For Dynamics 365 F&O and CE teams
Your sandbox gets the update seven days before production. That's your whole regression window.
Lastest replays your recorded F&O and CE business processes across every release wave and service update, diffing each screen — and it runs on your infrastructure, not ours.
The window you are given
Microsoft picks the date. You still have to prove nothing broke.
Microsoft's own documentation cannot agree on the length of the validation window: the release schedule says five business days between the sandbox update and the production update, the One Version FAQ says seven calendar days. Either way it is one week, it is unattended, and it comes round again next quarter. Version 10.0.49 went to preview on 27 July 2026 and reaches production environments by autoupdate on 2 October 2026.
Test it by hand inside the window.
Someone books the key users for a week, they click the processes they remember, and everything else goes untested. Most teams do not test platform changes at all — they let users find the bugs after the autoupdate lands. The cost of being wrong is asymmetric: there is no rollback once a service update has been applied.
- Pausing buys one cycle — only one consecutive update, since 19 February 2024
- You must take at least two service updates a year out of the four Microsoft ships
- CE release waves cannot be postponed at all — only paced, via the release channel
- Microsoft provides no free sandbox for this — you create and pay for the environment
Baseline before, diff after, sign off on evidence.
Record your order-to-cash, procure-to-pay, period-end and CE service processes once, while the sandbox is still on the current version. When the update lands, the same pack replays overnight and returns a screen-by-screen comparison — not a pass/fail count, but the actual before and after images of every screen your business runs on.
- Runs on your infrastructure — screens never leave your network
- Tests are Playwright, in your git repo — readable, editable, yours to keep
- Visual, structural and perceptual diffs catch the moved field and the renamed tab
- AI records once, replay is deterministic — the same run, the same evidence
- Timestamped artifacts for the go/no-go call and the change record
“We simply cannot trust that nothing will break. One week to test is not enough time, especially if we have no control over the timing of it.”
Karl Arensdorf, Corporate Controller, Mansfield Oil · Dynamics Communities forum
Part one · Finance and Operations
One Version, service updates, and a sandbox that updates first.
F&O application versions run on the 10.0.x train — 10.0.47, 10.0.48, 10.0.49 — each with a preview date, a self-update date, two autoupdate windows and an end-of-service date. The platform version shipped inside them is a separate 7.0.x number; it is not what your release calendar tracks.
01 · Record
Your business processes, captured before the update.
Walk a process through the browser the way a key user would: sales order, confirmation, packing slip, invoice, and the workspace tiles they check afterwards. Lastest emits deterministic Playwright code and a visual baseline for every screen. No Task recorder file to maintain, no separate parameter workbook to keep in step with the steps.
- Runs against your UAT sandbox, or any environment your network can reach
- Covers custom forms, extensions and ISV screens — whatever renders in the browser
- Tests land in your repository, so they survive the tool that created them
02 · Replay after the autoupdate
Diff the screens the week the sandbox updates.
A service update changes forms, grids, defaults and labels that no functional assertion is watching. A passing test can sit alongside a screen that has silently shifted. Lastest compares every captured screen with three diff engines — pixel comparison, structural similarity, and a perceptual model tuned to what a human eye notices — and shows you the before and after side by side.
- Text-region-aware diffing keeps dynamic dates and document numbers from firing false positives
- One overnight run covers the processes a week of manual clicking never reaches
- Timestamped screenshot artifacts you can attach to the change record
Part two · Customer Engagement
Release waves, release channels, and no postpone button.
CE runs on a different calendar and a different vocabulary. Release wave 1 covers April to September and release wave 2 covers October to March; the wave 2 plan publishes in September, and Microsoft has not yet posted the page, so treat any specific date you see quoted elsewhere as an estimate. Early access as a separate programme has been deprecated in favour of the monthly release channel, and there is no preview sign-up left to complete. What you choose in the Power Platform admin center is the channel — monthly or semi-annual — not whether the wave arrives.
03 · Model-driven apps and solutions
The customisation that quietly stopped working.
A wave lands, a form control renders differently, a command bar button moves, a business rule that depended on a control behaves differently — and nothing errors. Lastest baselines your model-driven app forms, views, custom pages and embedded canvas apps in a Dataverse environment, then replays them after the channel update and after each solution deployment.
- Same runner for CE and F&O — one regression pack, two release calendars
- Works alongside whatever you already run, including EasyRepro-based suites
- Compare a sandbox environment against production, screen by screen
RSAT alternative · migration
RSAT is gone after 15 May 2027. Your regression suite shouldn't die with it.
