Skip to main content

Creatio AI Studio updates: September 5, 2026

In this update, AI Twin handles long tasks and interruptions better. The chat waits for as long as the agent's run time budget allows, and an answer still arrives after the connection restarts. When AI Twin cannot complete a request, it tells you why and what to do next. Long tables in answers are easier to read, copy, and expand. Agents that work with Creatio data can also read rich text and answer from your saved folders. Learn more about all the new capabilities below.

AI Twin​

Answers scoped to your saved folders. When you name a team, group, segment, territory, or similar scope, AI Twin finds the matching saved folder. It then answers from the records in that folder. For questions about closed records or past periods, AI Twin uses the folder's definition and says how it got the answer. If folder tools are not available for your Creatio connection, AI Twin finds the scope through a related record and says so.

Clearer messages when AI Twin cannot complete a request. When AI Twin cannot accept or finish a request, it tells you why and what to do next. It may ask you to wait because a response is already in progress or the conversation is waiting for approval. It may ask you to refresh because the conversation belongs to another environment. It may also ask you to contact an administrator when the agent is unavailable or you have no access to the conversation. If the service is temporarily unavailable, it asks you to try again. The message is still there after you reload the page.

Answers delivered after a connection restart. Sometimes AI Twin restarts its connection while an agent is answering, for example, after your sign-in session is renewed. AI Twin then reconnects and shows the answer in the same tab when it is ready. You do not need to reload the page.

Long tables in answers. Tables in AI Twin answers are easier to read and reuse. A long table has a fixed height and scrolls inside the answer, with the header row pinned. A short table is shown in full. Each table has Copy table and Expand table actions. A copied table pastes as a table into documents and email, and as one cell per column into a spreadsheet. An expanded table opens over the chat. To go back, click Collapse table or press Esc.

Skills offered for the selected agent. When you type "/" in the AI Twin message box, the skill list shows only the skills attached to and enabled on the selected agent. Select one to run that skill for your request. If a skill is no longer available to the agent, AI Twin asks you to refresh the skill list and try again.

Skills​

Stage history skill. The Trace stage transitions vendor skill uses the stage history of a record. It tells you which stage the record was in before, where it came from, or which stage it was in when it closed. If the history is incomplete, the skill says so instead of giving a partial count. To use it, click Import skill → Import from vendor registry in the Skills library. Then attach the skill to an agent that uses the Creatio integration.

Channels and integrations​

Readable rich text and clearer empty results in Creatio data. Agents that work with Creatio data read rich-text fields as readable text, with line breaks and lists kept. Examples are the body of a knowledge base article or an email. Sometimes a filtered query or a folder returns nothing because the object has no records the agent can read. The Creatio tools then say so and name the object, so the agent can tell an empty result from a source with no data. If the Creatio instance cannot be reached, the tools report a connection problem. The feature works again as soon as the instance responds.

Clear reasons when an agent cannot send email. An agent that cannot send email tells you why and what an administrator needs to change. For example, no email channel may be bound to the agent, or outbound sending may be off for its mailbox. The agent may also be bound to more than one mailbox that can send. If only one of the agent's mailboxes can send, the agent uses that mailbox. This applies after you publish the agent again.

Trust and governance​

View permission paired with manage permission. When you save a role with Manage others' personal resources, the role also gets View others' personal resources. This way, its members can find the personal resources they can edit. Roles saved earlier get the view permission the next time you save them.

Runtime and deployment​

Long runs in the AI Twin chat. An agent working on a long task can finish its answer in the AI Twin chat. The chat waits for as long as the agent's Run time budget (seconds) allows, up to 3,600 seconds. Set the budget on the agent's Advanced tab. It applies once you deploy the agent version that has it.

Faster response to model provider outages. When the model provider is unavailable, an agent run ends in about a minute. The error says that the provider is unavailable, so you can try again sooner.