The Operating Layer Inside Your Working Day
Microsoft is turning Copilot from an assistant beside the apps into an agentic layer across the tenant. Work IQ provides organizational context, Chat answers, Cowork executes, and three model vendors now sit under the Microsoft 365 model picker: Anthropic, OpenAI and SpaceXAI. Two things shape a buying decision here more than the feature list does. Which Microsoft page you read decides what you believe, because its Cowork pages and its tenant-wide release notes carry different stamps and different content. And the compliance answer splits by purchase route: what Microsoft documents for Fable 5.1 is not what Anthropic's own direct terms say about the same model.
The teammate that starts with your organization’s context
Copilot’s moat is not a single model. It is the ability to reason over the work graph, meaning mail, meetings, files, people and activity, while respecting the permissions already defined in Microsoft 365.
Context-rich and process-aware
Copilot is strongest when the task depends on what your organization has already said, decided, created and permitted.
- Tenant-aware grounding
- Native business-app outputs
- Enterprise identity and governance
Work that should finish where it began
Meeting follow-up, document creation, spreadsheet analysis, presentation production, inbox work and governed agents.
- Less copy-and-paste
- Fewer tool handoffs
- Work stays connected to its source systems
Context and control outrank model loyalty
Use Copilot when the right answer depends on organizational context and must inherit permissions, compliance and brand systems.
- Your work already lives in Microsoft 365
- IT needs visibility and policy control
- The output belongs in a Microsoft app
Microsoft closed the gap it opened, and the lag moved to the page admins read first
On September 14, 2026 Microsoft rewrote the Cowork model page and the Cowork admin and governance page on the same day. The model page now lists Claude Fable 5.1 and GPT 6 Astra; the governance page names no Anthropic model at all and the Sonnet plus Opus Advisor pairing is gone from both. The thirty-one-day disagreement between the two, running from Advisor leaving the model page on August 14 to it leaving the governance page on September 14, is closed. What is not closed is the pattern behind it: the tenant-wide release notes for Microsoft 365 Copilot are stamped September 3 with a newest entry of August 25, eleven days behind the Cowork-specific page, and that is the page most admins open first. Meanwhile any flat conclusion about retention on the Fable line is too flat, because the answer depends on whether you buy through Microsoft or direct from Anthropic, Fable 5 has been replaced rather than promoted on the Cowork list, and a third model vendor arrived in the Microsoft 365 model picker. Checked September 18, 2026.
Microsoft reconciled both Cowork pages in one day Sept 14
The Cowork model page and the Cowork admin and governance page disagreed with each other for a month. That disagreement is gone. Both pages on Microsoft Learn now carry a last-updated stamp of 2026-09-14. The model list reads, verbatim: Auto; GPT 5.5 (Frontier); GPT 5.6 Sol; GPT 5.6 Terra; GPT 6 Astra; Opus 5; Claude Sonnet 5; Claude Fable 5.1. There is no “Preview” tag on Fable 5.1. The governance page no longer names Claude, Sonnet, Opus or Advisor at all; it says only that Cowork ships with several models described on the model page, and that an admin “can turn off the Anthropic model family in the Microsoft 365 admin center under Copilot settings.” Both pages are rewritten in place, so confirm the current wording directly before you quote either one.
- Both pages stamped 2026-09-14; the gap is zero days
- Fable 5.1 and GPT 6 Astra now on the model list
- Advisor and every Anthropic model name are absent from the governance page
- Both pages are rewritten in place, so treat any quote as perishable
The lag did not disappear, it moved The new instance
Microsoft’s Cowork pages and its tenant-wide release notes carry different stamps and different content, so which Microsoft page you read decides what you believe. Do not retire that habit; retarget it. The Cowork-specific what’s-new page carries a full September 2026 section. The tenant-wide “Release notes for Microsoft 365 Copilot” page carries an updated stamp of 2026-09-03 and its newest entry is August 25, 2026, with no September entries at all. Eleven days separate the two stamps, September 3 against September 14, and the tenant-wide page is the one an admin reaches for when someone asks what changed. Keep two numbers apart here, because they measure different things. The gap between the pages is eleven days. The staleness of the release notes is a separate figure: read on September 18, that stamp is fifteen days old and its newest entry is twenty-four days old. The lesson survives its original example: a Microsoft page describes the moment it was written, and two Microsoft pages about the same product write at different speeds. Check which page you are on before you conclude that nothing shipped.
- Cowork what’s-new carries September 2026; release notes stop at August 25
- Release notes stamped 2026-09-03, eleven days behind the Cowork page
- Separately, that stamp is fifteen days old as read on September 18
- The slower page is the one with tenant-wide scope
- Both are rewritten in place; confirm directly
Fable 5 is off the Cowork list, but not documented as gone Replaced, not retired
One naming point first, because it defeats searches: Fable 5 and Fable 5.0 are the same model. Microsoft writes it as Fable 5 on the Cowork pages and as Fable 5.0 on the subprocessor page, and each page below is quoted with its own spelling. Now the substance. Any evaluation plan written before September will name “Claude Fable 5 (Preview)” and its gate. The September entry on the Cowork what’s-new page introduces the Claude Fable 5.1 model as now available as the most advanced model, replacing Fable 5, and the Cowork model page rewritten on 2026-09-14 does not list Fable 5 at all, so it is not something a user picks in Cowork today. Do not upgrade that into a retirement. Microsoft publishes no retirement date, and the subprocessor page stamped 2026-09-10, four days earlier, still describes Fable 5.0 as a model an admin can turn on: off by default for all scenarios, with no user access until the tenant admin explicitly opts in. One page drops it from a list, another still documents a switch for it, and neither says the switch was removed. If your evaluation plan, pilot group or admin change request still names Fable 5, re-scope it to 5.1 and confirm in your own admin center what is actually selectable.
