Dynamics 365 Release Plans Are Retiring: What Replaces Them

TL;DR

Microsoft states plainly: "Release Plans will no longer be published starting in September 2026." Capabilities move to the AI at Work roadmap from the same month (Microsoft Learn).

Existing plans stay up: "Existing release plans will remain available for historical reference until further notice." Historical reference is not a planning document.

The change is already visible in product documentation. The Business Central update 29.0 preview page carries the line "Features are no longer linked to release plans."

2026 release wave 1 is the last full wave with a published plan. The wave 2 plan URL returns 404 rather than a page, which is what a retirement looks like from the outside.

Meanwhile update 29.0 is in public preview now, sandbox environments only, with general availability stated for October 2026.

What is Microsoft actually changing?

The publication, not the release cadence. Those are different things and conflating them will send an upgrade plan wrong.

The release-plans landing page now leads with a retirement notice: "Release Plans are retiring. Release Plans will no longer be published starting in September 2026. New Dynamics 365, Power Platform, and Dataverse capabilities will be published to the AI at Work roadmap beginning in September 2026." Microsoft frames the destination as "One always-on roadmap: Dynamics 365, Power Platform and Dataverse join the AI at Work roadmap".

So updates keep shipping. What stops is the twice-yearly document that told you, months ahead, feature by feature, what was coming and when.

Why does that document matter more than it sounds?

Because it is the input to almost every D365 upgrade process in existence.

A release plan gave a mid-market shop three things at once: a feature list to review against their configuration, a date to plan the sandbox test around, and a shape to the year. Two waves, two windows, two conversations with the partner. Whatever else you think of six-monthly waves, they made the year plannable by people who have other jobs.

An always-on roadmap gives you currency and takes away the shape. The information arrives when it arrives. That is better for a team watching continuously and worse for a team that reviews twice a year, and the second describes most Business Central customers.

Some arithmetic on what changes. Under waves, a customer reviewing each plan once spends perhaps 2 review sessions a year and is never more than 6 months behind the news. To stay equally current against a continuously updated roadmap you either check it on a cadence, say monthly, which is 12 sessions, or you accept being behind by however long your gap is. The same currency now costs about 6 times the check-ins, or it costs you the currency.

That is not an argument against the change. It is the cost nobody has put a number on.

Is the retirement already visible in the docs?

Yes, in two places, and both are worth checking yourself rather than taking on trust.

The Business Central update 29.0 public preview article carries the sentence "Features are no longer linked to release plans. Learn more in New AI at Work roadmap instead of Release Plans." That is the change landing inside product documentation rather than only in an announcement.

And the wave 2 plan simply is not there. A request for the 2026 wave 2 release-plan path returns a 404, while the landing page still lists 2026 release wave 1 (April 2026 to September 2026) and the two 2025 waves. The last full wave with a published plan is the one we are in.

What is in update 29.0 right now?

It is in public preview, and Microsoft is precise about the boundary: the preview "applies only to Business Central online sandbox environments and not to production or on-premises environments", with more detail promised "when version 29.0 is generally available in October 2026".

Three preview items stand out for anyone already running agents:

Area. Feature. Why it matters

Copilot and agents. Unified Copilot chat, aligned with the broader Microsoft Copilot experience. Reasons over business data, web results and general knowledge in one place

Copilot and agents. Payables Agent purchase order matching, now using line amounts and expected receipt dates. Matching on amount alone is where three-way match usually breaks

Copilot and agents. Run data queries with MCP Server. "query data for which no existing APIs exist"

That last one deserves attention. It extends the MCP surface we looked at when Microsoft put 650,000 actions behind the protocol into territory the API layer never covered. An agent that can define and run its own query is reaching data nobody wrote an endpoint for, which is powerful and is a governance question at the same time.

What else lands with 29.0 that is not a feature?

Two removals, and they matter more than most of the additions because they break things that work today.

We covered them when the deprecation list was published: exposing a Microsoft page as a SOAP endpoint is removed in v29.0, and on-premises deployments lose the ability to run data-defined permission sets only. Microsoft's remedy for the first is a copy, not a rewrite: take the page source into your own extension and publish it from there.

The October general-availability date now firms up that countdown. From 7 September, general availability in October is roughly 3 to 7 weeks depending where in the month it lands. The preview is available today, in sandbox, which is exactly where a SOAP inventory should be tested.

What should a Dynamics team change about how it plans?

Assign the roadmap to a person, with a cadence. The wave document used to arrive and demand attention. A roadmap does not arrive. Somebody has to go and look, and if that is nobody's named job it becomes nobody's job by about November.

Bookmark the historical plans while they are up. Microsoft says existing plans remain "until further notice", which is not a commitment. The 2026 wave 1 plan is the last complete feature list in the old format, and it is the reference your current configuration was reviewed against.

Use the sandbox preview as the new early-warning system. With no forward-looking plan, the preview build becomes the earliest concrete signal of what is coming. That is a shift from reading to testing, and testing is slower, so start earlier.

Separate "what is new" from "what is removed". The roadmap is a capabilities roadmap. The deprecation lists are separate documents and always were. Nothing about this change makes those less important, and the v29 removals are live proof.

FAQ

Are the release waves themselves going away?

Microsoft's notice is about publishing release plans, not about the update cadence. Update 29.0 is still described as 2026 release wave 2 with general availability in October 2026.

Where do I find upcoming features now?

The AI at Work roadmap, which Microsoft describes as one always-on roadmap covering Dynamics 365, Power Platform and Dataverse.

Do the old release plans disappear?

Not yet. Microsoft states they remain available for historical reference until further notice.

Can I try update 29.0 today?

The public preview is available for Business Central online sandbox environments only. Not production, not on-premises.

Does this change the deprecation timeline?

No. The v29.0 removals stand, and October general availability is the date they take effect.

The last mile

Microsoft automates the delivery of capability, and an always-on roadmap is honest about how software actually ships now. Waves were always a publishing convention rather than a description of engineering.

What a roadmap cannot tell you is which of the capabilities is worth your attention this quarter, against your configuration, your customisations and the one process that would actually move margin if it got better. That reading was the real value in a release plan, and it was never in the document. It was in the person who read it against the business. The ERP is the system of record. Turning what it records into the decision that should follow is the layer Cognilium works in, and a roadmap without waves makes that reading more valuable, not less.