Version 2026.318.2
Release Date: November 15, 2026
This release includes the customer-facing changes delivered since version 2026.254.5.
Availability: Features marked Cloud are available in cloud environments and are not included in the mindzieStudio Enterprise or mindzieDesktop packages. All other items apply where the corresponding workspace is enabled.
New Features
Process Mining and Process Intelligence Modules
The workspace modules are now named Process Mining and Process Intelligence. Process Intelligence is no longer a pre-release feature and is the default module for new tenants.
Guided Project Setup (Cloud)
New AI Process Intelligence projects walk you from raw data to your first analysis with a step bar and one clear next step at a time.
- Choose your path - after uploading, choose to be guided or to set the project up yourself.
- Understanding your data - the project reads the staged tables, says what it found, records the question you want answered, and checks whether the data can answer it.
- Data and analysis interviews - a short interview confirms how the tables connect and what the analysis should focus on.
- Live progress - each build step shows its progress, and a finished setup opens on the process map.
- Size warnings - the files screen warns when the tables are too big for the project's size.
Upload Files to a Project (Cloud)
Table files can be uploaded directly on the Files tab. Each file is recognized by its content as it arrives and sorted into the right place, so event tables, master data, and settings files no longer need to be prepared by hand first.
- Fast, checked uploads - large files are sent in compressed parts and checked on arrival.
- History and snapshot files - version-history exports and daily snapshot files are converted into the steps of each record, including snapshots uploaded one file at a time.
- Change tables - a table of changes next to its record table becomes the steps of that record.
- Status at a glance - the Files tab lists every upload by status, with Remove and Convert again, and shows a picture of the tables and how much storage they use.
- Build from your uploads - Build the OCEL takes the uploaded tables straight into the schema check.
- Space checked first - a conversion the project's disk cannot hold is refused before it starts, with the reason.
Schema Studio (Cloud)
Schema Studio, now in the project's left menu, shows how your tables become an object-centric event log before anything is built.
- Check the schema is the default step before a build.
- Split a table by one of its columns.
- Join explanations - when a join matches less than all rows, Schema Studio says why.
- Repeated events - repeated events are measured across every row, you choose how they are handled, and the forecast is shown beside what the build produced.
- Empty object types - you are warned before building when an object type has no events.
Change a Built Project (Cloud)
A project that is already built can be changed without starting again. Use Change the schema to adjust the object types and events and rebuild, or add more table files and rebuild with them.
Database Connections, Including Snowflake (Cloud)
The data connection screens have been redesigned. Test connection opens the database for real, the table chooser reads the database's own catalog, and Snowflake is supported as a connection with key-pair or password sign-in, so its tables can be pulled into a project.
AI Reports (Cloud)
- Ask in your own words - start a report from one of the ready-made cards or describe what you want; reports publish themselves when they are finished.
- Open without starting - published reports open and can be shared without starting the project.
- Your language - reports are written in the language you asked in.
- Conformance against your model - an AI conformance report can compare the log with your own BPMN model.
- Workload over time - reports and AI jobs chart the workload over time and name busy periods, including a ready-made "Where does work wait?" report.
- Edit and rename - rename a report from the Reports page, and select a section by clicking it in edit mode.
Create Rules, Filters, and Enrichments in Words (Cloud)
Rules, filters, enrichments, and costs can be added by hand or by describing them in your own words. Free-text values are read and labelled, and Try warns when a column leaves values unsorted.
When Is a Case Finished? (Cloud)
The Data screen suggests the activities that end a case and asks you to confirm them. Every analysis screen then uses that answer, so running cases are no longer treated as finished.
Process Map (Cloud)
- Headline - a one-line summary above the map names the main path share, the slowest handoff, and the most repeated step. Each fact can be clicked.
- Click panel - click an activity or a connection to see the steps before and after it with their waits, who performs the step, and actions to follow up.
- Show measures - the map can show frequency or waiting time, and the slowest connections are colored.
- Faster first map - the map is drawn with the page, and the first map after a build no longer waits for long calculations.
- Clearer labels - long activity names wrap onto two lines, the animation control is an Animate button, and only real amounts are offered for totals.
BPMN Map and Train Map (Cloud)
The BPMN Map opens on a full-height canvas, readable from the start event, with a minimap to find your way when you zoom in. The lane picker lists only useful attributes, best first, with a No lanes choice. Model quality is shown as short labels, and a validation problem is a calm warning with Download anyway. On the Train Map, every element explains itself on hover, a legend sits at the top, and each measured finding is a card with Show cases and Open on map. On a log with a single object type, the Train Map notes that there are no handoffs between types to show and offers Open the Process Map.
