capability practitioners describe positively, ranked by how many sources say so. Open one to read who said it.
Counted for Leaders and teams rolling it out.
Mixed
Quothling, Hacker News commenter
Comment on: The last six months in LLMs in five minutesHacker News comment
Teams transcribes every meeting very well and Cowork drafts most presentations, but the regular Office 365 Copilot was poor enough that many colleagues stopped believing in AI.
Positive
_Tomin_, Reddit commenter
Comment on: Anyone using Copilot for M365? What gains does it bring you? : sysadminReddit comment
Rolled Copilot out to some salespeople to transcribe meetings so they can focus on clients, and uses it to start policy and procedure documents.
Positive
HM Revenue & Customs
Evaluation report: phase 3 trial of Microsoft CopilotGovernment evaluation report, GOV.UK
In a self-reported survey of 1,364 staff from a randomly allocated cohort, HMRC users estimated saving about an hour a week after HMRC cut the figure by about 20%. 83% of licensed staff used Copilot, and security and privacy worries were the top reason non-users gave.
Positive
UK Department of Health and Social Care and NHS England
Major NHS AI trial delivers unprecedented time and cost savingsGovernment press release, GOV.UK
A Microsoft 365 Copilot pilot across 90 NHS organizations reported 43 minutes saved per person per day; Copilot Chat is now available across the NHS.
Mixed
Muneera Bano, Principal Research Scientist and Science Team Lead, CSIRO, with six co-authors
A Qualitative Study of User Perception of M365 AI CopilotResearch paper, arXiv
CSIRO's trial users found Copilot useful for email coaching, meeting summaries, and finding content. Usability limits and the need for people to check its output held back its wider impact.
Counted for Leaders and teams rolling it out.
Positive, 5
Positive
Antony Spence, Assistant Secretary for IT support, engagement and innovation, Australian Attorney-General's Department
Attorney-General's Dept backs Copilot Chat, Google NotebookLMTrade press report, iTnews
The department runs Copilot Chat in its own controlled tenancy at the protected level, so staff can use AI on their own data. This followed a public-data-only phase that fell out of favor, and humans still own every final version.
Positive
a_bonobo, Hacker News commenter
Comment on: Microsoft 365 Copilot's commercial failureHacker News comment
At a non-tech employer's slow rollout, non-technical participants love Copilot and IT values its defined data and risk policies; its agents handle only basic document questions.
Positive
Forumrider4life, Reddit commenter in IT administration
Comment on: Anyone using Copilot for M365? What gains does it bring you? : sysadminReddit comment
Prefers Copilot to alternatives for compliance: the organization can track usage in Microsoft Purview and see whether company information or PII is used.
Mixed
Different_Coffee_161, IT administrator
Claude M365 Connector vs Copilot: are we creating long-term technical debt? : sysadminReddit post
An admin weighing Claude's Microsoft 365 connector says Copilot disappointed at first but now fits accounting, HR, and general staff better through agents, Purview, DLP, and label support.
Mixed
3acctforcom, Hacker News commenter who manages reporting, analytics, and coding teams
Comment on: Claude Code is suddenly everywhere inside MicrosoftHacker News comment
A manager waiting on legal sign-off to buy Copilot for Office, Power BI, and GitHub says staff meanwhile paste work into ChatGPT.
Mixed or negative, 2
Positive
HM Revenue & Customs
Evaluation report: phase 3 trial of Microsoft CopilotGovernment evaluation report, GOV.UK
In a self-reported survey of 1,364 staff from a randomly allocated cohort, HMRC users estimated saving about an hour a week after HMRC cut the figure by about 20%. 83% of licensed staff used Copilot, and security and privacy worries were the top reason non-users gave.
Negative
wkat4242, Hacker News commenter who manages Microsoft 365 at a large company
Comment on: Microsoft 365 Copilot's commercial failureHacker News comment
An M365 admin lists SharePoint agents switched on by default, constant renames, and Copilot Chat adding history, which forced a full DPIA redo.
Counted for Leaders and teams rolling it out.
Positive, 4
Mixed
DramaticErraticism, Microsoft 365 admin at a large company
Great news, CoWork for CoPilot has left beta...bad news, they are charging for it!!! : sysadminReddit post
An admin at a large E5 company says users like Copilot Cowork. In the thread, metered credits put a mid-level task at US$4 to US$7, and only 1 to 5% of agents are really useful.