- Fable 5 and Fable 5.0 are the same model under two Microsoft spellings
- Fable 5.1 presented as the replacement for Fable 5 in Cowork
- Absent from the Cowork model list rewritten 2026-09-14
- No retirement date published; the subprocessor page still documents an admin opt-in for Fable 5.0
- Any pilot scoped to Fable 5 needs re-scoping; the picker in your tenant is the authority
Retention depends on the route you buy through Two contracts, one model name
“The Fable line is the retained line” is the flat answer, and it is too flat to survive contact with either vendor’s documentation. Microsoft documents a split inside its own services. The page “Anthropic models in Microsoft Online Services,” stamped 2026-09-10, says that Anthropic models with Data Retention, “such as Fable 5.0,” carry additional controls in the Microsoft 365 admin center and “are off by default for all scenarios,” with no user access “until the tenant admin explicitly opts in.” For the newer model it says something different: “For example, for certain organizations, Fable 5.1 is available with Anthropic acting as a subprocessor and Microsoft’s Product Terms and Data Protection Addendum (DPA) apply,” and under that path “Anthropic doesn’t retain customer content (including prompts and responses).” Read that for what it is. It is Microsoft describing the terms Microsoft has contracted for as the buyer of record, inside Microsoft Online Services. It is not a statement about Fable 5.1 wherever you meet it. Bought direct from Anthropic, the same model sits under Anthropic’s own commercial terms, and the Claude guide in this series records those terms as requiring 30-day retention for Fable 5.1 and as not making it available under zero data retention unless Anthropic expressly authorizes it. So the two vendors are not contradicting each other and neither one is wrong: same model name, two contracts, two retention answers, and the route you buy through decides which applies to you. Inside Microsoft the gate attaches to 5.0 and 5.1 depends on which commercial regime your organization is under; direct from Anthropic that is not the shape of the question at all.
- Inside Microsoft, Fable 5.0 is the gated, off-by-default retained model
- Through Microsoft, Fable 5.1 can run with Anthropic as a subprocessor under Product Terms and the DPA
- On that path Microsoft states Anthropic does not retain customer content
- Direct from Anthropic the answer differs: 30-day retention, and no zero data retention without express authorization
- A flat “the Fable line is retained” answer is unqualified rather than simply false
Neither page tells you which regime you get Read the picker
The standing rule holds here, for a new reason. The model page, newer by four days, presents Fable 5.1 as an ordinary selectable model with no preview tag, no off-by-default language, no retention gate and no banner. The subprocessor page says the retention regime attaches to Fable 5.0 and that 5.1 terms apply to “certain organizations.” Neither page tells a buyer which of those describes their own tenant. And neither one speaks to the other half of the picture, which is what Anthropic’s direct terms say about the same model: a separate contract, with 30-day retention on Fable 5.1 and no zero data retention without express authorization, per the Claude guide in this series. Three documents, one model name, and only your own agreement tells you which of them governs you. So the authority on what you can select, and on what terms, is still the picker in your admin center plus your own contract, not any of the documents.
- Model page: Fable 5.1 as an ordinary selectable model
- Subprocessor page: terms vary by organization
- Anthropic direct: a separate contract with a different retention answer
- No page maps a tenant to a regime
- Take the question to your contract, not the docs
The retention tails privacy review actually asks about Numbers to carry
Thirty days is the headline and it is the least interesting number. The subprocessor page, stamped 2026-09-10, states that Anthropic, not Microsoft, stores most inputs and outputs for up to 30 days before deleting them, and that if Anthropic’s trust and safety classifiers identify potential violations of Anthropic’s Usage Policy, Anthropic may retain content, meaning inputs and outputs, for up to two years, and trust and safety classification scores for up to seven years. Two years and seven years are the figures a privacy review will stop on, and a summary that quotes only the 30-day number is incomplete. The page is rewritten in place; confirm the current wording before you paste it into a data protection impact assessment.
- Up to 30 days for most inputs and outputs
- Up to two years on a usage-policy classifier hit
- Up to seven years for classification scores
- Anthropic holds the content, not Microsoft
GPT 6 Astra in Copilot Sept 4
Microsoft published “Available today: OpenAI GPT-6 Astra in Microsoft Copilot” on September 4, 2026, by Bas Brekelmans: “GPT-6 Astra is rolling out today to users in Copilot Cowork and Copilot Studio.” Access is administrator-managed, with “Admins can manage access and configure availability for their organization through the Microsoft 365 admin center,” and “Availability may vary by region and organization.” The Cowork model page now carries Astra as the “Latest model for tough problems.” For routing purposes that matters: Astra, not GPT 5.6 Sol, is now the top OpenAI option in the selector. The post publishes no pricing and no retention language.
- Copilot Cowork and Copilot Studio, per Microsoft
- Admin-managed through the Microsoft 365 admin center
- Availability may vary by region and organization
- Astra replaces Sol as the top OpenAI choice in the picker
- No pricing and no retention language in the post
Astra token pricing exists; Copilot seat pricing did not move Scope the price
Microsoft publishing an Astra price is not Microsoft repricing Copilot, and the two are easy to conflate. Microsoft published full GPT-6 Astra token pricing on September 3, 2026 in “GPT-6 Astra: Frontier intelligence for work, now generally available in Microsoft Foundry,” by Steve Sweetman and Naomi Moneypenny: Standard Global short context at input $10.00, cached input $1.00 and output $50.00 per million tokens; long context at $20.00, $2.00 and $75.00; the US Data Zone at roughly a 10 percent premium. Note what that is and is not. Those are Microsoft Foundry token prices, not Copilot prices. The sentence a buyer needs is that no Copilot seat price changed alongside it.
