ERP Roundup: Microsoft's Agentic ERP Roadmap Leaves Business Central Out, and Two Surveys Say Verification Is the Bottleneck
A quiet week for Business Central releases, a loud one around it. Microsoft used Directions for Enterprise in Copenhagen to set out what replaces the release plans, and two testing surveys a day apart reached the same conclusion from opposite ends of the trade. No named ERP failure, lawsuit or securities filing surfaced, so this is a four-story roundup rather than five.
1. Microsoft publishes an agentic ERP roadmap, and Business Central is not in it
On 23 September, Sameer Verma, Microsoft's chief product officer for Dynamics 365 AI ERP, published a roadmap of ERP capabilities "new and coming soon through March of 2027". It is the first real look at the AI at Work roadmap replacing the twice-yearly release plans, covered here when the wave model was retired.
The list is mostly agents: an Account Reconciliation Agent extended to clear exceptions in bulk, a Finance Agent with skills for reconciliation, variance analysis and collections, Procurement Agent impact analysis in public preview, and a Scheduling Operations Agent for Field Service. Two foundation items are marked generally available: the Dynamics 365 ERP MCP server and an ERP plugin for Copilot Cowork.
The products named are Finance, Supply Chain Management, Project Operations, Commerce, Field Service, Human Resources and Sustainability. Business Central does not appear. That is an observation about one post, not Microsoft's intentions, and Business Central has its own agent work shipping in version 29. But partners told a roadmap would replace the release plan have now seen it, and their product was not on it. Monday's piece on scoping the BC29 regression pass without a release plan assumed the planning document was simply gone; its replacement so far addresses the enterprise apps.
What it means for testing
If you scope regression passes from a published plan, you need a substitute for Business Central: the version 29 preview documentation on Learn, your AppSource dependency list and the cumulative update notes. Build it once and keep it, because an undated roadmap gives you no date to test against. The MCP server going GA matters more than it reads: it is a route into ERP data no screen-driven UAT script will exercise, so the question is not whether it works but what it can reach.
2. Directions for Enterprise: Microsoft pushes partners towards Activate
At Directions for Enterprise in Copenhagen, Maitame Munoz, partner marketing lead for Microsoft Business Applications, told partners to benchmark their AI readiness with an assessment Microsoft wants practice leads to take, as MSDynamicsWorld reported on 25 September. The concrete ask underneath is Dynamics 365 Activate, previously Project Mia, described in that coverage as "an agent-driven implementation experience designed to accelerate every D365 journey with lower risk and repeatable outcomes". Its agents cover requirements analysis, configuration, data migration, integration design, security roles, test planning, cutover, training, feature evaluation and process optimisation.
Check the primary source before planning around it. Microsoft's Learn documentation for Activate, last updated 9 September, says the product is in preview, not for production use, and currently supports migration from Salesforce to Dynamics 365 only. Testing appears as one word inside its execution phase, with nothing on who owns validation or sign-off. The agent list is intent; the shipping product is a Salesforce migration preview.
What it means for testing
A test planning agent is not a testing agent, and the difference is the whole job: an agent can propose what to test, but cannot be the person who accepts the result. If Activate reaches Business Central work, the risk is not a bad generated plan but a plausible one arriving with no named business owner and signed off because it looks complete. Treat generated coverage as a first draft to argue with, check it against the processes your client runs, and keep the sign-off trail human. The configuration and migration agents raise the same question: each produces output needing acceptance testing nobody scoped.
3. Two surveys in two days: verification is the bottleneck
On 23 September Qodo published its 2026 State of AI Code Quality Report, a Censuswide survey of 500 US developers and 300 engineering leaders fielded 7 to 14 August. Reviewing and validating AI-generated code was the top delivery bottleneck for 26% of developers and 26% of leaders, and 48% of leaders called it their largest quality and governance gap. The pair that should stop a consultant: 90% of leaders said they are confident reporting AI's impact to executives or board, while 45% have evidence tracing AI activity to code changes.
A day later, DeviQA published a survey of 4,000 QA professionals fielded 15 August to 15 September, all screened for testing AI-assisted software in the previous six months. 65% said features reach testing sooner. Then: 64% said more features arrive at once, 55% have watched QA queues grow, 52% reported longer test, fix and retest cycles, and 47% had hit AI-assisted functionality that worked as asked and broke something elsewhere. The most-cited remedy, at 77%, was clear acceptance criteria.
Both are vendor-commissioned surveys of self-reported experience, and both vendors sell into the problem they describe. Trust the direction, not the decimals.
