Microsoft Just Retired Release Waves. What it means for release planning and governance. (Markdown)

Back to article

There is no Power Platform 2026 Wave 2. Not delayed. Not renamed. On 25 August 2026, Microsoft announced that Dynamics 365, Power Platform and Dataverse would move onto a single always-on roadmap, and that the semiannual release wave model, the rhythm that has effectively governed Microsoft business applications since 2019, will end. September 2026 was the last traditional cycle.

The Release Planner, the tool thousands of organisations use to track what is coming, retires on 15 November 2026.

Most organisations have not noticed. The ones that have are starting to realise that their AI adoption plan was built on a calendar that no longer exists.

## What actually changed

The mechanics are simple enough.

Previously, Microsoft published two release plans a year. Each one listed every planned feature, its preview date and its general availability date, months in advance. You could download it, circulate it, and build a twelve-month plan against it. Change advisory boards could schedule reviews around it. Training could be timed to it. Risk assessments had a natural cadence.

From now on, capabilities are published to the unified **AI at Work roadmap** as soon as Microsoft confirms them, and they progress through development stages in real time. Filtering, CSV export and RSS subscriptions are available. What is not available is the thing that mattered most: a fixed horizon.

Microsoft's stated reasoning is sound. Organisations do not experience Dynamics 365, Power Platform and Microsoft 365 as separate products, so separate roadmaps created artificial seams. And in a market where agent capability is changing monthly, holding a finished feature for four months to fit a publishing calendar serves nobody.

But the consequence is real, and it is not a documentation problem.

## Why this is a governance event, not an administrative one

Release waves did something most organisations never explicitly designed for: they provided a **free governance clock**.

Twice a year, a document arrived that forced a conversation. What is coming? What do we want? What do we need to prepare for? Who needs training? What are we turning off? For a great many organisations, that document was the only thing that triggered a structured review of their Microsoft platform at all.

Remove it, and three things break quietly.

**Planning loses its horizon.** A roadmap that updates continuously cannot be consumed in an annual planning cycle. Teams either check constantly — which nobody does — or they find out when a feature appears in production. This becomes especially challenging during a transformation project.

**Change management loses its trigger.** Under the wave model, general availability dates were known months ahead. Now features arrive when they arrive. If your change process depends on advance notice, it is now dependent on someone noticing.

**Governance loses its natural checkpoint.** This is the serious one. The pace of agent capability is accelerating precisely as the mechanism that forced periodic review has been withdrawn. That is a widening gap, not a stable one.

## The pace is not slowing down

To understand why this matters now rather than eventually, look at what has shipped in Copilot Studio in the last five months alone:

- **Computer-using agents reached general availability in May 2026.** Agents can now operate browsers and desktop applications directly — clicking, typing, navigating a user interface — with no API required. The practical implication is significant: legacy systems that were previously unautomatable because they had no integration layer are now in scope.
- **Microsoft Entra Agent ID is generally available and mandatory at environment level.** Every agent now has a real identity with scoped permissions, and an organisation-wide agent inventory is auditable from the admin centre, API and Azure Resource Graph.
- **Memory and Skills reached general availability.** Agents retain context across interactions and can share reusable instruction modules.
- **Agent-to-agent delegation is in preview**, moving organisations from isolated departmental bots toward orchestrated agents behind a single front door.

That is a substantial change in what is possible inside an organisation, delivered in under half a year. Each item individually is a capability announcement. Together they represent a shift in operating model — from AI that assists a person to AI that performs work with an identity, a memory and a permission set.

The question is not whether your organisation can adopt these. It is whether anyone in your organisation is positioned to notice when they arrive.

## Four things worth doing now

**1. Assign ownership of the roadmap.** Someone needs to own monitoring the AI at Work roadmap with a defined review cadence — monthly is workable, quarterly is probably too slow. Without a named owner, continuous publication means nobody reads it.

**2. Replace the wave cycle with your own.** The governance clock the release waves provided for free now has to be built deliberately. A standing quarterly platform review, with a defined agenda and the right people in the room, does what the wave document used to do. However, your review clock might need to be more regular during a project.

**3. Build your agent inventory before you need it.** The tooling now exists to see every agent in your tenant. Most organisations have more than they think, and the number is growing faster than their awareness of it. An inventory you maintain is worth far more than one you assemble during an incident.

