Back to the changelog

• Feature

  Conversational profiles

  Management gains the Conversational profiles section (premium feature): define how the assistant talks to external participants, pick the channels of each collaborative phase in the Designer and add notes for the agent to the flow. The actual conversations arrive in the next versions.

• Security

  Data subject requests

  In Settings → Data subject requests the administrator records an access, erasure or rectification request, previews the cases involved and executes it: personal data is pseudonymised leaving the process structure intact, selected documents are deleted, and everything stays traced in history and audit. To find the cases, just declare which fields identify the person.

• Security

  Data subject export

  An access request produces a ZIP archive with the person's data, a reading guide and the linked documents, ready to be checked before handing it over.

• Security

  Profile: your data and your account

  From your Profile you can download your personal data as JSON and ask for your account to be deleted: administrators receive the request and fulfil it from the Data subject requests section.

• Feature

  AI decision explanation

  Every decision taken by AI in a case (an automatic state, a task outcome, a rule reading AI-produced data) now has an explanation you can open and print from the history: which assistant and model decided, with which answers and confidence levels, which rule won and why. The explanation stays readable after the call log is purged; read-only users see the evidence but not the values. Assistant configurations are versioned: every change creates an immutable revision.

• Security

  Decisions about people

  In the flow settings you can declare that the process decides about people: publishing then rejects a path where AI alone leads to a negative outcome without an intermediate human review, and flags unverifiable scripts. The AI Act chapter is now in the in-app guide.

• Improvement

  AI summary and history

  Existing AI summaries remain available without automatic regeneration (no cost at update time); new summaries no longer include AI execution comments, and read-only users see pre-update summaries only once the case has renewed them. The title recalculated by a formula appears in the history as "Title recalculated".

• Security

  Data flows and sub-processors

  In Settings → Data flows and sub-processors you find the inventory, generated from the configuration, of every destination FlowSharp may send data to (AI providers, email, documents, n8n, webhooks, MCP, federation, licensing portal): host names only, never credentials or full addresses, with the declared jurisdiction of each provider, ready to export as JSON or print for the processing register.

• Security

  Allowed AI hosts

  The administrator can restrict AI providers to a list of allowed hosts: a provider outside the list cannot be saved and a task that would use it fails without calling the model, with an audit trace. When a provider's model or host changes, the administrators of tenants using it in published flows receive a notice.

• Improvement

  Management

  The Extension section no longer shows a double title, sections are recognised even with a trailing slash in the address, and the Implementation Plan counter refreshes every minute.

• Improvement

  Guides

  New chapters in the Analytics/AI and governance, Flows and Licensing guides; new runbook for data subject requests.