Mixed
Quothling, Hacker News commenter
Comment on: Microsoft Copilot Cowork Exfiltrates FilesHacker News comment
An IT team piloting Copilot Cowork is impressed, but is holding a wider rollout until Microsoft adds admin controls over what users can do with it.
Mixed
Quothling, Hacker News commenter
Comment on: The last six months in LLMs in five minutesHacker News comment
Teams transcribes every meeting very well and Cowork drafts most presentations, but the regular Office 365 Copilot was poor enough that many colleagues stopped believing in AI.
Positive
p_ing, Hacker News commenter
Comment on: Microsoft 365 Copilot's commercial failureHacker News comment
The commenter's organization uses Copilot Studio heavily: the lite version reaches SharePoint and Teams data, and the full version any data with a connector, REST API, or MCP server.
Mixed or negative, 6
Negative
pbronez, Hacker News commenter
Comment on: Microsoft lowers AI software growth targetsHacker News comment
Deploying Copilot Studio in a GCC tenant, the commenter hit incomplete Power Platform integration, nondeterministic agents, no secure way to share agents, and a slow editor.
Negative
CharlesW, Hacker News commenter
Comment on: Microsoft is struggling to sell Copilot to corporations - employees want ChatGPTHacker News comment
Spent months trying to make a Copilot Studio agent in Office Copilot work as well as an OpenAI GPT and never could; Microsoft staff offered no useful direction.
No opinion stated
Mayank-Draup, GitHub user
Streaming: custom engine agent responses longer than 2 minutes can't be delivered in M365 CopilotGitHub issue, OfficeDev/microsoft-365-agents-toolkit
When the team's agent takes longer than 2 minutes, Microsoft 365 Copilot rejects the stream and the user gets no answer, even with progress updates every 25 seconds.
No opinion stated
ankitjaggi, GitHub user
MCP agent connector with DCR works in Copilot Cowork but returns "User does not have sufficient permissions" in Copilot Chat and ResearcherGitHub issue, OfficeDev/microsoft-365-agents-toolkit
The same MCP agent connector works in Copilot Cowork but fails every tool call in Copilot Chat and Researcher, which Microsoft's Store validation team reproduced and which blocks store certification.
Positive
a_bonobo, Hacker News commenter
Comment on: Microsoft 365 Copilot's commercial failureHacker News comment
At a non-tech employer's slow rollout, non-technical participants love Copilot and IT values its defined data and risk policies; its agents handle only basic document questions.
No opinion stated
Tamir Ishay Sharbat, Zenity Labs
AgentFlayer: When AIjacking Leads to Full Data Exfiltration in Copilot StudioSecurity research post, Zenity Labs
One emailed prompt injection made a Copilot Studio agent read its knowledge file of customer records and send the whole file to the attacker.
Every concern practitioners raise, most raised first.
Counted for Leaders and teams rolling it out.
Negative
mugenjin23, Hacker News commenter working in Australian local government
Comment on: Microsoft abandons personal AI chatbot race with Copilot rebootHacker News comment
In Australian local government, organizations struggle to explain the many Copilot products and features to staff, and each rebrand adds work.
Mixed
Quothling, Hacker News commenter
Comment on: Microsoft Copilot Cowork Exfiltrates FilesHacker News comment
An IT team piloting Copilot Cowork is impressed, but is holding a wider rollout until Microsoft adds admin controls over what users can do with it.
Negative
wkat4242, Hacker News commenter who manages Microsoft 365 at a large company
Comment on: Microsoft 365 Copilot's commercial failureHacker News comment
An M365 admin lists SharePoint agents switched on by default, constant renames, and Copilot Chat adding history, which forced a full DPIA redo.
Mixed
cowprince, IT administrator
An admin on E5 with Copilot licenses finds Office apps inconsistent when trying to disable Copilot in individual apps, as Microsoft's article suggested was possible.
Negative
bongodongobob, Hacker News commenter in IT support
Comment on: OpenAI reaches agreement to buy Windsurf for US$3BHacker News comment
An IT admin says the many products named Copilot make help desk requests slow, because staff must first work out which Copilot a user means.
Negative
pbronez, Hacker News commenter
Comment on: Microsoft lowers AI software growth targetsHacker News comment
Deploying Copilot Studio in a GCC tenant, the commenter hit incomplete Power Platform integration, nondeterministic agents, no secure way to share agents, and a slow editor.