- Foundry token pricing published September 3, 2026
- Short context: $10.00 input, $1.00 cached input, $50.00 output per million tokens
- Long context: $20.00, $2.00 and $75.00; US Data Zone around 10 percent higher
- No Copilot seat price changed
- Vendor pricing as published by Microsoft
Grok from SpaceXAI, a third vendor in the Microsoft 365 picker Sept 12
Microsoft 365 Copilot’s own model picker is no longer a two-vendor product. Scope the claim before you repeat it, because on the wider Copilot brand SpaceXAI was already a model vendor inside a Copilot-branded Microsoft product: the Grok guide in this series records Grok 4.5 reaching GitHub Copilot on July 28, 2026 and Grok 4.6 on August 14, 2026, roughly six weeks earlier. Different product, different admin gate, different purchase. September 12 is the date for Microsoft 365 Copilot, not the date SpaceXAI became a Copilot vendor. “Expanding model choice in Copilot with Grok,” by xiasong, published September 12, 2026 and updated September 14, puts Grok into Microsoft Word, Excel and PowerPoint through the Microsoft Frontier Program. Not Cowork, and not Copilot Studio. The admin gate is stricter than the Anthropic one: “Access to SpaceXAI models is managed through a dedicated administrative setting that is disabled by default.” Geography is limited too: “Grok is not available to Frontier customers in the European Union (EU), European Free Trade Association (EFTA), and the United Kingdom during the preview,” with no end date given. Microsoft has published no retention terms for SpaceXAI models, which makes this a third subprocessor question for privacy review rather than a fourth model in the same picker.
- Word, Excel and PowerPoint via the Microsoft Frontier Program
- Not in Cowork and not in Copilot Studio
- Dedicated admin setting, disabled by default
- Not available to Frontier customers in the EU, EFTA and the UK during the preview
- Grok has been selectable in GitHub Copilot since July 28, 2026, and 4.6 since August 14: a different product with a different gate
- No published retention terms; treat as an open subprocessor question
Grok’s absence from Cowork is unstated, not denied Boundary, not conflict
The Cowork model page does not mention Grok, SpaceXAI or the Frontier Program. That reads as consistent with the Grok post, which scopes the rollout to Word, Excel and PowerPoint, and the Grok post was itself updated on the same day the model page was rewritten. Record it accurately, though: no Microsoft page states that Grok is unavailable in Cowork. The Cowork page simply omits it. An unstated boundary is a weaker thing to plan around than a documented one, so if a Cowork policy needs to exclude SpaceXAI, write the exclusion into your admin setting rather than inferring it from a page that does not mention the vendor.
- Grok, SpaceXAI and Frontier Program all absent from the Cowork model page
- No page says Grok is unavailable in Cowork
- Treat as an omission, not a documented limit
- Enforce the boundary in the admin setting, not by inference
Opus 5 Bare name on the page
Still listed on the Cowork model page, still as the bare string “Opus 5” rather than “Claude Opus 5,” while the adjacent rows read “Claude Sonnet 5” and “Claude Fable 5.1.” The September 14 rewrite did not change that. It matters only because it defeats a search: a compliance reviewer looking for the full product name on that page will find nothing. The full name appears on Microsoft 365 Roadmap item id 568372, dated July 24, 2026, status Launched (Generally Available), but the roadmap cannot be used to confirm it: it is a client-rendered application and a direct feature-id lookup returns nothing. Carry that as an unconfirmed reading.
- Renders as bare “Opus 5” on the Cowork model page
- Roadmap item id 568372 cannot be confirmed by a direct fetch
- The roadmap is client-rendered; a feature-id fetch returns nothing
- Described on the page as for complex, high stakes work
“Claude Opus 5” in full, on a different Microsoft property Scope the absence
It is tempting to say that no page in the Microsoft 365 Copilot documentation set names Claude Opus 5 in full. A doc-set-wide absence cannot be proven by fetching pages, so scope the claim to what was actually checked. On the Cowork model page, checked September 18, 2026, the string renders as bare “Opus 5.” Separately, the Azure Databricks page “Databricks-hosted foundation models” on Microsoft Learn, stamped August 28, 2026, describes “Claude Opus 5” in full as Anthropic’s most capable hybrid reasoning model with a focus on agentic coding, code review and long-horizon autonomous work, and exposes the endpoint databricks-claude-opus-5. One Microsoft property names the model in full. The page your Copilot admins actually read does not. Both pages are rewritten in place.
- Cowork model page: bare “Opus 5,” checked September 18, 2026
- Azure Databricks foundation models page, stamped August 28, 2026, names it in full
- Endpoint
databricks-claude-opus-5 - The absence claim is limited to the pages checked, not the whole doc set
Claude Sonnet 5 Everyday tier
Announced on July 2, 2026 as rolling out in Copilot Cowork and PowerPoint, and still carried on the Cowork model page, verbatim, as “For everyday tasks and fast responses such as drafting, quick lookups, and day-to-day work.” Claude Sonnet 4.6 remains off the table. Read that beside the Opus 5 description and you have Microsoft’s own routing hint: Sonnet 5 is the everyday tier, Opus 5 is the escalation for complex, high-stakes work, and Fable 5.1 is now the top of the Anthropic ladder.
- Replaced Sonnet 4.6, which is no longer listed
- Documented for everyday tasks and fast responses
- Escalate to Opus 5 when the stakes justify it
- Fable 5.1 sits above both on the current list
Sonnet plus Opus Advisor is gone Resolved
The paired mode, where Sonnet handled the turn and Opus reviewed the result, came off the Cowork model page in August and survived on the admin governance page. It no longer survives anywhere. The governance page rewritten on 2026-09-14 names no Claude, Sonnet or Opus model and does not mention Advisor. Microsoft resolved the disagreement by deleting the losing side rather than by reconciling the two lists, which is worth noticing: a contradiction can close because a page was corrected, or because a page was emptied. If you built a review workflow on Advisor, the documented basis for it is now absent from both pages.
- Absent from the model page since August 14
- Now absent from the governance page as well
- The contradiction closed by deletion, not by reconciliation
- Design your own review pass instead
Effort levels, now documented with a default On the model page
The effort ladder is documented on the Cowork model page itself, with detail the August 2026 blog roundup did not carry. Medium is the default, and the page notes that the default can vary across models to meet the Cowork quality bar. Light is offered for lighter tasks, High and Extra High for harder ones, and Max for your hardest work. The line to put in front of a budget owner is the last one: higher reasoning effort uses more credits. Effort is therefore a cost control as much as a quality control, and it is set per task by whoever is running it.
