ProcessMind 2026.11

ProcessMind 2026.11 connects the documents your organization already has to the models you maintain. Attach a Visio diagram, a PDF procedure, a Word policy or an email thread, review what the documents propose, and decide what enters the model. The release also lets you define model quality per architecture level, rebuilds the welcome page around what needs you today, and makes the app start with a fraction of the JavaScript it used to.

From documents to process updates

• Propose Updates from Documents. Open a process, choose Propose Updates from Documents from the process menu, and ProcessMind merges the documents already extracted for that process into one proposal. It compares the documents with the current model and groups the result into New elements, Renamed elements, New connections, New lanes, Documentation, RACI, Roles, and Conflicts to review.
• Two ways to analyze. Analyze documents uses the deterministic extraction; Analyze with AI drafts also drafts documentation text and RACI suggestions and counts against your AI plan.
• Nothing changes until you decide. Import into canvas uses the standard BPMN import: it fills an empty canvas or replaces an existing model after confirmation. You can also Download BPMN to review the proposal elsewhere. RACI and role suggestions stay in the review; the BPMN import applies the model and element documentation.
• More documents count as sources. Attachments now include Visio drawings (.vsdx, .vssx, .vstx), draw.io files, PDFs, Word, Excel and PowerPoint (including macro-enabled and OpenDocument variants), email (.eml, .mbox), images (text is read with OCR), audio recordings and videos (transcribed), Markdown, text, CSV, JSON and ZIP archives. Legacy binary formats such as .vsd, .doc, .xls, .ppt, .msg, .7z and .rar are not converted.
• Everything stays searchable. Extraction runs in the background when you upload a document. The artifact moves through Waiting to process, Processing and Searchable, and its content becomes available to Global Search and to the AI assistant.

Compare versions and processes

• Compare a version with the draft or another version. Open a saved version and choose Compare with draft, or start from the draft with Compare Changes. Both sides are picked as Original and Changed, the dialog groups the graph changes, and On canvas highlights added, modified and removed elements. Documentation changes are listed too and jump to the changed section.
• Approve or reject what you reviewed. Approve records a review decision without changing either version; Reject records a rejection and keeps the group visible. Undo reverses a decision, and Show reviewed groups brings approved groups back.
• Compare two processes. Select exactly two processes in the catalog (Ctrl/Cmd-click) and choose Compare in the bulk bar. Equivalent activities are matched by name, documentation is compared across both, and a swap button flips the sides. Mappings and model constants are not part of the comparison. See Versioning for the full workflow.

Model quality, defined per architecture level

• Choose the checks that count. In Settings → Process Architecture, open a level and use Model Quality → Configure. Model Quality Checks lets you switch individual checks on or off, with a Reset all shortcut that re-enables every check and clears the tuning, and a running X / Y enabled count per category.
• Tune how strict a check is. Checks that support it offer relaxed, standard, strict or strictest, and every option states the rule the check itself applies — how many outgoing paths a split may have, how long a name may be, or how much routing complexity is tolerated. An untuned check runs its standard option. Gateway naming is required wherever a decision split exists: name the gateway or label every outgoing path.
• The score means what your level requires. Disabled checks are excluded from the score, so 100% always means everything this level asks for. They are also hidden in the process’s Model Quality panel unless you tick Show disabled criteria; revealed checks show a Disabled label and the reason.
• Existing processes follow the level. Saving a level with changed quality settings re-scores the processes on that level, so the catalog stays consistent.

A clearer default architecture, with repair

• Six levels by default. New and existing workspaces start from Value Stream → Process → Subprocess, plus Folder for organization, and Activity and VSM Element Level for documenting model elements. Value streams can contain processes; folders sit at the top level or inside another folder; Activity and VSM Element Level are never process types.
• Your structure is kept. Existing processes, names and custom levels survive; the starter set is a default, not a reset.
• Clean up when the hierarchy drifts. If an allowed-child list points at a deleted level, uses an element level as a child, or an element level carries a code prefix, Settings → Process Architecture shows Architecture issues with a Clean up action. The catalog also flags items that sit in an invalid place without hiding them.