Performance Analysis (Cloud)
- Biggest wait first - the largest wait on the main path is shown above the tiles, with actions to see the cases, the map, or a simulation.
- Cycle time by month - a monthly chart, with trend lines on the Cycle time and Cases tiles. Click a month to see only the cases that finished in it.
- Filter from the histogram - click a bar of the cycle-time histogram, or drag across several bars, to keep only those cases.
- Target on the tile - set the cycle-time target directly on the Cycle time tile, which also shows the 90th and 95th percentile.
- Handoff kinds - every handoff says what kind of wait it is (Internal, Customer, Partner, or System), and you can change it.
- One Bottlenecks table - group the bottlenecks by handoff or by activity.
Conformance and Rules & KPIs (Cloud)
- Use your own model or a variant - check the log against your own BPMN model, or use the most common variant (or any of the top ten) as the target in one click.
- Time cost of deviating - a new tile shows how much time deviations cost.
- Fairer scores - running cases are not counted as deviations, and no conformance score is shown against a model discovered from the same log.
- Rules & KPIs - each rule shows one rate bar with its counts, worst rule first, together with its category, owner, broken cases, and what uses it.
Costs and Automations (Cloud)
Costs can be edited in place, suggested cost templates appear next to the quantity they would price, and each cost shows who made it and when. Automations lists the opportunities with what and why for each candidate; steps done by customers, partners, or systems are no longer suggested, and a start-time column is read so hands-on time is measured apart from waiting.
Task Mining (Cloud)
The Task Mining Agent records desktop work, and its recordings can be uploaded to an AI Process Intelligence project and read as AI reports. Recordings can be joined to the process by the case ID typed during recording, so desktop tasks appear alongside the process events.
AI Settings - mindzie AI, Your Own Models, or AI Off
AI settings (previously Copilot settings) now offer one clear choice: mindzie AI, your own models, or AI off. Provider types show their proper names, the provider dialog helps fill itself in, you can name a model for agent work, and a tenant can remove its last provider. When an administrator turns AI off, screens say so instead of failing. In Offline Mode, your own model can be reached through a list of private hosts.
AI Usage and Credits (Cloud)
mindzie AI usage is measured in credits. A credits meter in the header shows what is left, and the Usage tab shows usage this month, what used it, and who may do what. Administrators can buy credits by card and turn on automatic top-up, and usage emails arrive in the tenant's language. AI that runs on your own models is never charged; the Usage tab shows its tokens instead.
Larger and Automatic Project Sizes (Cloud)
Three larger AI Process Intelligence project sizes are available: Extra large (4 vCPU, 64 GB), Huge (6 vCPU, 128 GB), and Gigantic (8 vCPU, 256 GB). With Automatic size, a project moves itself to a larger size when its data needs it, keeps its files, and waits for you before continuing.
Process Intelligence on mindzieDesktop and Enterprise
mindzieDesktop can run AI Process Intelligence projects through Docker Desktop, and setup checks what a new Windows computer needs. Enterprise servers without internet access can install the project image from the Enterprise package; a daily check offers newer images and never moves backwards. Uninstalling mindzieDesktop keeps your data unless you ask to remove it.
Command Line and Hosted MCP Server
A single mz command-line tool now covers AI Process Intelligence projects, BPM Studio, tenants, and users, and can run as a local MCP server. A hosted MCP endpoint lets AI agents manage tenants with one address, shown next to your API keys.
Improvements
Reading Real Exports (Cloud)
- Choose the encoding of a file that is not UTF-8, with a preview, including Windows and Latin-1 exports.
- Report titles and total rows in an export are found and left out, and the screen says so.
- A file with no column-name row is read from its first line, with a warning.
- A decimal comma in a semicolon-separated file no longer makes the separator ambiguous, and a column of durations is no longer mistaken for a time column.
- SAP-style dates (YYYYMMDD), mixed time-zone offsets, and date-only events are read correctly.
- References between tables are still joined when an export has damaged some of them, and master data is treated as objects while code lists stay lookups.
- Object-centric (OCEL 2.0) imports keep object-to-object relations and attribute history, and a project's log can be exported as OCEL 2.0 JSON.
- Builds are faster on common date formats.
