> For the complete documentation index, see [llms.txt](https://doc.aissist.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://doc.aissist.io/release-notes.md).

# Release Notes

*Last updated: October 30, 2026*

{% updates format="full" %}
{% update date="2026-10-26" %}

## Drive and website assets, database actions and asset usage

* **Google Drive and OneDrive assets:** pick files, or follow a whole folder (up to 200 files), from Google Drive, OneDrive or SharePoint. Docs, Sheets, Slides, PDF, Word, PowerPoint, Excel and images are read, and kept in sync every 6 hours.
* **Website assets:** follow a whole website, or part of it, or pick the pages you want. Pages are re-checked on a schedule, and you can exclude pages you don't want used.
* **Database actions:** connect a database and let the AI look up records in it during a conversation, limited to the tables and fields you allow.
* **Usage counts for assets, actions and sub-agents:** see how often each asset is quoted in replies, down to the page, file or article; how often each action is called; and how many conversations each sub-agent handles. Find what is never used at a glance.
* **Clearer asset pages:** the sub-agents that answer from an asset are now at the top of its page, and each page or article has a button to open it in a new tab.
* **Connector capabilities:** each connector's page lists what it can do: its actions, the knowledge it gives, the gateways it serves and more.
* **NPS:** NPS now uses the standard formula, with passives counted, in the console and in weekly reports.
* **Ask Engine:** look into specific conversations, such as why a conversation was or wasn't handed to a person, and check whether a message would start a sub-agent.
  {% endupdate %}

{% update date="2026-10-19" %}

## Help desk actions and knowledge, safer actions and per-project reports

* **Safer actions:** actions that create or update records in your systems now get an extra check before they run.
* **Sub-agent instructions guide actions:** when an action's own instructions are thin, the AI uses the instructions of the sub-agent handling the conversation to call it correctly. You can also name an action in a sub-agent's instructions.
* **Help desk actions:** the AI can create tickets in Zendesk, Intercom and Freshdesk, and work on the customer's own ticket: close or reopen it, assign it, tag it, fill in its fields or add an internal note. Each action runs only when your instructions call for it.
* **Help desk knowledge:** import your help center articles, and macros where your help desk has them, as assets that stay in sync.
* **Zendesk:** tickets and messaging now share one Zendesk connector: messaging is added after the Zendesk sign-in, and its API key is optional.
* **Weekly reports by project:** weekly reports are now written per project, with new agent insight sections. The onboarding analysis was improved too.
* **Alerts name their drivers:** alerts now say what drove a change, with the figure behind each driver.
* **Ask Engine:** clearer charts, and add or update conversation examples by asking, with each change shown to you before it is saved.
* **Stronger protection against prompt injection.**
* **Fixed:** some conversations received an empty reply.
  {% endupdate %}

{% update date="2026-10-12" %}

## More accurate replies, escalations and in-note commands

* **More relevant knowledge:** each reply is now built from the most relevant parts of your assets, with weak matches filtered out.
* **More accurate sub-agent detection:** conversations are matched to the right sub-agent more reliably.
* **More accurate escalations:** decisions to hand a conversation to a person are checked more carefully.
* **Better web search:** more reliable results, with more search options.
* **In-note commands use previous conversations:** commands can now take a customer's earlier conversations and tickets into account.
* **Ask Engine can add actions and assets:** create actions and assets by asking for them in plain language.
* **Onboarding:** connect your systems while you onboard, and choose whether to run the conversation analysis.
* **Website:** pages now open in your language automatically.
  {% endupdate %}

{% update date="2026-10-05" %}

## Reply accuracy and Intercom events

* **Fewer unsupported claims:** replies are checked claim by claim, so fewer statements go out without support.
* **Intercom:** the AI can now respond to tag and attribute changes on Intercom conversations.
* **More reliable tagging.**
  {% endupdate %}

{% update date="2026-09-28" %}

## Connectors and the sales playbook

* **Connectors:** gateways and smart actions now connect through connectors, so each system you use is connected once and reused everywhere.
* **Sales playbook:** sales conversations get a playbook of their own, aimed at closing the sale.
* **More accurate tagging:** every tag is checked a second time before it is set.
* **Ask Engine:** rate its answers, and search conversations faster.
  {% endupdate %}