**4. Separate capability from readiness.** Computer-using agents can automate your legacy systems. Whether they should, where, and under what controls, is an entirely different question — and it is the one that determines whether the investment returns value or creates exposure.

## The underlying point

Microsoft has not made anything harder. It has removed a scaffold that many organisations were relying on without realising it. Organisations with genuine platform governance — a named owner, a review cadence, a clear view of what is deployed and why — will barely register this change. Organisations that have been navigating by the release wave calendar have just lost their instrument, at exactly the moment the terrain became more complex.

Which of those describes your organisation is worth knowing before the next capability lands unannounced.

---

## Key takeaways

- Microsoft ended the semiannual release wave model on 25 August 2026. There is no Power Platform 2026 Wave 2; features now publish continuously to the unified AI at Work roadmap.
- The Release Planner retires on 15 November 2026.
- The practical loss is not information — it is the governance rhythm the wave cycle provided for free.
- Agent capability is accelerating simultaneously: computer-using agents, Entra Agent ID, Memory and Skills all reached general availability between May and July 2026.
- The organisations least affected will be those that already have a named platform owner, a standing review cadence and a maintained agent inventory.

---

*Xperiens Digital is a Microsoft AI Cloud Partner helping Australian organisations get real operational value from their Microsoft investment. Our Dynamics 365, Power Platform and AI enablement services cover the full path — platform governance and agent readiness, solution design and delivery, and the capability uplift that makes it stick. If the roadmap change has left you without a plan, our [AI Value Diagnostic](/services#discover) is the place to start.*