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.Organization-built agents go through admin approval before users can add them. Admins can enable, block, or remove agents in the Microsoft 365 admin center. Manage agents in the Microsoft 365 admin center
Open them in the recordCounted for Leaders and teams rolling it out.
Mixed
wkat4242, Hacker News commenter who helps manage Microsoft 365 at a large company
Comment on: How We Rooted CopilotHacker News comment
Explains that staff label documents public to avoid restrictions, so classification cannot be relied on to keep sensitive content away from Copilot.
Mixed
New-Present-9862, IT team planning an enterprise Copilot rollout
Microsoft... Pre-Enterprise Rollout of Copilot: How Are You Mitigating oversharing links? : sysadminReddit post
Planning an enterprise Copilot deployment, the poster worries that licensed Copilot will surface sensitive SharePoint and OneDrive files shared with everyone or the whole organization.
Mixed
Adziboy, Reddit commenter in IT administration
Comment on: Microsoft... Pre-Enterprise Rollout of Copilot: How Are You Mitigating oversharing links? : sysadminReddit comment
Found oversharing links very hard to prevent, so the team labels all data with sensitivity labels and keeps data that cannot go to the cloud on premises.
Negative
Dzemal Sukurica, AI Lead for Microsoft Copilot, Eindhoven University of Technology (TU/e)
Experts raise privacy concerns about Microsoft 365's CopilotUniversity news article, Cursor (tue.nl)
TU/e kept Microsoft 365 Copilot to a small pilot after SURF's privacy assessment. Its Chief Privacy Officer warns that Copilot can draw on any document open to the whole university, intranet pages included.
Mixed
EquivalentPace7357, Reddit commenter in IT administration
Comment on: Microsoft... Pre-Enterprise Rollout of Copilot: How Are You Mitigating oversharing links? : sysadminReddit comment
Before enabling Copilot the team audited org-wide sharing links with PowerShell and Graph, fed them into Purview labeling and DLP, and restricted indexing by label.
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.SharePoint Advanced Management, included with a Copilot license, offers restricted content discovery to keep high-risk sites out of Copilot responses. It also offers reports that find overshared sites and sharing links. SharePoint Advanced Management overview
Open them in the recordCounted for Leaders and teams rolling it out.
Raised as a concern, 5
Mixed
DramaticErraticism, Microsoft 365 admin at a large company
Great news, CoWork for CoPilot has left beta...bad news, they are charging for it!!! : sysadminReddit post
An admin at a large E5 company says users like Copilot Cowork. In the thread, metered credits put a mid-level task at US$4 to US$7, and only 1 to 5% of agents are really useful.
Mixed
jabroni_salad, Hacker News commenter
Comment on: Microsoft drops AI sales targets in half after salespeople miss their quotasHacker News comment
Likes finding facts across Teams chats but runs out of context quickly; the workplace is reclaiming licenses from light users because US$21 a month is too much for a novelty.
Negative
Vegenoid, Hacker News commenter who evaluated Microsoft licensing for an employer
Comment on: The Microsoft 365 Copilot launch was a disasterHacker News comment
Tasked with recommending Microsoft licenses including Copilot, the commenter found the offerings hard to understand because of frequent renames and overlapping licenses.
Negative
mugenjin23, Hacker News commenter working in Australian local government
Comment on: Microsoft abandons personal AI chatbot race with Copilot rebootHacker News comment
In Australian local government, organizations struggle to explain the many Copilot products and features to staff, and each rebrand adds work.
Negative
bongodongobob, Hacker News commenter in IT support
Comment on: OpenAI reaches agreement to buy Windsurf for US$3BHacker News comment
An IT admin says the many products named Copilot make help desk requests slow, because staff must first work out which Copilot a user means.
Discussed without concern, 1
Positive
p_ing, Hacker News commenter
Comment on: Microsoft 365 Copilot's commercial failureHacker News comment
The commenter's organization uses Copilot Studio heavily: the lite version reaches SharePoint and Teams data, and the full version any data with a connector, REST API, or MCP server.
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.Microsoft 365 Copilot lists at US$30 per user per month paid yearly. Copilot Business lists at US$21, with US$18 a first-year promotion on annual orders placed July 1 to December 31, 2026. E7 bundles Copilot at US$99. Microsoft Copilot for business: plans and pricing
Open them in the recordCounted for Leaders and teams rolling it out.
