Changelog
What's new in Airclou
Customer-visible updates across Helpdesk, Playbook, and the platform — newest first, dated when they reached production.
See what's coming nextSteer the AI agent to a different draft
Approval cards gained a Steer button: say in one line what the draft should do differently and the AI agent tries again — so asking for a second draft no longer means rejecting the first.
Deleted organizations are erased
A deleted organization’s data is now erased for real once its grace period ends, by a nightly job scoped to that one organization.
The AI agent can retry a failed payment
Asked to retry a payment, the AI agent now finds the open invoice and actually retries it inside your approval envelopes, then replies with the real result instead of restating the dunning email.
Your brand on visitor-facing surfaces
The chat widget and help center now render the brand the visitor arrived through — its own name, logo and colours — rather than one workspace-wide look.
Dashboard overview shows only measured numbers
Some figures on the Dashboard overview were placeholders rather than anything we measured; they have been removed until the metering behind them is real.
Switch workspace without signing out
Change workspace from inside the app — including to a workspace in a different organization, which previously meant signing out and back in.
Choose your app when you sign in
Signing in with no app in mind now lands on a launcher instead of a silent default, and signing in from an app’s own link takes you straight there — the Application dropdown on the login page is gone.
Placeholders in the new-ticket composer
Placeholder variables now resolve when you compose a brand-new ticket, not only when you reply to an existing one.
Knowledgebase reporting
Article views and top articles moved into Reporting alongside the rest of your numbers, in the same layout — the separate knowledgebase analytics page is gone.
Chats nobody picked up find an owner
A live chat that started with no operator online used to sit unassigned until its SLA quietly breached; unowned chat tickets are now recovered and routed, and the SLA safety net sees them.
What the help center shows a visitor
Internal articles no longer surface in the public help center’s popular list, category counts now match what the reader can actually open, and empty categories and sections stay hidden.
AI replies sign as the right brand
A drafted reply now signs as the brand the ticket belongs to, including for brands whose sender name and address differ.
Linking tickets
A usability pass on the link-ticket modal — finding the ticket you mean and confirming the link is quicker.
Unsubscribe links in notification emails
The unsubscribe link in subscriber notification emails pointed at the customer portal and always landed on a 404 — it now reaches a working unsubscribe page.
SOP template library
Thirteen free, adaptable SOP templates — read them, copy them, or import one into Playbook and make it yours.
Browse the templatesGates: author the checks the AI agent must pass
The AI agent’s Playbooks tab became Gates and gained an authoring UI: define the per-brand checks a reply or action must clear before it goes out.
Inspectable steps on approval cards
Approval cards now include the AI agent’s steps — every lookup and check it ran before proposing the action — so you can examine its work before you approve it.
Group SLA timers on email tickets
Group-ownership SLA timers now start for tickets created by inbound email, matching tickets created in the app.
SLA policy editor
Clearable target inputs, a per-row unit toggle, and a cleaner targets grid in the SLA policy editor.
Multi-brand workspaces
Run several brands from one workspace — each with its own email addresses, help center, and AI agent settings.
Provenance on every AI message
Every AI-authored message now carries a provenance panel — the sources it cited and the reasoning behind the reply — and a live progress timeline shows the AI agent’s work while it runs.
Custom connectors
Custom connectors gained multi-region base-URL fallback; the editor got a wider sidebar, shared selects, and deep-linkable actions.
Attachment serving
Ticket attachments are now served through authenticated, opaque URLs, and outbound email embeds inline images instead of linking back to the helpdesk.
AI replies grounded in attachments
The AI agent now reads the documents customers attach — invoices, exports, PDFs — and grounds its draft replies in them.
Cap on consecutive auto-replies
A hard cap on how many times the AI agent can reply in a row on one ticket without a human stepping in — configurable per workspace, with a system-wide backstop.
Auto-approval envelopes
Set per-action envelopes — a refund ceiling, for example — inside which the AI agent’s pre-approved actions run without a click. Everything outside an envelope still queues for human approval. Configurable per workspace and per brand.
Replies grounded in past tickets
Draft replies are now grounded in how your team answered similar tickets before — not just the knowledgebase.
This changelog starts in May 2026. For the arcs before then — and what's being built now — see the roadmap.