Copilot4DevOps Announces V10, Its Most Comprehensive AI Release to Date

TORONTO, Canada, [September 2026] — Copilot4DevOps today released Version 10. The release changes two things: what the AI does when a request is incomplete, and what it hands back when the work is done. Four modules now ask clarifying questions before generating anything, and Chat answers with interactive widgets rather than paragraphs of text.

Chat, Elicit, Generate Documents and the Mockup module each present a clarifying questions card before they produce output, surfacing gaps in a request instead of filling them with assumptions. In Elicit and Mockup the questions arrive in sequenced passes with a progress indicator, and each accepts a short text answer, a single selection or multiple selections. In Elicit, the suggested options are drawn from the anchor work item, the parent work item details, the elicitation setup, and any custom instructions rather than from a static list. Any question can be skipped, the whole card can be skipped at once, and a running count of answered questions stays visible.

Chat’s answers have changed shape entirely. Seventeen widget types, spanning thirty-nine distinct layouts, now render inline whenever Chat creates, updates, finds or reads something in Azure DevOps.

The set covers work item creation and updates, test case steps, query results, hierarchy trees, linked items, board snapshots, iteration summaries, commit history, pipeline status, file diffs, wiki reads and activity feeds. Update receipts show each field before and after, flag any item that failed rather than dropping it quietly, and offer an undo action on single item changes. Eight repository write actions and four wiki write actions return their own confirmations, and higher risk or destructive actions require explicit confirmation before they are carried out.

A full backlog hierarchy in one run

Elicit can now generate an entire multi-level work item hierarchy from a single source item. An Epic, Feature and User Story chain is produced in one run rather than requiring a separate pass per level with manual re-parenting afterwards. Teams can use their project’s native process template chain or build a custom hierarchy with reorderable levels and configurable link types, then name, save and share it at personal, project or organization scope. Auto mode runs every level in one continuous pass, while Manual mode stops at each level for review before the next is drafted, carrying any edits into the context used below.

Results appear as an expandable tree, with safeguards that prevent deselecting a parent from silently publishing orphaned children. On publish, Copilot4DevOps creates the items in Azure DevOps, applies the hierarchy’s link types top down, and returns each new work item ID with a direct link. Partial failures are reported without leaving dangling links behind.

Title generation also stopped producing a fixed number of results. It now continues until requirement coverage is sufficiently addressed, so the count reflects the actual scope of the requirement. Every AI-generated title carries a relevancy score from 0 to 10 that the candidate list can be sorted by, and those scores persist with the run.

Documents that keep their history

Generate Documents, previously the SOP/Document Generator, gained a full version history. Applied edits increment a minor version automatically and record the instruction that caused them, while major versions can be created with an optional release note. Any two versions can be compared side by side, and any version can be restored, which creates a new version rather than erasing what came between. Every proposed AI change appears as a marked diff that the user applies or discards, never applied silently.

Documents can be published straight to an Azure DevOps Wiki with heading hierarchy, numbering, lists and tables preserved as Wiki Markdown, attached to any work item as Word or PDF, or downloaded as Word, PDF or, new in this release, Markdown. A prompt template picker supplies the document type, its section order and per-section guidance, and a chat panel alongside the draft refines it by instruction rather than by regenerating from scratch.

Mockups you can watch being built

The Mockup module added Theme, Fidelity and Platform controls above the prompt, so the look and target form factor no longer must be described in the prompt text. Variants can be generated against the same brief to compare directions without losing the original, and a viewport switcher previews any design at desktop, tablet and mobile sizes.

Generation itself is now visible: the page assembles in the canvas top down as it is produced rather than sitting behind a progress bar. Mockups export as HTML without scripts, design tokens in JSON, CSS variables or a Tailwind theme, alongside the existing file and image options.

Also in Version 10

• Field-level AI editing. Suggest Edits, Length, Polish and Reading Level, previously available only across a whole work item, now apply to an individual field such as Description. Version history moved into the work item context menu, with side-by-side comparison and a running count of removed and added content.
• AI-generated email templates. In Outreach Email Configuration, a template can be described in plain language and generated complete with name, subject, body and the relevant placeholder variables. Images can now be pasted or dragged straight into a template body.
• Fairer quality scoring in Analyze. Criteria that do not apply to a given work item are now treated as not applicable rather than scored low, so the overall percentage reflects only what was evaluated. The interface has been redesigned alongside it.
• Chat thread management. Threads can be sorted, filtered by status and last activity, and grouped across both personal and shared views, with bulk selection and a reset to defaults. Each reply now shows the agent and skill that produced it.
• Skill referencing in Chat. Skills can be referenced directly in a conversation from the commands menu or by typing a tilde for an inline picker. If an agent is already selected both apply; otherwise the skill runs with the default Worker agent.
• Persistent history in Elicit. Past elicitation runs are kept as threads in a searchable history rail and restore in full, including candidates, work item type and source context. Threads can be renamed, pinned and deleted.

Four modules were also renamed for clarity. The SOP and Document Generator is now Generate Documents, the Q&A Assistant is now Topic Explorer, Generate is now Generate Artifacts, and Impact Assessment is now Impact Analysis. The user interface has been refreshed across the application.