Changelog
Content in this guide changes as Microsoft ships features and UK guidance evolves. The most recent full review was on 9 July 2026.
2026-10-26
Updated Copilot limits, sensitivity-label governance, Copilot Studio protected-content guidance and SharePoint administration controls.
- Clarified that Microsoft 365 Copilot provides priority access rather than unlimited use: Auto covers everyday AI, while manually selected models can be subject to fair-use or model-specific limits and advanced work can use Copilot Credits.
- Added a practical 'human access allowed, no AI processing' pattern using sensitivity labels with Microsoft Purview DLP.
- Documented the direct-prompt-upload DLP limitation and the role of endpoint/browser DLP controls in stronger no-AI policies.
- Updated Copilot Studio guidance for sensitivity-labelled content, including VIEW and EXTRACT rights where encryption applies.
- Updated SharePoint guidance for general-availability rollout, the 1 November 2026 retirement of preview PowerShell controls, Site AI settings, Restricted Content Discovery and advanced Copilot Credit usage.
- Expanded governance health checks to cover model/usage limits, AI-exclusion labels, endpoint/browser controls and SharePoint AI settings.
2026-10-09
Added evidence-based guidance on using personal and sensitive data with Copilot.
- Added a dedicated guidance page that starts with the organisation's own AI and data-protection policy rather than a blanket prohibition on PII.
- Added published examples from DSIT Redbox, the Single Source Regulations Office and the Greater London Authority showing different legitimate policy positions on sensitive and personal data.
- Added Microsoft Purview DLP examples showing how organisations can technically block selected sensitive information types or sensitivity-labelled content from Copilot processing.
- Added an ICO-aligned decision path covering authorisation, lawful purpose, data minimisation, special category/criminal-offence data, retention and consequential decisions.
- Updated the survey-comments use case so anonymisation is presented as a risk-minimisation choice where appropriate, rather than an absolute prerequisite for every use.
2026-10-09
Updated the main Guidance page to use risk-based language for personal and sensitive data.
- Removed wording in the Start here tier that implied candidate, safeguarding, HR, malpractice, appeals and complaints information should simply be treated as content not to upload.
- Replaced it with a policy-first decision model covering purpose, necessity, data minimisation, access, retention and approval requirements.
- Updated Premium Copilot guidance to state that organisations may legitimately permit, restrict or prohibit different sensitive-data uses depending on policy, risk and Microsoft 365 controls.
- Aligned the main Guidance page with the new Personal & sensitive data guidance and the revised sensitive-meeting use case.
2026-10-09
Rebuilt the Guidance experience around generic AI capabilities.
- Changed the main journey to capability-led sections: Ask & create, Work with documents, Work with data, Meetings/research/planning, Automate repetitive work, Add AI to automation, followed by clearly marked advanced sections.
- Archived narrow awarding-organisation examples from the beginner path and rewrote retained items as generic capabilities.
- Added or highlighted generic capabilities such as asking questions about documents, comparing documents, consolidating documents, completing templates from source material, extracting information, consolidating data, creating charts and using voice.
- Kept detailed Power Platform, agent, licensing, capacity and governance material in advanced sections rather than repeating it against every beginner capability.
- Moved Personal & sensitive data under Admins & Governance, Licensing & pricing under Costs & licences, and Cowork under My Copilot Access; hid the old Use cases by access navigation item.
- Retained AI Redlines as a separate set of hard boundaries rather than mixing them with normal capability guidance.
2026-10-06
Updated Guide 365 v2 for October 2026 Copilot Studio roadmap changes covering persistent Agent identity, self-learning, reusable Skills and Computer use.
- Expanded Future capabilities with Persistent Agent Identity, self-learning, the organisational Skills catalogue, Computer use, app-and-intent-first Agent creation and October maker-sharing status changes.
- Updated MCP Apps to reflect October preview availability and November rollout planning.