Clearer Figures on Every Screen (Cloud)
A How this is calculated drawer explains each figure. Labels appear only where a figure is not plain data, technical timings are kept behind a Diagnostics switch that is off by default, and a filtered map or variants list says how many of the cases it shows. Screens use plain business words throughout and meet WCAG AA color contrast.
Variants, Predictions, and Working Calendar (Cloud)
The main-path share names its definition, and running variants can be hidden with Completed only. Parallel variants on large logs are calculated much faster and with far less memory. Predictions offer only attributes that can be predicted, and a model can no longer learn from columns calculated from its own target. The working calendar's Measuring as at date is never in the future, and the calendar shows one working-time figure with its unit.
Projects Page (Cloud)
A new project asks only for a name, with the size under More options. Start keeps you on the project list and opens the project in a new tab, a ready project is shown as ready, and recently used projects appear in Pick up where you left off. Only owners and administrators may rename, share, resize, delete, or restore a project, and a project always keeps one owner. Backups, restore, and project-file import are available.
Project Status and Reliability (Cloud)
- A busy project that is slow to answer stays Running and says how long it has not answered.
- A project that stops itself after being idle reads Stopped instead of Failed to start, and running out of memory names the project size.
- A start that does not come up is retried once, then reported in one plain sentence.
- Large projects no longer run out of memory on AI questions, wide logs, or start-up calculations.
- Saving something made with Create with AI no longer makes a small project stop answering, and Create with AI also works through the API.
BPM Studio
- AI-generated diagrams must pass the BPMN checker before they become project diagrams, and are checked against their source.
- Importing and exporting BPMN keeps collaborations, pools, lanes, data artifacts, identifiers, and every element's content.
- Data associations, black-box pools and their message flows, and artifacts inside sub-processes are drawn and kept.
- Generated layouts route connections more cleanly and keep labels inside their own sub-process, lane, and pool.
- Turkish wording for the industry templates.
AI Jobs
A failed AI job says why, with Retry on its row, and a finished job stays finished after a page reload. A job shows one status word, Queued or Running, on both the AI Jobs and Reports screens.
Help
Help links in mindzieStudio now open Ask mindzie on the documentation site at docs.mindziestudio.com/ask, which answers in the language of the page. A new guide explains how to prepare a real ERP export for AI Process Intelligence.
Sign-In and Settings
Signing out takes one click. Settings has a labelled way back to the modules, and a tenant administrator on a single-module tenant can always reach Settings.
Security and Administration
License Key Protection
The license key is no longer written to any log.
Signed Installers
The Enterprise, mindzieDesktop, and Task Mining setup programs are code-signed.
Updated Web Component
A web component was updated to address a published security advisory.
Bug Fixes
| Area | Description |
|---|---|
| Upload Files | Remove, Convert again, and Build the OCEL now work on a new project |
| Upload Files | Snapshot files that carry their own snapshot date are converted instead of refused |
| Upload Files | Daily snapshot files uploaded one at a time are converted as one set |
| Upload Files | A version history with few columns is converted |
| Upload Files | A settings file goes to Additional data, and a binary file is no longer told to save as UTF-8 |
| Projects | A project that stopped during start-up no longer always reads Failed to start |
| Projects | A double press on Start, Stop, or Cancel acts once, without a false error |
| Create with AI | A small project keeps answering while a Create with AI result is saved |
| Create with AI | A log whose time column has a non-standard name is accepted |
| Create with AI | Kept date and number columns are no longer treated as free text, and dates are read in any common format, including day-first |
| Create with AI | Asking to change an existing enrichment saves a new version of it with its name kept, instead of adding a second one |
| Costs | The top-handoff waiting cost uses the median wait times the number of cases, not a sum of days |
| Costs | A cost that applies to any rule is no longer added on top of per-rule costs |
| Train Map and BPMN Map | The maps no longer show a second scroll bar inside the page |
| Pivot | Validation in the aggregation dialog shows warnings, and a one-row pivot tolerates an empty aggregation list |
| Case Stage Alerts | The alerts dialog no longer fails when the threshold is 1 or less |
| Data Connections | The table list is read on Oracle and Vertica when a schema is entered |
| Python API Client | File uploads are sent correctly |
| Task Mining Agent | Dialogs fit at every display scale, and long recordings get their report |
Upgrade Notes
- Projects created from the command line before this release may need to be reloaded once before Create with AI accepts a non-standard time column name.
Known Issues
- After a file is kept by Upload files, uploading a file with the same name as a flat event log is refused. Rename the file, or build from the Upload files list.