- Medium is the default effort level
- The default can vary across models
- Light, Medium, High, Extra High and Max
- Higher effort consumes more Copilot Credits
- Model page, stamped 2026-09-14; confirm directly
The /app skill and app building Sept 10
“Build business apps with Copilot Cowork and Copilot Studio,” by Ryan Cunningham, CVP for Copilot Studio and Power Platform, September 10, 2026, adds a new agent surface. The /app skill generates a working app scaffold from a natural-language description. In Cowork it is available through the Microsoft Frontier program; in Copilot Studio it entered public preview described as rolling out over the following week. Admins get centralized visibility in the Microsoft 365 admin center, and published apps live at managedapps.cloud.microsoft.com. It is billed on the usage-based model with Copilot Credits, so app generation is a spend event, not a free feature. The Cowork what’s-new September section corroborates it as “App skill (Frontier).” The stated one-week Studio rollout window closed on September 17 with no confirmation post, so verify availability in your own tenant.
/appgenerates a working app scaffold from a description- Cowork through the Frontier program; Copilot Studio in public preview
- Centralized admin visibility; published apps at managedapps.cloud.microsoft.com
- Billed with Copilot Credits
- The stated rollout window closed without a confirmation post
Brand Kit in PowerPoint Sept 14
“Brand Kit and Skills in Copilot in PowerPoint,” by Derek Johnson, September 14, 2026, covers templates, imagery, icons, fonts and other assets, and adds a stricter mode in which Copilot follows approved layouts aligned with company standards. The governance half is the part to plan for: admins can publish approved Skills to specific users, groups or the entire organization, which turns deck production standards into something distributed centrally rather than trained into people. It is generally available with a Microsoft 365 Copilot license.
- Templates, imagery, icons, fonts and other assets
- A stricter mode that follows approved layouts
- Admins publish approved Skills to users, groups or the whole organization
- GA with a Microsoft 365 Copilot license
Domain Exclusion is back Sept 9
“Update: Domain Exclusion for Microsoft 365 Copilot,” September 9, 2026, restores an admin control over Copilot web grounding that had been withdrawn. It is described as available again. This belongs on the governance checklist rather than the feature list: it is one of the few controls that changes what Copilot may read from the open web, as opposed to what it may read from your tenant. If your web-grounding policy was written during the withdrawal, it was written without this lever and is worth re-opening.
- Admin control over Copilot web grounding
- Withdrawn earlier, now available again
- Governs open-web reading, not tenant content
- Re-check any policy written while it was unavailable
Measuring the value of Cowork Sept 9
“Measuring the value of Cowork: From AI interactions to completed work,” by Mike_Walsh, September 9, 2026, is Microsoft naming the metric that actually matters: completed work rather than interaction volume. The post body did not render on fetch, so treat it as dated and topical rather than as a source of quotable claims. The framing in its title is the useful part, and it lines up with the habit in the playbook below: measure the artifact that shipped, not the number of chats that happened.
- Microsoft framing Cowork value as completed work
- Body did not render; cited as dated and topical only
- Pairs with the outcome-measurement habit
- Do not quote claims from it until the body is readable
New connectors August 31 roundup
The August 2026 roundup names new connectors including Harvey, iManage Work, Mercury, Xero, Statista and ZoomInfo. Harvey and iManage Work are the two to notice if you run a legal or professional-services tenant: they put matter and document systems into the grounding set, which changes what a Copilot answer can be built from and therefore what your permission review has to cover.
- Harvey, iManage Work, Mercury, Xero, Statista, ZoomInfo
- Legal and finance systems entering the grounding set
- Every new connector widens the permission review, not just the answer
Local browser use reaches GA August 28
The August 28, 2026 Cowork what’s-new entry moves local browser use from Frontier to general availability for all Microsoft 365 Copilot tenants. It requires Edge. The same entry adds workspace file input for plugin tools, and event-driven tasks that fire on a matching email or Teams message, including @mentions. An agent that can drive a browser on a user’s machine is a different risk surface from one that reads the graph, so review it before it is on by habit.
- Local browser use GA for all Microsoft 365 Copilot tenants
- Requires Edge
- Workspace file input for plugin tools
- Event-driven tasks on a matching email, Teams message or @mention
Visual creation in flow
Cowork can create deck graphics, document illustrations and email imagery without forcing the user into a separate creative app. The August 25, 2026 release notes add image generation in Cowork via ChatGPT Images 2.0. With Brand Kit now shipping in PowerPoint, the approved-asset path and the generated-asset path meet inside the same deliverable, so the brand review has to cover both.
- Image generation in Cowork via ChatGPT Images 2.0
- Integrated into deliverables
- Brand Kit and generated assets now coexist in one deck
- Still requires brand and truth review
What the Copilot add-on costs Read Sept 18
Microsoft’s Copilot pricing pages list Microsoft 365 Copilot at $30.00 user/month, paid yearly, as an annual subscription that auto-renews, and state that a separate license for a qualifying Microsoft 365 plan is required. Microsoft 365 Copilot Chat is listed as included for all Microsoft Entra account users with an eligible Microsoft 365 subscription. So the commercial shape is: Chat included, the full Copilot add-on at $30.00 on top of a base plan, and Cowork execution metered separately on top of that. These pages carry no publication date; prices were read on September 18, 2026 and should be confirmed with your reseller before a budget is signed.
- Microsoft 365 Copilot add-on at $30.00 user/month, paid yearly
- A qualifying Microsoft 365 plan license is required separately
- Copilot Chat listed as included with an eligible subscription
- Cowork execution is metered on top of both
- Pricing pages are undated; read September 18, 2026
Copilot Credits, with real numbers Read Sept 18
“Metered” is not a budget line. The published numbers are these. The Microsoft Copilot Studio Pre-Purchase Plan is listed at $200.00 per license for Copilot Credit capacity packs of 25,000 Copilot Credits each, with savings of up to 20 percent through up-front purchase of Copilot Credit Commit Units, and automatic pay-as-you-go coverage when prepaid credits run out. A pay-as-you-go option with no up-front license commitment also exists, billed through an Azure subscription using the Azure meter set up for the service. Microsoft describes Copilot Credits as a common currency for eligible Microsoft services with usage-based billing. One number is still missing and it is the one you want: the unit price of a single Copilot Credit is unpublished, as are per-task Cowork and Work IQ API consumption rates. So you can forecast the size of a pack but not the number of tasks it buys.