- Added Copilot Studio guidance distinguishing Microsoft Entra Agent IDs from the newer persistent Microsoft 365 working identity for Agents.
- Added Copilot Studio guidance on reusing governed organisational Skills rather than duplicating instructions and workflows across Agents.
- Expanded Build & Automate safety options for persistent Agent identities, Computer use and self-learning recommendations.
- Added Computer use to the Build & Automate tool-selection decision table as a governed fallback where no suitable API or connector exists.
- Expanded lifecycle health checks and the minimum governance register to cover Agent identities/accounts, reusable Skills and accepted/rejected self-learning recommendations.
- Updated Agent 365 guidance to include first-class Agent identities, sponsors, permissions and identity retirement.
- Published a new Copilot Studio use case: Automate a legacy admin task with Computer use, including least-privilege, testing, monitoring and human-supervision safeguards.
- Updated the Copilot Studio tier to include Computer use, organisational Skills and Agent identity considerations.
- Refreshed relevant Microsoft sources and review dates to 6 October 2026.
- Ran the site-wide internal-link check after the changes.
2026-09-21
Updated Guide 365 v2 for September 2026 Copilot Studio rollout changes, with new cost visibility and a practical external-change tracking example.
- Updated Future capabilities to mark September Copilot Studio safety and workflow features as rolling out, with tenant availability caveats.
- Added Agent-level cost visibility to Future capabilities, covering Preview/History, Agent Evaluations and Monitor.
- Updated Copilot Studio guidance to explain that building, previewing, testing, evaluating and running Agents can consume Copilot Credits and that makers can increasingly inspect those costs directly.
- Expanded Credits & capacity from forecast-only planning to forecast, test, inspect development/evaluation consumption, monitor production and review value.
- Updated governance health checks and the minimum register to include execution identity, actual development/evaluation/production consumption and review of approval gates.
- Updated the qualification-review Notebook use case with a practical example of comparing a revised DfE implementation plan against earlier plans to identify moved milestones, stale assumptions and actions needing review.
- Refreshed cited Microsoft and DfE sources and review dates to 21 September 2026.
- Checked internal links after the amendments.
2026-09-16
Added two practical automation examples covering document consolidation and AI-assisted invoice processing.
- Added a Premium Copilot use case showing how multiple completed forms can be consolidated into an approved master Word template using Copilot in Word and referenced source files.
- Added a Power Automate with AI use case for invoice intake: detect invoice files, extract key fields with AI Builder, write data to an Excel table, rename the invoice with a reference and move it to a processed folder.
- Added review, confidence, duplicate-trigger and AI credit/capacity caveats so the invoice example is suitable for production guidance rather than just a demo.
- Linked both use cases to current Microsoft support and Learn documentation.
2026-09-15
Expanded licensing guidance to cover annual commitments, Power Platform pricing and AI/Copilot Credit capacity limits.
- Changed the pricing presentation so monthly-equivalent prices clearly show the underlying annual commitment and annual cost.
- Added UK pricing for Power Automate Premium, Process, Hosted Process, Power Apps Premium and Copilot Studio capacity/pay-as-you-go.
- Added guidance that licensing and AI capacity are separate: correctly licensed Power Automate/Power Apps users can still be blocked when AI Builder and Copilot Credit capacity is exhausted.
- Added current Power Platform request allocations and explained that automation limits can be hit independently of AI credits.
- Added the November 2026 transition note for seeded AI Builder credits, which increases the importance of Copilot Credit budgeting.
- Clarified that eligible Microsoft 365 Copilot agent usage in Microsoft 365 channels can be included for licensed users, while other Copilot Studio and advanced agent scenarios may remain metered.
2026-09-15
Added dedicated Microsoft 365 licensing and pricing guidance for Copilot and Agent 365.
- Added current UK public list prices for Microsoft 365 E3, E5 and E7, Microsoft 365 Copilot and Agent 365.
- Added comparison totals showing E3/E5 plus Copilot and Agent 365 against E7.