Raised as a concern, 4
Negative
sandos, Hacker News commenter
Comment on: Claude Opus 4.6Hacker News comment
At a global conglomerate that enrolled at least 1,500 people in business Copilot plans, the commenter believes few use even 10% of their quota.
Negative
nosianu, Microsoft 365 global admin (self-described)
Comment on: Microsoft lets bosses spot teams that are dodging CopilotHacker News comment
A Microsoft 365 global admin for a 50-person business ignores Copilot prompts because nothing explains what it would do; the company pays for ChatGPT.
Negative
ezst, Hacker News commenter
Comment on: GPT 4.5 level for 1% of the priceHacker News comment
After the commenter's organization rolled out Copilot in Office, they doubt writing help gets past the novelty or is something most people will pay for.
Mixed
jabroni_salad, Hacker News commenter
Comment on: Microsoft drops AI sales targets in half after salespeople miss their quotasHacker News comment
Likes finding facts across Teams chats but runs out of context quickly; the workplace is reclaiming licenses from light users because US$21 a month is too much for a novelty.
Discussed without concern, 1
Mixed
Australian Treasury, internal evaluation of its Copilot trial
Evaluation of a trial of generative AI (Copilot) in The TreasuryGovernment evaluation report, evaluation.treasury.gov.au
Treasury found it could deploy Copilot in its environment, and almost two-thirds of users found it useful for basic administrative tasks. Limits in the product and privacy and security restrictions held back its value for complex work.
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.Copilot controls include measurement and reporting through Copilot Analytics for readiness, adoption, and impact. Copilot controls overview
Open them in the recordStatements this group's sources agree on, then the ones close to agreement.
Close to agreement: each needs 12 supporting sources
Use depends more on training, named use cases, and colleagues who show what works than on assigning a license.
8 supporting sourcesSummaries of meetings, email threads, and documents, and search across them, save people time they would spend catching up.
6 supporting sourcesCopilot earns its place on first drafts of documents, decks, and emails that a person then finishes.
5 supporting sourcesQuestions the sources split on, with the count on each side.
Evidence remains mixed
Everyone at once, 3 sources
Some organizations license departments or the whole company, then measure use.
Selected roles first, 8 sources
Others pilot with volunteers or selected roles, fund licenses by team, or reclaim licenses from light users.
The three groups, then every practitioner by the context they work in, and how emphasis changes with scale.
Three groups drawn from the same sources. Choose a group to select whose experience counts. Within that group, each author counts once in the practitioner sentiment, with their latest view. Signals open on Leaders and teams rolling it out.
57 sources. These accounts focus on licensing, oversharing, admin controls, and measured rollout results, from IT leaders, departments, and organization-wide evaluations.
51 sources. These accounts focus on answer quality, Microsoft 365 integration, and day-to-day use for people using Copilot in their own work.
108 sources. This gives the broadest read, but combines individual users with leaders deciding licenses for a whole organization, who can reach different conclusions.
108 sources from practitioners, by the context they work in. Filled rows count for this group.
Emphasizes Answer quality, Excel and PowerPoint help, and meeting recaps
Emphasizes Team pilots, Copilot Studio agents, and meeting notes
Emphasizes Licensing, oversharing, admin controls, and agent costs
Emphasizes Measured time savings, adoption rates, and license scope
What changes with scale. Individual users talk about whether Copilot's answers and edits hold up in Word, Excel, and PowerPoint. IT leaders and admins talk about licensing, oversharing in SharePoint, and settings that arrive switched on. Organization-wide evaluations report measured minutes saved, adoption rates, and the training it took to reach them.
Where this group's sources were published.
Counted for Leaders and teams rolling it out, from 57 sources. Change group
73 sources for this group, 16 of them documentation
Older evidence is retained when it still applies. Sources that depend on an app name or setting Microsoft has since changed stay in the history, marked Needs a recheck.
See sources by originWho each source is, how it knows, whether it still applies and what it counts toward.
Newest first, on its own page. Every count above opens its sources there.
Read all 124 sourcesAbout this read
A running review of what people using Microsoft Copilot at work report, from individual users to company-wide rollouts. Each finding links to its sources, and each conflict shows both sides.
Counts cover only the sources GILD reviewed.