- $200.00 pre-purchase pack of 25,000 Copilot Credits
- Up to 20 percent off with Copilot Credit Commit Units
- Automatic pay-as-you-go overflow when prepaid credits run out
- Pay-as-you-go billed through an Azure meter
- Unit credit price and per-task consumption rates are unpublished
Microsoft 365 E7, Agent 365 and the tiers around them Read Sept 18
The Frontier Suite tier was announced March 9, 2026 and reached general availability May 1, 2026. Prices re-read on September 18, 2026 on Microsoft’s own pricing pages are unchanged: E7 at $99.00 user/month, paid yearly on an annual commitment, a no-Teams SKU at $90.45, and Agent 365 standalone at $15.00 user/month, paid yearly. The surrounding enterprise tiers are E5 at $60.00 and E3 at $39.00, and each carries a no-Teams variant that is easy to leave out of a comparison: E5 No Teams at $51.45 and E3 No Teams at $30.45. On the business side the annual and monthly figures diverge enough to matter. The full picture: Business Premium with Copilot at $32.00 paid yearly or $38.40 monthly; Business Standard with Copilot at $23.50 paid yearly or $28.20 monthly; Copilot Business starting from $18.00 paid yearly, down from $21.00; the monthly figure for that tier was not re-read on September 18 and is deliberately not stated here, because any monthly figure derived from the old $21.00 yearly price is now wrong. Cowork is listed as metered on every Business tier that carries it. These pages are undated; confirm with your reseller.
- E7 $99.00, E7 No Teams $90.45, Agent 365 standalone $15.00
- E5 $60.00 and E5 No Teams $51.45; E3 $39.00 and E3 No Teams $30.45
- Business Premium with Copilot $32.00 yearly or $38.40 monthly
- Business Standard with Copilot $23.50 yearly or $28.20 monthly
- Copilot Business from $18.00 yearly, down from $21.00; monthly figure not re-read, so not stated
- Cowork metered on top of whichever seat you buy
Microsoft 365 G7 for government Launches Oct 1
“Introducing Microsoft 365 G7: Intelligence + Trust for the mission ahead,” by Chris Barry, President of US Public Sector Industries, September 15, 2026, announces the government counterpart to the Frontier Suite for GCC. It bundles G5, the Entra Suite, Copilot with Work IQ, Researcher and Analyst, and Agent 365. Launch is October 1, rolling out in phases as customers complete required GCC authorization milestones. No pricing is published. If you read the E7 card above as a public-sector buyer, this is the line that concerns you, and the phased authorization language means an October 1 launch date is not an October 1 availability date for any particular tenant.
- GCC counterpart to the Frontier Suite
- Bundles G5, Entra Suite, Copilot with Work IQ, Researcher, Analyst and Agent 365
- Launch October 1, phased by GCC authorization milestones
- No pricing published
Consumer Copilot, which this guide has never covered Read Sept 18
Employees arrive with opinions formed on the consumer tiers, so a buyer should know what those are. Microsoft’s individuals pricing page lists Microsoft 365 Personal at $9.99 per month or $99.99 per year with advanced Copilot chat at higher usage than free; Family at $12.99 per month or $129.99 per year, for one to six people with AI for the subscription owner only; Premium at $19.99 per month or $199.99 per year with extensive Copilot chat usage plus limited agentic AI usage; and Pro at $99.99 per month, monthly billing only, with the highest chat usage and the highest agentic AI usage. Two cautions. “Copilot Pro” as a standalone consumer SKU no longer resolves; the ladder is Personal, Family, Premium and Pro. And no Microsoft consumer page names which models any of these tiers get, so no consumer model access is stated here. The page is undated; prices read September 18, 2026.
- Personal $9.99 monthly or $99.99 yearly
- Family $12.99 monthly or $129.99 yearly, one to six people, AI for the owner only
- Premium $19.99 monthly or $199.99 yearly, plus limited agentic AI
- Pro $99.99 monthly, monthly billing only
- The old standalone Copilot Pro page no longer resolves
- No first-party page states which models consumer tiers receive
Consumer-side retirements Underway
Microsoft’s support FAQ on retiring Copilot features carries retirements that get almost no coverage, and they land on the consumer side where your employees notice first. Group Chat: account updates began on August 18, 2026 and roll out gradually, so that one has come due. Mico, the animated character in Copilot Voice, is being retired in waves that started in August, with no fixed end date. Copilot Labs is retiring as part of a Copilot update, with timing varying by account and no published date. The FAQ exposes no publication or update date at all, so treat every date in it as a reading taken on September 18, 2026 and confirm directly before you build a communication plan on it.
- Group Chat: account updates began August 18, 2026
- Mico retiring in waves from August, no fixed date
- Copilot Labs retiring with timing varying by account
- The FAQ exposes no date; confirm directly before relying on it
The Agent 365 licensing contradiction, both sides Two Microsoft pages
Take this to a licensing desk, and take both citations with you. The Microsoft Agent 365 overview, still stamped 2026-08-20 and still verbatim correct, says: “Microsoft Agent 365 works best when using Microsoft E5 as a pre-requisite. At least one user must be licensed with a qualifying Microsoft Agent 365 license to enable Agent 365.” That is “works best when,” not “requires.” The older page “Prepare for Microsoft Copilot by comparing E3, E5, and E7 license features,” stamped 2026-06-30, says the opposite in a different direction: Agent 365 “is the control plane for AI agents, which is a new component included exclusively in Microsoft 365 E7,” and marks Agent 365 as unavailable for both E3 and E5. But Agent 365 has a published $15.00 standalone price on the pricing page and on its own overview, which is hard to square with “exclusively.” The comparison page is the older artifact. Neither statement should become a budget line without confirmation.
- Overview, 2026-08-20: “works best when using Microsoft E5 as a pre-requisite”
- Comparison page, 2026-06-30: “included exclusively in Microsoft 365 E7”
- A $15.00 standalone price contradicts the exclusivity reading
- No first-party page states a hard prerequisite
- Both pages are rewritten in place; confirm with a licensing desk
The context layer is the real product advantage
Work IQ connects the people, content and activity already present in Microsoft 365. That turns “what does the model know?” into “what can the system retrieve, permission and use for this employee?”