- Clarified that Agent 365 enterprise prerequisites mean E3 customers may also need Defender Suite and Purview Suite, so the £11.50 add-on alone is not the full cost.
- Clarified that Agent 365 is licensed per user rather than per agent, and distinguished foundational capabilities from premium Agent 365 capabilities.
- Added a visible Licensing & pricing navigation item.
2026-09-07
Refocused the latest Copilot updates on practical use cases, future capability planning and proportionate safety controls rather than adding marking-specific guidance.
- Published a new Future capabilities page with cited Microsoft sources covering interactive MCP Apps, Work IQ in Copilot Studio, agents as workflow steps, maker roles, safer identity/sharing controls and human approval for selected Agent actions.
- Added Future capabilities to the Build & Automate navigation and linked it from relevant build, admin and Copilot Studio guidance.
- Added a practical safety-options table covering read-only access, sandboxes, scoped sources, designed execution identities, human approval, read-only Work IQ, evaluation, spend limits and review dates.
- Updated IT/admin guidance to present safety as proportionate configuration choices rather than simply enable/disable decisions.
- Updated Copilot Studio and advanced-tier watch-outs for safe execution identities, human approval gates, read-only connected tools and evaluation evidence.
- Updated the qualification/policy approval Agent use case to show controlled actions with human approval rather than autonomous approval.
- Updated Copilot Notebooks use cases to include selected Outlook email and Teams meeting context.
- Updated the centre-support analysis use case to include Python-backed Excel analysis with validation safeguards.
- Updated the centre-support Agent example around tightly scoped approved SharePoint knowledge.
- Expanded the safe Agent sandbox use case with dummy data, test identities, read-only/scoped designs and non-production connections.
- Checked all 24 published pages for broken internal links; none were found.
2026-09-05
Updated Guide 365 v2 for late-August and September 2026 Microsoft 365 Copilot changes affecting governance, knowledge sources, chat, Notebooks and Excel.
- Expanded Copilot Studio guidance with human approval for sensitive tool calls, Agent Readiness, richer evaluations and structured knowledge sources including SharePoint Lists, Dataverse and Azure SQL.
- Added SharePoint Authoritative Sites guidance for prioritising trusted canonical content in Copilot Search.
- Updated Copilot access guidance for the unified web/work chat experience and Work IQ control.
- Added a new published Copilot Notebooks page covering files, SharePoint, Outlook emails, Teams meetings, permissions, evidence completeness and sharing.
- Updated Copilot in apps guidance for Python-powered advanced analysis in Excel, including validation requirements for consequential assessment-sector work.
- Strengthened lifecycle governance checks to record approval gates and test/evaluation evidence and to rerun assurance after material changes.
- Added Copilot Notebooks to the discovery journey, related links and site navigation.
- Checked all internal links across the site; no broken internal links were found.
2026-09-05
Restructured the guide around user discovery, access and lifecycle
- Added a new Start Here journey that begins with what users can see and what they are trying to do, rather than internal tiers.
- Added user-facing pages for current Copilot access, missing/disabled features, Cowork, managers, admins/governance, build/automate decisions and lifecycle health checks.
- Introduced the recurring model: visible ≠ enabled ≠ configured ≠ permitted ≠ included/free.
- Added 'What to ask your admin' guidance and awarding-organisation risk prompts.
- Added lifecycle guidance covering ownership, review dates, offboarding, project closure and retirement of licences, agents, connectors, Cowork access and automations.
- Added new custom navigation: Start Here, My Copilot Access, Cowork, Build & Automate, Managers, and Admins & Governance.
- Checked all standalone-page internal links; no broken links remain.
2026-09-05
Expanded the user discovery journey from basic Copilot through automation and Copilot Studio
- Added a Personal Agents page that positions early Microsoft 365 Agents as saved specialist helpers with limited organisational grounding, not autonomous AI workers.