--- title: "Microsoft Just Retired Release Waves. What it means for release planning and governance." slug: "end-of-release-waves" date: "2026-09-21" author: "Ben Dexter" topics: ["Business Strategy", "AI Governance", "Agentic AI", "Microsoft Dynamics 365"] summary: "Microsoft has ended the six-month release wave cycle for Power Platform, Dynamics 365 and Dataverse. Features now ship continuously. Here is why that quietly breaks the way most organisations plan and govern their Microsoft AI adoption — and what to do about it." description: "Microsoft has ended the six-month release wave cycle for Power Platform, Dynamics 365 and Dataverse. Features now ship continuously. Here is why that quietly breaks the way most organisations plan and govern their Microsoft AI adoption — and what to do about it." keywords: "Microsoft release waves, Power Platform 2026 release wave 2, Dynamics 365 release planning, AI at Work roadmap, Release Planner retirement, Copilot Studio governance, agent governance, Microsoft Entra Agent ID, Microsoft platform governance" aeo_summary: "On 25 August 2026 Microsoft ended the semiannual release wave model for Dynamics 365, Power Platform and Dataverse. Features now publish continuously to the unified AI at Work roadmap, and the Release Planner retires on 15 November 2026. Organisations lose the fixed planning horizon and the periodic governance checkpoint the waves provided, just as Copilot Studio agent capability accelerates. The response is to name a roadmap owner, run a standing platform review cadence, maintain an agent inventory, and separate what agents can do from what they should do." cta: "Talk to Ben about replacing the release wave calendar with a governance cadence of your own" --- There is no Power Platform 2026 Wave 2. Not delayed. Not renamed. On 25 August 2026, Microsoft announced that Dynamics 365, Power Platform and Dataverse would move onto a single always-on roadmap, and that the semiannual release wave model, the rhythm that has effectively governed Microsoft business applications since 2019, will end. September 2026 was the last traditional cycle. The Release Planner, the tool thousands of organisations use to track what is coming, retires on 15 November 2026. Most organisations have not noticed. The ones that have are starting to realise that their AI adoption plan was built on a calendar that no longer exists. ## What actually changed The mechanics are simple enough. Previously, Microsoft published two release plans a year. Each one listed every planned feature, its preview date and its general availability date, months in advance. You could download it, circulate it, and build a twelve-month plan against it. Change advisory boards could schedule reviews around it. Training could be timed to it. Risk assessments had a natural cadence. From now on, capabilities are published to the unified **AI at Work roadmap** as soon as Microsoft confirms them, and they progress through development stages in real time. Filtering, CSV export and RSS subscriptions are available. What is not available is the thing that mattered most: a fixed horizon. Microsoft's stated reasoning is sound. Organisations do not experience Dynamics 365, Power Platform and Microsoft 365 as separate products, so separate roadmaps created artificial seams. And in a market where agent capability is changing monthly, holding a finished feature for four months to fit a publishing calendar serves nobody. But the consequence is real, and it is not a documentation problem. ## Why this is a governance event, not an administrative one Release waves did something most organisations never explicitly designed for: they provided a **free governance clock**. Twice a year, a document arrived that forced a conversation. What is coming? What do we want? What do we need to prepare for? Who needs training? What are we turning off? For a great many organisations, that document was the only thing that triggered a structured review of their Microsoft platform at all. Remove it, and three things break quietly. **Planning loses its horizon.** A roadmap that updates continuously cannot be consumed in an annual planning cycle. Teams either check constantly — which nobody does — or they find out when a feature appears in production. This becomes especially challenging during a transformation project. **Change management loses its trigger.** Under the wave model, general availability dates were known months ahead. Now features arrive when they arrive. If your change process depends on advance notice, it is now dependent on someone noticing. **Governance loses its natural checkpoint.** This is the serious one. The pace of agent capability is accelerating precisely as the mechanism that forced periodic review has been withdrawn. That is a widening gap, not a stable one. ## The pace is not slowing down To understand why this matters now rather than eventually, look at what has shipped in Copilot Studio in the last five months alone: - **Computer-using agents reached general availability in May 2026.** Agents can now operate browsers and desktop applications directly — clicking, typing, navigating a user interface — with no API required. The practical implication is significant: legacy systems that were previously unautomatable because they had no integration layer are now in scope. - **Microsoft Entra Agent ID is generally available and mandatory at environment level.** Every agent now has a real identity with scoped permissions, and an organisation-wide agent inventory is auditable from the admin centre, API and Azure Resource Graph. - **Memory and Skills reached general availability.** Agents retain context across interactions and can share reusable instruction modules. - **Agent-to-agent delegation is in preview**, moving organisations from isolated departmental bots toward orchestrated agents behind a single front door. That is a substantial change in what is possible inside an organisation, delivered in under half a year. Each item individually is a capability announcement. Together they represent a shift in operating model — from AI that assists a person to AI that performs work with an identity, a memory and a permission set. The question is not whether your organisation can adopt these. It is whether anyone in your organisation is positioned to notice when they arrive. ## Four things worth doing now **1. Assign ownership of the roadmap.** Someone needs to own monitoring the AI at Work roadmap with a defined review cadence — monthly is workable, quarterly is probably too slow. Without a named owner, continuous publication means nobody reads it. **2. Replace the wave cycle with your own.** The governance clock the release waves provided for free now has to be built deliberately. A standing quarterly platform review, with a defined agenda and the right people in the room, does what the wave document used to do. However, your review clock might need to be more regular during a project. **3. Build your agent inventory before you need it.** The tooling now exists to see every agent in your tenant. Most organisations have more than they think, and the number is growing faster than their awareness of it. An inventory you maintain is worth far more than one you assemble during an incident. **4. Separate capability from readiness.** Computer-using agents can automate your legacy systems. Whether they should, where, and under what controls, is an entirely different question — and it is the one that determines whether the investment returns value or creates exposure. ## The underlying point Microsoft has not made anything harder. It has removed a scaffold that many organisations were relying on without realising it. Organisations with genuine platform governance — a named owner, a review cadence, a clear view of what is deployed and why — will barely register this change. Organisations that have been navigating by the release wave calendar have just lost their instrument, at exactly the moment the terrain became more complex. Which of those describes your organisation is worth knowing before the next capability lands unannounced. --- ## Key takeaways - Microsoft ended the semiannual release wave model on 25 August 2026. There is no Power Platform 2026 Wave 2; features now publish continuously to the unified AI at Work roadmap. - The Release Planner retires on 15 November 2026. - The practical loss is not information — it is the governance rhythm the wave cycle provided for free. - Agent capability is accelerating simultaneously: computer-using agents, Entra Agent ID, Memory and Skills all reached general availability between May and July 2026. - The organisations least affected will be those that already have a named platform owner, a standing review cadence and a maintained agent inventory. --- *Xperiens Digital is a Microsoft AI Cloud Partner helping Australian organisations get real operational value from their Microsoft investment. Our Dynamics 365, Power Platform and AI enablement services cover the full path — platform governance and agent readiness, solution design and delivery, and the capability uplift that makes it stick. If the roadmap change has left you without a plan, our [AI Value Diagnostic](/services#discover) is the place to start.*