Copilot briefing for 20 to 26 October 2026: five developments that matter to IT leaders and enablement teams. Copilot Code is beginning to roll out through Frontier, bringing application creation to more business users. Other updates cover agent architectures, departmental workflows and adoption measurement. Here is what each one means and what to do about it.
The common thread is that Copilot is moving from a tool that helps people with content to a platform that lets people build applications, run agents and act on business systems. That shift is good news for productivity. It also means the governance questions IT has spent a decade answering for SaaS, Power Platform and low-code now apply to Copilot, with a faster clock and a wider user base.
1. Copilot Code brings application publishing to business users
Copilot Code has entered Frontier preview. Eligible users can describe an app, dashboard, script or internal tool in natural language, see a live preview and iterate. Where enabled, they can connect GitHub repositories and publish internal applications through Copilot Managed Runtime. The documented sharing scope is within the organisation, rather than a general public publishing service.
For me, this is the most consequential announcement of the week because it moves more application development into the hands of people closest to the work. The lesson from citizen development is that useful applications still need owners, inventories and a maintenance lifecycle. Copilot Code includes enterprise governance controls, but an internal tool can still become business-critical before anyone has agreed who will support it. That is where the shadow IT risk sits.
The questions to answer before enablement is switched on:
- Who is allowed to create applications, and who is allowed to publish them?
- What data can a Copilot Code application reach, and under whose identity?
- Which execution locations, data-access settings and auto-approval options are permitted?
- Who owns an application once its creator changes role or leaves?
- How are applications inventoried, reviewed, maintained and retired?
- How is consumption controlled, given Code sits within usage-based Copilot Credits?
What to do now: Treat Copilot Code as an application platform with an onboarding, ownership and retirement lifecycle. Reuse the governance practices from your Power Platform centre of excellence, with named owners for the Copilot-specific controls. Review execution location, GitHub connections, organisational data access and auto-approval settings before enabling it.
Microsoft announcement: Copilot Code | Copilot Code admin documentation
Copilot Code is in Frontier preview and rolling out. Availability depends on Microsoft Copilot licensing, organisational enablement and Copilot Credits for usage-based workloads.
2. Copilot Studio now has three agent architectures
Copilot Studio has introduced the Customer Conversation harness in public preview, built for customer-facing voice and chat agents with specialist sub-agents and human escalation. That brings the number of harness options to three, and the choice is now a genuine architecture decision.
| Harness | Appropriate workload |
|---|---|
| Standard | Structured, predictable conversations with a defined path |
| GitHub Copilot | Reasoning-heavy, multistep business processes |
| Customer Conversation | Customer-facing voice and chat, with routing to specialists and escalation to people |
Two points matter commercially. First, Microsoft’s documentation confirms that the harness you choose affects billing, so architecture and cost need to be assessed together. Second, a customer-facing agent carries a different risk profile from an internal one. Authentication, escalation paths, evaluation and monitoring need to be part of the design.
What to do now: Choose the agent architecture around the workload, not the most capable model. For contact-centre use cases, assess voice and chat channels, authentication, human escalation, evaluation and operating cost before committing, and remember the capability is still in preview.
Microsoft announcement: Customer Conversation harness | Copilot Studio agents overview
The Customer Conversation harness is in public preview. Building, testing, evaluating and using these agents may consume Copilot Credits. Check the preview’s channel and feature limits before designing around it.
3. Cowork gains role-based plugins
Cowork now offers Sales, Service, Education and Legal plugins: predefined skills for repeatable departmental workflows such as sales follow-up or service case triage.
This matters for anyone running an AI enablement programme. A predefined skill makes the workflow, inputs and expected output more explicit. That gives a team something repeatable to test and measure, although it does not guarantee a correct result. Departmental workflows let you attach training to a specific business outcome rather than relying on prompt counts as evidence of value.
What to do now: Move enablement from “how to prompt” towards “which business outcomes we are automating”. Pick one workflow per department, baseline it, enable the relevant plugin and measure the difference. The business outcome matters more than the number of prompts anyone writes.
Microsoft announcement: What’s new in Cowork
Role-based plugins are available to eligible Cowork users, subject to administrator controls.
4. Cowork can now act on business systems
Two further Cowork updates extend it beyond Microsoft 365 content. Power Platform connectors and event-driven tasks let Cowork respond to changes in CRM and other connected systems. Separately, an Azure DevOps plugin lets users work with backlogs, bugs, user stories, pull requests and related project information directly from Copilot.
