AI-created D365 F&O tickets in Azure DevOps: how it works

AI can turn a business request written in natural language into a structured ticket (user story, bug, task) in Azure DevOps: it rephrases the need, proposes acceptance criteria, identifies the D365 F&O objects involved, then creates the work item. Technically there are two routes: the Azure DevOps Work Items REST API, or Microsoft's Azure DevOps MCP server, which gives an AI agent read and write tools on work items.

The point is not to remove human validation: it is to get complete tickets from the start, which the ERP lead or IT lead validates before development.

Microsoft offers an Azure DevOps MCP server, as a hosted remote version (recommended, Microsoft Entra ID authentication) or a local one. Its work item tools let an agent read an item, run a WIQL query, list the backlog and, for writes, create a work item, update it or add child items. A configuration header restricts the server to read-only, and you can limit which toolsets are exposed.

Good practice: start read-only (the agent analyses the backlog and proposes), then only open writes to a defined scope and project.

Whether AI or a person writes it, a usable ticket contains:

• the business need and the process involved (module, form, report);
• the expected behaviour and verifiable acceptance criteria;
• sample data and the company concerned;
• constraints: security (roles), performance, affected integrations;
• the likely type of development (table extension, Chain of Command, report, data entity, batch…);
• links to existing tickets or code.

With Azure DevOps, a branch policy can then require a pull request to be linked to a work item before merging: ticket → code traceability is automatic.

• Letting an agent create tickets without validation: duplicates and vague requests multiply.
• Giving the agent write access to the whole Azure DevOps organisation.
• Accepting unverifiable acceptance criteria ("must be fast").
• Not linking the ticket to the delivered code: history is lost.
• Putting sensitive production data in the ticket description.

At Skalp AI, chat and needs analysis are free. When the smart contract is activated, Skalp itself creates the work item in the customer's Azure DevOps project: Task type, title carrying the TK-<year>-<number> reference, description = HTML specification structured as F-xx (forms), REP-xx (reports), DE-xx (data entities) and PARAM-xx (parameters). The code is then delivered as a pull request, validated by your technical architect. See Azure DevOps D365 integration, the bigger picture in building X++ without a consultant in 2026, and the Basic, Pro and Max plans (1 credit, ≈ €400 on Basic, €200 on Pro, consumed only when the code is generated).

• Microsoft Learn, "Set up the remote Azure DevOps MCP Server", updated 21 Sep 2026, accessed 18 Oct 2026.
• GitHub, Microsoft, "Azure DevOps MCP Server (README)", accessed 11 Oct 2026.
• Microsoft Learn, "Set and manage branch policies", updated 15 Jul 2026, accessed 18 Oct 2026.
• Microsoft Learn, "Work Items - Create (REST API)", accessed 18 Oct 2026.