Organizational relationships
Who owns the work, who attended, who collaborates and which roles matter to the task.
Files and communications
Documents, presentations, spreadsheets, emails, chats and meeting artifacts. The August 25, 2026 release notes add Viva Engage private content as a Copilot grounding source.
What changed and when
Recent edits, decisions, conversations and project movement provide operational context.
The security boundary
Copilot should surface only the information the requesting user is already entitled to access.
What Copilot may read outside the tenant
Domain Exclusion, the admin control over Copilot web grounding, was restored on September 9, 2026 after being withdrawn. It is the lever that limits open-web sources, and it is worth revisiting if your policy was written while it was unavailable.
Plugins and Dynamics
Fabric, Dynamics 365 and partner plugins extend work beyond the Microsoft content graph. New connectors in the August 2026 roundup include Harvey, iManage Work, Mercury, Xero, Statista and ZoomInfo.
Skills and instructions
Custom skills encode the process, rules and output expectations the organization wants repeated. From September 14, 2026 admins can publish approved PowerPoint Skills to specific users, groups or the entire organization.
Brand Kit and asset libraries
Brand Kit in Copilot in PowerPoint covers templates, imagery, icons, fonts and other assets, with a stricter mode in which Copilot follows approved layouts aligned with company standards.
Work IQ REST endpoint
The August 25, 2026 release notes add a unified Work IQ REST endpoint for agents and workflows, which turns the context layer into something an in-house agent can call directly rather than only something Copilot consumes.
Using the launch brief, the most recent steering meeting, the finance workbook and the project channel, build the leadership decision pack. Reconcile conflicting dates, name missing owners and end with the decisions required this week.
From today’s meeting, update the project plan, draft the stakeholder email and create the five-slide progress deck using our approved template. Do not send or publish anything until I approve.
Cowork changes Copilot from assistant to operator
The value is not a longer answer. Cowork takes responsibility for the sequence of work: planning, tool use, file creation and delivery, while the organization retains policy and cost control.
Skills
Custom instructions that make Cowork follow the organization’s preferred process and output standards.
- Create and edit in the Customize area
- Combine personal, partner and organization-wide skills
- Use for recurring work
The /app skill Sept 10
The /app skill generates a working app scaffold from a natural-language description. In Cowork it is available through the Microsoft Frontier program; in Copilot Studio it entered public preview on September 10, 2026. Published apps appear at managedapps.cloud.microsoft.com with centralized visibility in the Microsoft 365 admin center, and generation is billed with Copilot Credits.
- Scaffolds an app from a description
- Frontier in Cowork, public preview in Copilot Studio
- Centralized admin visibility
- Consumes Copilot Credits
Plugins
Connect Cowork to partner data and business systems without hard-coding every integration into the prompt. The August 28, 2026 what’s-new entry adds workspace file input for plugin tools, and the September entry adds plugins on the mobile app, discoverable and configurable through the attach menu under Skills.
- Dynamics and Fabric support
- Plugins now discoverable on mobile under Skills
- Partner marketplace
- Admin-managed availability
Automations Renamed Aug 31
The Scheduled tab is now called Automations, per the August 2026 Copilot roundup. The surface is the same idea it always was: work that runs over time, making the home page a task-management view rather than a chat history. Update your internal training, because the old label is in every screenshot you have.
- Review in-flight work
- Adjust automated tasks
- Keep ownership visible
Event-driven tasks Restated Aug 28
Cowork tasks can fire on an event rather than a clock: a matching email arriving, or a Teams message or @mention. The August 28, 2026 what’s-new entry restates this alongside the new capabilities. It moves Cowork from something a person starts to something the work starts.
- Trigger on a matching email
- Trigger on a Teams message or @mention
- Define the stop condition before you enable it
Local browser use GA Aug 28
Local browser use moved from Frontier to general availability for all Microsoft 365 Copilot tenants on August 28, 2026. It requires Edge. An agent driving a real browser on a real machine is a materially different risk surface from one reading the graph, so decide the policy before the capability arrives by default.
- GA for all Microsoft 365 Copilot tenants
- Requires Edge
- Review what a browsing agent may sign in to
Cost management
Usage-based execution requires budgets, monitoring and an explicit answer to who is allowed to spend organizational agent capacity. The /cost skill reports the percentage of the monthly credit limit remaining. Capacity is bought as Copilot Credits: a $200.00 pre-purchase pack of 25,000 credits, savings of up to 20 percent through Commit Units, and automatic pay-as-you-go coverage through an Azure meter when prepaid credits run out.
- Cost Management Dashboard
- $200.00 for 25,000 Copilot Credits
- Up to 20 percent off with Commit Units
- Pay-as-you-go overflow billed via an Azure meter
- Unit credit price is unpublished
DLP controls
Purview can restrict sensitive information and email behavior inside Cowork workflows.
- Policy follows the work
- Protect external actions
- Test real edge cases
Completed deliverables
The success metric is a usable file, update or outcome, not the sophistication of the plan. Microsoft published its own framing of this on September 9, 2026 under the title “Measuring the value of Cowork: From AI interactions to completed work,” which is the right unit of measurement even before the post body is readable.
- Inspect the result
- Require completion evidence
- Count completed work, not interactions
- Keep human sign-off where needed
The interface disappears into the applications people already use
Copilot becomes more valuable when the user does not need to export context, explain the file structure or translate the output into the application’s native format.
Draft, edit and reshape
Create and revise documents in context, with stronger control over structure, tone and targeted edits. Word for Web gained an “Enable Sonnet 5 in models menu” setting in the August 25, 2026 release notes, and Word is one of the three apps where SpaceXAI’s Grok became selectable on September 12, behind an admin setting that is disabled by default.
- Use existing templates
- Anthropic model selection since August 11
- Grok available via the Frontier Program, off by default
- Review tracked changes
Analyze with skills and rules
Use agent-style assistance, personalization and explicit rules for spreadsheet work. The August 25, 2026 release notes add Python in Edit with Copilot for Excel, and Excel is a Grok surface as of September 12, 2026.