A welcome home that tells you what needs you

• A new start page. The welcome page now opens with the next thing that needs you: a review waiting for your decision, your first dataset, roles to define, a process to publish. Not now hides a suggestion for a week; a review waiting on you always stays.
• My Work. One ranked list of up to ten items involving you, with a short reason on every row: you own it, you follow it, you were mentioned, or you hold a RACI role such as Accountable. Mentions have their own section above the list.
• The rest of the page at a glance. At a glance shows workspace numbers; Bookmarks & Recent keeps your last processes across Recent, Owned and Bookmarked tabs; Process landscape jumps into the catalog; Set up your workspace shows only open checklist items; Company & portal and Quick links round it out.
• Onboarding starts you on the right path. The final step preselects the path that fits your goal and still lets you choose Process Modeling, Process Mining, or Demo. Invited members get a shorter Welcome → details → Finish flow, and an interrupted run resumes where you left off.
• Portal settings. Environment → Portal now configures the viewer experience: Headline, Welcome text (Markdown) with Preview, Support email and Support link, plus Allow viewers to comment and Allow viewers to export processes. Branding is inherited from the workspace theme and company profile.
• Portal viewers get their own version. They see the configured headline and welcome text, a Need help? contact line, their conversations, and the published processes shared with them.

Collaboration that follows the work

• Follow a process. Choose Follow this process in the process menu, or Follow in the catalog row menu. The menu shows the follower count, and commenting on or reviewing a process can follow it automatically.
• See everything you follow. Settings → User settings → Notifications → Processes I follow lists your followed processes, with Unfollow on each row.
• Choose your email. The Notifications page groups your email into Workspace & account, Product & analysis and Digest, with Enable All / Disable All. Mentions & replies has its own switch.
• Tune process mail. Under Process notifications, choose Instant email, Weekly digest or Muted, then switch the events in Process status, Model versions, Quality and People. If delivery to your address was suppressed, Resume delivery brings it back.

A lighter, faster start

• Much less JavaScript before you can work. Route pages, admin settings, the super-admin console and the AI assistant load only when you open them. In our measurements the app now starts with roughly a quarter of the JavaScript it needed before.
• Something to look at while it loads. A branded loading screen appears immediately instead of a blank page, and the session calls start while the bundle is still parsing.
• No dead end after a deploy. A tab left open across a release recovers from a missing code chunk with one guarded reload; a route that cannot load shows a retry screen instead of an endless spinner.
• Data-friendly on slow connections. Idle prefetch and image loading step back on save-data, 2G/3G and other constrained links.
• Less tracking in the app. The app no longer loads Google Analytics or Google Ads. Microsoft Clarity remains, and only after you consent to analytics cookies.

API and MCP

• New MCP tool: propose_process_updates. On Enterprise, AI tooling connected through the MCP server can now turn a process’s extracted documents into one reviewable proposal of model, documentation, RACI and role updates. The tool is read-only: nothing is applied until you use the existing apply tools.
• MCP server 1.2.0. The server adds this tool alongside the existing read, AI, write and memory tools.

Also improved

• Delete a process with a clear confirmation. Administrators can choose Delete; the Delete process dialog states what will be removed, child items move to the parent level, batch delete skips locked processes, and deleting a folder keeps its contents.
• Bulk BPMN import. The catalog’s Import BPMN files dialog accepts up to 100 .bpmn files at once, or a .zip archive on Process Architecture and higher, and keeps the folder paths from the archive.
• Document previews inline. PDFs preview in the artifact panel; Word, Excel and PowerPoint preview through the Office viewer up to 4 MB, with a clear fallback for larger files.
• Charts explain themselves. Each chart has a Chart description, and chart settings now open in Basic with an Advanced option for the full set.
• The AI Assistant can propose role edits and process deletion on the process page. Both still arrive as a diff you review and confirm.
• A visible answer trail. The AI Assistant shows a numbered tool timeline for each answer and accepts feedback with an optional note.
• Global Search and the AI assistant now cover a wider set of attached documents, including Visio structure, email threads, audio transcripts and archive contents.
• Artifact cards show a clearer preprocessing state and explain why a file is not searchable.
• A one-time AI overview on Process Health. During a trial, the Process Health dashboard adds an AI overview of the current CURES snapshot: the strongest and weakest signals, improvement opportunities and what to do next. It is generated once per trial organization.

Try everything in this release with a free trial. If something does not work the way you expect, tell us.