{% update date="2026-09-14" %}

## Ask Engine answers in more detail

* **Hour-by-hour answers:** Ask Engine now answers about periods shorter than a day, such as "the last two hours" or "this morning".
* **Why conversations were escalated:** ask why conversations were handed to a person, for a whole workspace or one sub-agent, with exact counts and rates.
* **Many conversations at once:** ask what happened across a set of conversations, such as which sub-agents handled them, what was found, and which actions ran.
* **Complete quality scores:** conversation summaries, quality scores and search now catch up on conversations missed during traffic spikes.
  {% endupdate %}

{% update date="2026-09-07" %}

## Guardrails for actions that change data

* **Extra guardrails:** actions that update records in your systems are guarded more strictly, and are never triggered by detection alone.
* **More complete context:** long conversations are summarized so the AI keeps the whole picture.
  {% endupdate %}

{% update date="2026-08-31" %}

## Closing conversations and sub-agent quality

* **Closing conversations:** the AI can now close conversations that are finished.
* **Quality by sub-agent:** traffic and quality reports now break results down by sub-agent.
* **Instructions in evaluations:** evaluation sets and agent insights now take instructions.
* **Simpler sub-agent setup:** sub-agents now work out from their instructions which images, assets and actions they need, so fewer settings are required.
* **Faster tagging.**
  {% endupdate %}

{% update date="2026-08-24" %}

## Ask Engine creates, notifications explain

* **Ask Engine creates for you:** Ask Engine can now create sub-agents and evaluation sets, and analyze a single conversation in depth.
* **Notifications explain themselves:** alerts and notifications now say what is likely behind a change, not just the numbers.
* **Realistic simulations:** simulations and evaluations can now be built from real conversations and agent insights.
* **Smarter workspace creation.**
* **Redesigned setup:** a new way to create integrations, actions and gateways.
  {% endupdate %}

{% update date="2026-08-17" %}

## Introducing Ask Engine

* **Ask Engine:** a new engine that answers questions about your workspace from its own data: traffic, conversations, escalations, quality and more.
* **Reports powered by Ask Engine:** weekly reports and agent reports are now written by Ask Engine.
* **Faster conversation search.**
* **Sign in with Google:** sign up and sign in faster with your Google account.
  {% endupdate %}

{% update date="2026-08-10" %}

## Console refresh and role-based agent insight

* **Refreshed console:** a cleaner layout across the console.
* **Role-based agent insight:** agent insight is now organized by role.
* **Better replies:** improvements to how replies are written, and a refined escalation strategy.
  {% endupdate %}

{% update date="2026-08-03" %}

## More accurate replies and better agent insight

* **More accurate replies:** irrelevant context is filtered out before a reply is written.
* **More reliable output:** structured output is more consistent.
* **Clearer agent insight:** agent insight comments now include specific examples.
* **Faster reports.**
  {% endupdate %}

{% update date="2026-07-27" %}

## See what shaped each reply

* **Sources of a reply:** see which instructions, sub-agents, actions and assets contributed to each reply, and which part of each.
* **Actions first:** the actions page now opens on a table of your actions.
* **Better image handling.**
  {% endupdate %}

{% update date="2026-07-20" %}

## More accurate tagging and actions

* **More accurate tagging:** fewer tags set by mistake.
* **More reliable actions:** simple, one-step actions run more reliably.
  {% endupdate %}

{% update date="2026-06-29" %}

## Higher resolution and stability

* **Higher resolution:** conversations are routed using more signals about how important the reply is and how long it should be.
* **Stability improvements.**
  {% endupdate %}

{% update date="2026-06-15" %}

## Smarter routing

* **Smarter routing:** each conversation is now matched to the level of AI it needs, from a simple greeting to a complex request.
* **More consistent output.**
  {% endupdate %}

{% update date="2026-06-04" %}

## Improved image processing

Image processing now uses a better AI model.

This improves image understanding and processing capability.
{% endupdate %}
{% endupdates %}