The Azure DevOps plugin is a neat example of the shift from summarising information to acting on it: turning approved meeting notes or requirements into structured user stories and backlog items is a workflow most delivery teams will recognise.
The event-driven CRM scenarios raise the governance question that runs through this whole week. There is a clear difference between an agent preparing an action and an agent approving a consequential change. Microsoft’s announcement says Cowork prepares messages and record changes, then waits for approval. That approval boundary belongs in your workflow design.
What to do now: For every connected action, define its authority and approval gate. Preserve Cowork’s approval step for messages and record changes. Across your wider automation estate, only pre-authorise routine, low-risk actions with explicit scope, limits and an audit trail. Consequential changes still need a person to approve.
Microsoft Developer Blog: Azure DevOps in Copilot | Microsoft: Cowork connectors and event-driven tasks
Power Platform connector availability for Cowork is rolling out and not universally available. The Azure DevOps plugin is available for eligible Cowork users with the appropriate Azure DevOps Services permissions. Microsoft says it does not support Azure DevOps Server.
5. Copilot usage is an adoption signal, not a value measure
Microsoft 365 Insights Agent V2, in Frontier preview, lets agent owners explore submitted user feedback in Copilot Chat, subject to administrator-controlled sharing.
Usage counts tell you an agent is being opened. They do not tell you whether it is helping. Combining feedback with usage statistics gives agent owners a way to prioritise improvements based on what users say is broken or missing, rather than on which agents happen to be popular.
What to do now: Put feedback alongside usage and business outcome on the adoption dashboard. Check the administrative feedback-sharing settings before relying on it, and treat the feature as preview.
Microsoft announcement: Insights Agent user feedback
Insights Agent V2 is in Frontier preview and requires Microsoft 365 Copilot, Frontier eligibility and administrative authorisation.
A note on cost
Copilot Code and Cowork are covered by Microsoft 365 Copilot’s usage-based cost management. The Customer Conversation harness also uses Copilot Credits, with its own Copilot Studio billing guidance. A shared credit currency does not mean identical rates, included entitlements or administrative controls. Keep the per-user licence, agent consumption and hosted application costs visible in your forecast, and check which spending controls apply to each service.
Microsoft Learn: Usage-based billing and Copilot Credits | Customer Conversation preview and billing guidance
What this means for IT leaders and enablement teams
| Development | Who owns it | First action |
|---|---|---|
| Copilot Code | CIO, application governance | Define who can create and publish, and the application lifecycle |
| Copilot Studio harnesses | Agent architects, contact centre | Match harness to workload and cost before building |
| Role-based Cowork plugins | Department heads, L&D | Pick one workflow per department and measure it |
| Cowork business-system actions | IT, process owners | Set approval gates for every consequential action |
| Insights Agent feedback | Agent owners, adoption lead | Add feedback to the adoption dashboard |
If you are rolling out Copilot and want to know whether the governance, skills and data foundations are in place before Code and agents reach your users, start with the AI Readiness Assessment.
For five practical prompt frameworks and worked examples your team can use, download The AI Adoption Playbook. The guide is free, with your name and email required to download it.
And if you want a Copilot App and Agent Governance Assessment covering permissions, publishing, lifecycle ownership, consumption and deployment standards, get in touch.
Gethyn Ellis is a Microsoft Data Platform MVP, Microsoft Certified Trainer and founder of GRE Solutions, a UK consultancy specialising in the Microsoft data platform, AI enablement and training. He is the author of AI That Actually Works and host of the Putting the Human into Technology podcast.
PakarPBN
A Private Blog Network (PBN) is a collection of websites that are controlled by a single individual or organization and used primarily to build backlinks to a “money site” in order to influence its ranking in search engines such as Google. The core idea behind a PBN is based on the importance of backlinks in Google’s ranking algorithm. Since Google views backlinks as signals of authority and trust, some website owners attempt to artificially create these signals through a controlled network of sites.
In a typical PBN setup, the owner acquires expired or aged domains that already have existing authority, backlinks, and history. These domains are rebuilt with new content and hosted separately, often using different IP addresses, hosting providers, themes, and ownership details to make them appear unrelated. Within the content published on these sites, links are strategically placed that point to the main website the owner wants to rank higher. By doing this, the owner attempts to pass link equity (also known as “link juice”) from the PBN sites to the target website.
The purpose of a PBN is to give the impression that the target website is naturally earning links from multiple independent sources. If done effectively, this can temporarily improve keyword rankings, increase organic visibility, and drive more traffic from search results.