- Inspect formulas
- Model scenarios
- Grok selectable through the Frontier Program
- Explain assumptions
Build on Brand Kit
Brand Kit, generally available from September 14, 2026 with a Microsoft 365 Copilot license, covers templates, imagery, icons, fonts and other assets, with a stricter mode in which Copilot follows approved layouts aligned with company standards. PowerPoint is also a Grok surface.
- Approved layouts and assets
- Admin-published Skills
- Source-aware narrative
- Generated visuals in flow
Move from inbox to action
Customize and refine responses using the surrounding thread, calendar and work context.
- Draft in the user’s style
- Summarize decisions
- Confirm before sending
Capture meeting reality
Use transcripts, chat and meeting artifacts to create summaries, actions and follow-up materials.
- Name owners
- Resolve ambiguous commitments
- Link back to the discussion
Curate a working context
Bring a deliberate set of references together for a sustained piece of research or planning. The August 2026 roundup adds a dual OneNote experience for Copilot Notebooks.
- Control the source set
- Add references over time
- Keep citations attached
Turn language into accountable work
Planner agents can help translate decisions into tasks, ownership and progress.
- Set dates and owners
- Review dependencies
- Avoid silent task creation
Measure and manage
Adoption, policy, cost and agent activity become part of the deployment, not an afterthought. Model access for Anthropic and SpaceXAI is set here, and the SpaceXAI setting ships disabled.
- Usage and value signals
- Model and plugin policy
- Risk and spend controls
A third vendor lands in the Office apps, not in Cowork
Grok from SpaceXAI became available in Word, Excel and PowerPoint through the Microsoft Frontier Program, behind a dedicated administrative setting that is disabled by default, and not available to Frontier customers in the EU, EFTA or the UK during the preview. It is not in Cowork and not in Copilot Studio. Model choice in the Office apps is now a three-vendor question.
- Word, Excel and PowerPoint only
- Admin setting disabled by default
- EU, EFTA and UK excluded during the preview
- No published retention terms
Unified chat and a Work IQ button
The August 25, 2026 batch unifies web and work chat behind a single surface with a dedicated Work IQ button, and adds model selection for Researcher in Chat. The button is the interesting part: it makes the grounding decision explicit to the user instead of leaving it to a mode they may not have noticed they were in.
- Unified web and work chat
- Dedicated Work IQ button
- Model selection for Researcher in Chat
Viva Engage becomes grounding
Viva Engage private content can now ground Copilot answers. If your organization runs private Engage communities as a knowledge channel, this widens what a Copilot answer can be built from, and therefore what your permission review has to cover.
- Private Engage content in the grounding set
- Check community membership as an access control
- Re-run the oversharing review after enabling it
A Work IQ endpoint for your own agents
A unified Work IQ REST endpoint for agents and workflows moves the context layer from a Copilot feature to a platform an in-house team can call. That is the difference between buying Copilot and building on Work IQ.
- REST access for agents and workflows
- Plan authentication and scope before the first prototype
- Governance follows the endpoint, not the app
A dashboard for credit consumption
The Consumption Dashboard tracks Copilot credit usage for Cowork and WorkIQ API services. If you are running metered Cowork, this is the surface that tells you what it costs before the invoice does, and the /cost skill adds a percentage-remaining view inside the session itself. Neither one publishes the unit price of a credit, which remains unavailable.
- Covers Cowork and WorkIQ API services
- Pair it with admin budgets
- Unit credit price still unpublished
- Measure cost against completed work
MCP agents reach the Office apps
Agents built on the Model Context Protocol are available inside Word, Excel, PowerPoint and Outlook, the point at which an external tool stops being a Chat sidebar and starts operating on the document in front of the user.
- Word, Excel, PowerPoint and Outlook
- Scope tool access per agent
- Review what the agent can change, not only what it can read
Audit this workbook, correct only verified formula issues, run the agreed sensitivities, then create the executive deck using our Brand Kit template. Every chart must link back to the workbook source.
Turn this meeting into an action plan. Distinguish decisions, proposed actions and unresolved discussion. Create tasks only for explicit commitments and ask me to approve owners and dates.
Review this thread and the related project context. Draft a concise response in my normal style, identify the decision being requested and list any commitments the draft would create. Do not send.
Review this document against our policy and the cited source materials. Suggest changes with reasons, preserve the author’s voice and flag every statement the evidence cannot support.
Eight habits for making Copilot an operating layer, not another sidebar
The organizational value appears when teams change how they reference context, delegate work and standardize recurring processes.
Reference, do not retype
Point to the authoritative meeting, file, person or project context.
Choose Chat or Cowork
Use Chat for an answer; Cowork when the responsibility is a completed deliverable and you accept a metered bill for it.
Let Auto route inside Cowork
Auto routing is still Cowork-only. Model choice is not: Word gained Anthropic model selection on August 11, Researcher in Chat and Word for Web followed on August 25, and Grok reached Word, Excel and PowerPoint on September 12 behind an admin setting that is off by default. Elsewhere, pick deliberately, and only when the task has a property that benefits.
Finish in the native app
Ask for the Word file, Excel model, PowerPoint deck or Planner update you will use.
Preserve provenance
Keep source links, workbook references and meeting context attached to claims.
Standardize successful work
Turn repeated prompts into governed skills, plugins or agent templates, and publish the approved ones centrally rather than training them into people.
Treat permissions as quality
Incomplete or excessive access both degrade the usefulness and safety of grounding.
Measure outcomes
Track time saved, decision quality and completed work, not chat volume. And when two Microsoft pages about the same product disagree, as the Cowork what’s-new page and the tenant-wide release notes do today, the picker in your own tenant is the only answer that binds.
The context advantage becomes a liability without information hygiene
Copilot inherits the strengths and weaknesses of the Microsoft 365 environment. Permissions, labels, stale files, duplicate sources and ownership determine the quality of the AI experience. So does knowing which Microsoft page to believe, and which of three model vendors is processing the request.
Permission sprawl
Copilot can make overshared information easier to discover. Fix access before celebrating retrieval. Every connector added in the August 2026 roundup, and Viva Engage private content as a grounding source, widens the surface this applies to.