Microsoft has marked the Regression Suite Automation Tool for deprecation, with support ending 15 May 2027. Teams that invested in it already know the maintenance story: RSAT decouples test parameters from test steps and stores them in Excel data files, drives F&O through Task recorder recordings and the acceptance test library, and depends on a browser driver that has to be kept in step with the browser. Practitioners describe suites that broke more often than they caught anything, and days lost to a driver version bump.
Moving off it is a chance to stop renting your test assets. Record the same processes once, get Playwright tests in your own repository plus a visual baseline for every screen, and run the whole thing on your own infrastructure. Nothing in the resulting suite depends on a Microsoft tool, a vendor cloud, or a model being consistent at replay time.
10.0.49 preview 27 July 2026, autoupdate to production 2 October 2026 — the next window to have a pack ready for.
# baseline captured before the update
$ lastest run --suite order-to-cash
42 processes · 318 screens · 11m 04s
✓ 305 screens identical
! 13 screens changed
# evidence written locally
$ ls ./artifacts/10.0.49/
before/ after/ diff/ run.json
✓ nothing left your network
What you get, and what you keep
Dynamics 365 release testing — common questions
- How long do I actually get to test a Dynamics 365 F&O service update?
- About a week, and Microsoft's own documentation is inconsistent about exactly how long. The release-schedule page describes a five-business-day gap between the sandbox update and the production update; the One Version service updates FAQ describes seven calendar days. Either way it is one week, it is unattended, and it repeats for every service update. Lastest is built to fit that window: the regression pack you recorded before the update replays overnight against the updated UAT sandbox and hands you a screen-by-screen diff in the morning.
- Can I just pause the update until we have time to test?
- Only once, and only so far. Microsoft ships up to four service updates a year and you must take at least two of them. Since 19 February 2024 you cannot pause more than one consecutive update, regardless of industry or business schedule, and both autoupdate windows have to be paused explicitly in Lifecycle Services. Pausing buys you one cycle of breathing room, not a testing strategy.
- What happens if a service update breaks something in production?
- There is no rollback once a service update has been applied to a production environment. Your options are a support case, a hotfix, or a configuration workaround while users are already blocked. That asymmetry is why the sandbox week matters: it is the only point in the cycle where finding a problem is cheap.
- Is Lastest an RSAT alternative?
- Yes, and it is a migration path rather than a like-for-like replacement. Microsoft has marked the Regression Suite Automation Tool for deprecation, with support ending 15 May 2027. RSAT drives F&O through Task recorder recordings and keeps its test parameters in separate Excel data files. Lastest records the same business processes through the browser and emits plain Playwright tests into your own git repository, plus a visual baseline for every screen the process touches. Nothing depends on a tool Microsoft is retiring.
- Does this work for Customer Engagement apps as well as Finance and Operations?
- Yes, and the page treats them separately because the release mechanics are different. F&O is governed by One Version service updates, LCS environments and autoupdate windows. CE is governed by release waves, the release channel you pick in the Power Platform admin center, and Dataverse solution deployments. Lastest drives whatever renders in the browser, so model-driven apps, custom pages and F&O workspaces are all recorded and replayed the same way.
- Can I postpone a Customer Engagement release wave?
- No. Release waves are mandatory and cannot be postponed. What you can choose is the release channel — monthly, which brings features earlier and in smaller increments, or semi-annual, which batches them. Early access as a separate opt-in programme has been deprecated in favour of the monthly channel, and there is no longer a preview sign-up to complete. Your control is over pace, not over whether the wave arrives.
- Does Lastest need access to our Dynamics 365 environment from the internet?
- No. Lastest is self-hosted. It runs on your infrastructure, inside your network, against your UAT sandbox or your Dataverse environment, and the screenshots never leave your perimeter. That matters when the screens under test contain customer, financial or pricing data that a security review will not let you send to a vendor cloud.
- Is there AI in the execution path?
- No. AI is used once, at authoring time, to turn a recorded business process into a Playwright test. Replay is deterministic: the same test produces the same run, the same screenshots and the same diff every time. Nothing about the evidence you hand to a release board depends on a model behaving the same way twice.
Map your next
forced release date
to a regression scope.
Thirty minutes. We take your next service update or release wave, work out which processes actually need to be in the pack, and show you what the evidence looks like when it comes back. Dynamics 365 is one of five platforms covered on the regression testing for vendor-forced releases pillar.
Book a release-readiness review →# F&O — One Version service updates
10.0.49 preview 27 Jul 2026
10.0.49 autoupdate → prod 2 Oct 2026
# the gap you are handed
sandbox first, production a week later
unattended · no rollback after apply
# CE — release waves
wave 1 Apr → Sep
wave 2 Oct → Mar mandatory
✓ have a pack ready before each one