What it means for testing
Two samples, one of developers and one of testers, landed on the same shape: output went up, verification did not, and failures show up sideways in components nobody changed. That describes an AL extension on a client tenant, and it is the argument for a regression suite that persists between passes, because the 47% figure is only caught by tests covering what was not touched. If a developer cannot tell you which processes a change touches, that is a test scope gap, not a documentation complaint for later.
4. CSP growth margins get an eligibility API
Microsoft told CSP partners on 23 September that distributors and direct bill partners can now validate customer eligibility for growth margins through APIs, with discovery and validation also in the Partner Center Pricing workspace. Transactions go generally available on 1 October 2026.
What it means for testing
Commercial, not functional, but it lands on the same desk. If your billing or quoting tooling reads Partner Center, a new eligibility call changes a system whose output clients see. That is the practical half of pricing UAT work properly: the margin you quote and the margin you are eligible for should match before 1 October, not after the first invoice.
The through-line: the plan is agentic, the accountability is not
Taken together, the week describes a supply chain of generated work with one human checkpoint at the end. Microsoft's roadmap adds agents that reconcile, procure and schedule. Activate adds agents that gather requirements, migrate data and plan the tests. Two surveys report what happens downstream: more arriving faster, verification unable to keep pace, 45% of leaders able to trace what the machines changed. None of it removes the need for a named person to say a process is acceptable to the business. It puts more in front of them with thinner evidence of where it came from, which is the argument in autonomous ERP: who tests the machine.
Free ERP UAT checklist (Excel)
A structured workbook covering the full UAT cycle: pre-UAT preparation tasks, test execution tracking, an issue log with severity guide, and post-UAT wrap-up. Built from real ERP implementations.
No spam. We will occasionally send UAT and ERP implementation resources. Unsubscribe any time.
Sources and further reading
- Microsoft Dynamics 365 Blog: Build the future of agentic ERP with new Microsoft Dynamics 365 capabilities, Sameer Verma (23 September 2026)
- MSDynamicsWorld: Directions for Enterprise 2026, it's time for partners to benchmark their AI and agent alignment, Microsoft says (25 September 2026)
- Microsoft Learn: Dynamics 365 Activate overview (preview), page updated 9 September 2026
- Qodo: 2026 State of AI Code Quality Report, survey conducted by Censuswide (23 September 2026)
- DeviQA: 2026 report on the impact of AI-assisted development on software testing, 4,000 QA professionals surveyed (24 September 2026)
- Microsoft Partner Center: September 2026 announcements, including CSP growth margin eligibility APIs (23 September 2026)
More Generated Work, Same Test Window
LogicHive gives ERP implementation teams a regression suite that survives between releases: structured test cases, named assignment, linked defect tracking, and sign-off built from real execution data. When an agent proposes a plan or a change lands without notice, running a pass is a scheduling decision rather than a project. The free plan runs a real project end to end.
Written by Sharif George
LogicHive Frontman
Sharif is the frontman for LogicHive. He writes the ERP news roundups and the Business Central release coverage, following each Microsoft, SAP and Oracle change through to what it means for regression scope and acceptance testing on live implementation projects.
View all articles by Sharif GeorgeRelated reading
ERP & Enterprise
Business Central 29 Ships in October Without a Release Plan: How to Scope the Regression Pass
BC29 is generally available in the first week of October, and Microsoft confirmed there would be no 2026 release wave 2 announcement. The shipping cadence has not changed; the planning document partners scoped their upgrade regression pass from has gone. How to rebuild the scope, and why your preview sandbox has a deletion date.
21 September 2026 · 6 min read
ERP & Enterprise
Business Central 29 Moves Extension Fields Into the Base Table: The Change With No Screens to Test
From version 29, the fields your extensions add to a table are stored in that table rather than in a companion table beside it. No screen changes, so a feature-led test pass finds nothing — while query plans, index behaviour and the field capacity every extension on that table shares all change. What to put in the October regression pass.
28 September 2026 · 6 min read
ERP & Enterprise
Business Central's Record-and-Replay Test Tool Reached GA This Week. It Is Not UAT.
Microsoft moved the Business Central page scripting tool from preview to general availability with version 29, thirty months after the preview opened, adding multi-row grid selection and dialog-text validation — neither of which the reference documentation describes yet. What a recording actually proves, what it is documented not to reach, and why a replay pipeline is a regression asset rather than an acceptance one.
5 October 2026 · 6 min read