Stale authority
Several plausible versions of the same file can lead to confident grounding in the wrong source.
Model availability
The documented list changed again on September 14, 2026, and this time both Cowork pages changed together. The Cowork model page now reads Auto, GPT 5.5 (Frontier), GPT 5.6 Sol, GPT 5.6 Terra, GPT 6 Astra, Opus 5, Claude Sonnet 5 and Claude Fable 5.1, with Fable 5 replaced and Advisor gone from both pages. Do not treat that as a settled lineup. The tenant-wide release notes page is stamped September 3 and stops at August 25, so a second Microsoft page still describes an older world. Claude, OpenAI, SpaceXAI and Frontier capabilities depend on plan, rollout stage and administrator policy: read the picker first, then the dated announcement, then the Learn page, and check the stamp on all three.
Retention, and which regime your tenant gets
Rebuild any note that says the Fable line is the retained line, full stop, because the answer now depends on the route you buy through. Microsoft’s subprocessor page, stamped 2026-09-10, gates Fable 5.0: additional admin-center controls, off by default for all scenarios, and no user access until a tenant admin explicitly opts in. For Fable 5.1 it says that for certain organizations the model is available with Anthropic acting as a subprocessor under Microsoft’s Product Terms and Data Protection Addendum, and that under that path Anthropic does not retain customer content. Where retention does apply, Anthropic stores most inputs and outputs for up to 30 days, may retain content for up to two years on a usage-policy classifier hit, and may keep classification scores for up to seven years. Neither Microsoft page maps a tenant to a regime, and none of it describes the direct route: bought from Anthropic rather than through Microsoft, the same Fable 5.1 carries 30-day retention and is not available under zero data retention without express authorization, per the Claude guide in this series. The route you buy through changes the answer, so the answer comes from your contract.
A third subprocessor
SpaceXAI is now a model vendor inside Copilot, in Word, Excel and PowerPoint through the Frontier Program. The access setting is dedicated and disabled by default, and Grok is not available to Frontier customers in the EU, EFTA or the UK during the preview. Microsoft has published no retention terms for SpaceXAI models. Until it does, enabling that setting is a data protection decision, not a productivity one, and your subprocessor register needs a third row.
Web grounding controls
Domain Exclusion for Microsoft 365 Copilot returned on September 9, 2026 after being withdrawn. It is the admin control over what Copilot may ground on from the open web. Re-open any web-grounding policy written during the withdrawal, because it was written without this lever available.
Agent 365 licensing gate
Two Microsoft pages disagree and both should go to your licensing desk. The Agent 365 overview, stamped 2026-08-20, says Agent 365 “works best when using Microsoft E5 as a pre-requisite” and that “at least one user must be licensed with a qualifying Microsoft Agent 365 license to enable Agent 365.” That is “works best when,” not “requires.” The older comparison page “Prepare for Microsoft Copilot by comparing E3, E5, and E7 license features,” stamped 2026-06-30, calls Agent 365 “a new component included exclusively in Microsoft 365 E7” and marks it unavailable for E3 and E5, which sits badly beside a published $15.00 user/month standalone price. No first-party page states a hard prerequisite outright. Treat any hard prerequisite as unconfirmed.
Usage-based cost
Cowork can create value and variable spend. Capacity is bought as Copilot Credits at $200.00 for a 25,000-credit pack, with up to 20 percent off through Commit Units and automatic pay-as-you-go overflow billed through an Azure meter. Budgets should map to business outcomes, and note two gaps: the unit price of a credit is unpublished, and reasoning effort silently changes consumption because higher effort uses more credits.
External actions
Sending, publishing, changing records or contacting customers requires explicit authority and review. Local browser use reached GA on August 28, 2026, which extends this from records to anything a browser session can reach.
DLP edge cases
Policy must be tested against real agent workflows, not only ordinary document sharing.
Generated brand risk
Brand Kit and approved layouts reduce visual drift but do not guarantee factual or narrative quality. A deck can be perfectly on brand and wrong.
Agent and app proliferation
Skills, plugins, agents and now generated apps need ownership, lifecycle management and retirement, not just creation. The /app skill lowers the cost of creating an app to a sentence, which raises the cost of governing what exists.
| Control layer | Question to answer | Useful evidence |
|---|---|---|
| Identity | Who is asking and what can they access? | Effective permission review |
| Information | Which source is authoritative and current? | Labels, ownership and version policy |
| Model | Which models may process this task? | Admin configuration and the live picker first; then the dated announcement; then the Learn page and its stamp, remembering that the tenant-wide release notes run behind the Cowork page |
| Subprocessor | Which vendors process our content, and under what terms? | Anthropic, OpenAI and SpaceXAI are all in scope; the SpaceXAI setting is off by default and carries no published retention terms |
| Retention | Does this model retain our prompts and responses? | The subprocessor page stamped 2026-09-10: Fable 5.0 gated and off by default, Fable 5.1 terms varying by organization, and where retention applies, up to 30 days, up to two years on a classifier hit and up to seven years for classification scores |
| Commercial | Do we hold the prerequisite licenses? | The Agent 365 overview says “works best when using Microsoft E5 as a pre-requisite”; the older E3, E5 and E7 comparison page calls Agent 365 an E7 exclusive; a $15.00 standalone price exists. Confirm with your licensing desk |
| Cost | What will execution consume? | Copilot Credits at $200.00 per 25,000-credit pack, Commit Units, the Consumption Dashboard and the effort level chosen per task |
| Agent | What can it change and spend? | Tool scope, budgets, action logs and whether local browser use is enabled |
| Outcome | Did the work create measurable value? | Completed artifact, cycle time and quality review |
First-party evidence behind this guide
Every item below is a dated first-party article. Microsoft’s reference documentation is a different kind of artifact: those pages are rewritten in place and carry a last-updated stamp rather than a publication date, so they are attributed inline with their stamp instead of listed as sources, and should be confirmed directly. The callouts below name exactly which claims rest on that kind of page, and where Microsoft’s own pages disagree with each other, both sides are cited with a note of which is older.
AI Mindset