- Added a Microsoft 365 Copilot Premium page explaining the practical unlock: richer work grounding through Microsoft Graph and Work IQ, while making clear that admin-controlled features can still remain unavailable.
- Added a Copilot in apps page covering Word, Excel, PowerPoint, Outlook and Teams, including why users can see different capabilities in the same app.
- Added a Shared Agents page covering recipient licensing, source permissions, admin sharing controls, co-ownership and lifecycle.
- Added a Copilot in SharePoint page covering scoped knowledge, SharePoint Agents, workflow features, permissions, preview/rollout differences and site/admin controls.
- Added a Power Automate page focused on persistent processes, non-AI-first automation, orphaned-flow risk and ownership/offboarding.
- Added a Copilot Studio page for cases where an Agent becomes an organisational service requiring actions, connectors, channels, environments, publishing, credits and formal ownership.
- Added a dedicated Copilot Connectors page explaining visible vs enabled vs configured vs permitted vs healthy, including self-service connector models and access-risk checks.
- Expanded the My Copilot Access and Build & Automate navigation to expose these stages directly.
- Updated Start Here with a numbered discovery path from current access through Personal Agents, Premium, in-app Copilot, shared Agents, SharePoint, Cowork, Power Automate and Copilot Studio.
- Updated related links across access, Cowork, admin and build/automate guidance; all 16 standalone pages pass the internal-link check with no broken links.
2026-09-05
Reworked the use-case library around the user access journey
- Replaced user-facing tier labels with journey language: Start here, Personal Agents, Premium Copilot, Richer Agents, Automate basics, Add AI to flows, Build business tools, Organisational Agents, and Advanced & governed.
- Kept the internal tier IDs unchanged for maintenance and filtering.
- Updated the advanced stage to reflect that Copilot Cowork is now generally available, while still requiring admin discoverability, usage-based billing and organisational controls.
- Added a new 'Browse use cases by what you have access to' page explaining how to read the Library and why a later stage is not automatically better.
- Added 'Use cases by access' to the My Copilot Access navigation and linked it from Start Here.
- Updated representative use cases across Chat, Personal Agents, Premium Copilot, work-grounded Agents, Power Automate, Power Apps, Copilot Studio and Cowork with plain-English licence and credit dependencies.
- Updated Cowork examples to current availability and credit-control language, including user/group limits and /cost.
- Removed temporary maintenance test use cases created while checking internal routing.
- Ran the internal-link checker across all standalone pages; no broken links remain.
2026-09-05
Clarified Cowork model choice, external connectivity and safe summarisation
- Updated Cowork guidance for UK tenants: Anthropic/Claude is optional and Cowork can continue with allowed non-Anthropic models such as GPT models.
- Updated the Cowork access diagnostic so model availability is clearly separate from Cowork access, billing and admin enablement.
- Added the same UK model-governance rule to the admin page and the Advanced & governed library stage.
- Added a new 'AI can process information — but it does not automatically connect to it' page explaining the difference between AI Builder/prompt processing and connectors, HTTP/API calls or external integration.
- Added a manual-vs-automated DfE example: a web-grounded Agent can support a person-led check, while unattended monitoring needs a scheduled Power Automate design with an approved fetch/integration step.
- Added a new use case for automatically monitoring a DfE page for changes, including Premium/integration dependencies, optional AI interpretation, source checking and lifecycle ownership.
- Updated the existing public regulatory research Agent use case to distinguish interactive web research from unattended monitoring.
- Added a new 'Summarising safely' page explaining that summaries are selective, users should define what matters, use coverage/gap checks and validate critical points against the source.
- Updated the public-guidance summarisation use case with omission risks, AO-specific importance criteria and a two-pass coverage method.
- Added navigation entries for Summarising safely and AI vs connectivity, and linked both into the main journey.
- Ran the internal-link checker across all 19 standalone pages; no broken links remain.