Choose a group to select whose experience counts. Within that group, each author counts once in the practitioner sentiment, with their latest view. GILD files each source in a group from how the account describes its author. A role or employer marked self-described is what the source states about themselves. Documentation checks facts for every group and never counts toward sentiment.
A source's date and whether it still applies are separate: a recent post can describe an old version, and an old one can still hold.
Flagged for a recheck, of the sources that take a side: leaders and teams 0 of 40, individual users 2 of 30, all users 2 of 70.
Each row opens its sources for the group you are reading.
| Finding | Leaders and teams rolling it out | Individual users | All users | Sources |
|---|---|---|---|---|
| Current readCurrent sentiment | 12 positive, 15 mixed, 13 negative | 6 positive, 8 mixed, 16 negative | 18 positive, 23 mixed, 29 negative | Sources |
| PraiseMicrosoft 365 Reach | 4 of 4 positive (under the minimum) | 5 of 12 positive | 9 of 16 positive | Sources |
| PraiseMeeting Recaps | 5 of 5 positive | 4 of 5 positive | 9 of 10 positive | Sources |
| PraiseAgents & Cowork | 4 of 10 positive | 2 of 3 positive (under the minimum) | 6 of 13 positive | Sources |
| PraiseEnterprise Data Terms | 5 of 7 positive | 2 of 2 positive (under the minimum) | 7 of 9 positive | Sources |
| PraiseAnswer Quality | 1 of 4 positive (under the minimum) | 3 of 12 positive | 4 of 16 positive | Sources |
| ConcernInconsistent Output | 4 of 4 raise it (under the minimum) | 6 of 6 raise it | 10 of 10 raise it | Sources |
| ConcernLicense Cost & Confusion | 5 of 6 raise it | 1 of 1 raise it (under the minimum) | 6 of 7 raise it | Sources |
| ConcernOversharing | 5 of 5 raise it | 0 of 0 raise it (under the minimum) | 5 of 5 raise it | Sources |
| ConcernAdmin Controls | 6 of 6 raise it | 2 of 2 raise it (under the minimum) | 8 of 8 raise it | Sources |
| ConcernLow Adoption | 4 of 5 raise it | 1 of 2 raise it (under the minimum) | 5 of 7 raise it | Sources |
| ConcernFeature Gaps | 4 of 4 raise it (under the minimum) | 4 of 4 raise it (under the minimum) | 8 of 8 raise it | Sources |
| ConcernSecurity Flaws | 2 of 2 raise it (under the minimum) | 1 of 1 raise it (under the minimum) | 3 of 3 raise it (under the minimum) | Sources |
| ConsensusSummaries save time | 6 supporting (below threshold) | 2 supporting (under the minimum) | 8 supporting (below threshold) | Sources |
| ConsensusAdoption needs enablement | 8 supporting (below threshold) | 0 supporting (under the minimum) | 8 supporting (below threshold) | Sources |
| ConsensusAnswers follow the data estate | 4 supporting (under the minimum) | 0 supporting (under the minimum) | 4 supporting (under the minimum) | Sources |
| ConsensusCheck before it goes out | 3 supporting (under the minimum) | 1 supporting (under the minimum) | 4 supporting (under the minimum) | Sources |
| ConsensusUseful for first drafts | 5 supporting (below threshold) | 2 supporting (under the minimum) | 7 supporting (below threshold) | Sources |
| DisagreementShould Microsoft Copilot be the organization's main AI assistant? | Copilot as the main assistant 1, Another assistant first 5 (under the minimum) | Copilot as the main assistant 4, Another assistant first 4 | Copilot as the main assistant 5, Another assistant first 9 | Sources |
| DisagreementShould Copilot licenses go to everyone, or to selected roles first? | Everyone at once 3, Selected roles first 8 | Everyone at once 0, Selected roles first 0 (under the minimum) | Everyone at once 3, Selected roles first 8 | Sources |
| DisagreementDoes the paid Copilot license earn its price over what Copilot Chat or another tool provides? | Worth the license 1, Not worth it yet 4 (under the minimum) | Worth the license 1, Not worth it yet 1 (under the minimum) | Worth the license 2, Not worth it yet 5 | Sources |
| Current readEvidence mix | 73 sources | 67 sources | 124 sources | Sources |
GILD's checks of claims against official documentation, kept apart from the source counts.