2026-09-05
Added a manager-facing pricing, licensing and credit-capacity guide
- Added a new top-level 'Costs & licences' section with current UK list-price guideposts checked 5 September 2026 and clear warnings that pricing, agreements and promotions can change.
- Added a commercial snapshot covering Copilot Chat, Microsoft 365 Copilot, Power Automate Premium/Process, Copilot Studio capacity, Agent 365, Microsoft 365 E3/E5/E7 and Cowork.
- Explained annual-commitment risk so monthly-looking prices are translated into approximate annual exposure for experiments and pilots.
- Added guidance on bundle effects, including that Microsoft 365 E7 currently includes Microsoft 365 E5, Microsoft 365 Copilot, Agent 365 and Microsoft Entra Suite.
- Added a dedicated 'Credits & capacity' page explaining AI Builder credits, Copilot Credits, pooled tenant/environment capacity, overage, monthly resets, prepaid capacity and pay-as-you-go.
- Added a prominent warning that the 5,000 AI Builder credits currently seeded by each Power Automate Premium licence are scheduled to be removed on 1 November 2026 and do not convert automatically into Copilot Credits.
- Added current Copilot Credit rate examples and a capacity-estimation method for AI-enabled automations.
- Added a dedicated Agent 365 page explaining that Agent 365 is a governance/control plane, not an Agent builder; it is licensed per user and is available standalone or included in Microsoft 365 E7.
- Updated the existing Agent 365 use case with current licensing and availability information.
- Added commercial snapshot callouts to Microsoft 365 Copilot Premium, Power Automate, Copilot Studio and Cowork pages.
- Updated manager guidance to ask for annual commitment, variable consumption and total use-case cost rather than a single monthly licence figure.
- Added Costs & licences to the main navigation, Credits & capacity beneath it, and Agent 365 under Admins & Governance.
- Linked pricing and capacity guidance from Start Here, Build & Automate, Managers and Admins pages.
- Added a warning that a Power Automate trial does not necessarily trial the full AI capacity model because AI Builder trials are discontinued.
- Ran the internal-link checker across all 22 standalone pages; no broken links remain.
2026-09-04
Added Chat vs Cowork decision guidance
- Added a new published guide item: 'Choose between Copilot Chat and Copilot Cowork'.
- Added the core decision rule: use Chat for questions and interactive help; consider Cowork for defined multi-step jobs and deliverables.
- Added guidance on Copilot Credits, prerequisites, limitations and a fallback when Cowork is unavailable.
- Added the podcast episode as the source and marked the item last checked on 4 September 2026.
2026-07-11
Added surface picker, security readiness checker, sensitive-conversations guidance and machine-readable feeds.
- New page: "Which Copilot should I use?" surface picker and surface explainer, including Web / Work / SharePoint / Studio agent and Cowork.
- New page: security readiness checker (SharePoint hygiene, DLP, labels, transcription, agent controls, credit controls).
- New guidance section: sensitive AO conversations (safeguarding, malpractice, complaints, appeals, reasonable adjustments, HR, live assessment).
- Repositioned Copilot Cowork as a future app for UK AOs — plan for, do not depend on.
- Refreshed Office Agent (stride.microsoft.com) wording to blue / on-radar preview.
- Added machine-readable feeds: /llms.txt, /data/use-cases.json, /data/sources.json.
- Sitemap now uses absolute URLs and includes new pages.
- Added optional provenance fields to use cases: last_checked, source_url, uk_availability, licence_dependency, credit_dependency.
2026-07-11
Applied first structured content update batch.
- Updated watch-outs for tiers t0, t0a, t1, t1a, t5, t6.
- Added 14 new use cases across tiers t0, t0a, t1, t1a, t2, t4, t5, t6.
2026-07-09
Initial launch of The 365 Copilot Guide for Awarding Organisations.
- Nine access tiers (t0 – t6) with 80+ use cases.
- Guidance page: red lines, licensing traps, governance model, staff allocation.
- Per-use-case feedback with work-email validation.