Skip to main content

Creatio AI Studio updates: October 15, 2026

With this update, your agents can run on a schedule. Set an agent to run every weekday at 9:00, on the first day of each month, or once at a time you choose. Each run posts its result to a conversation in the AI Twin chat, so you find it waiting. You can build a schedule in a short dialog or ask AI Twin to do it. Agents with knowledge sources also answer questions from those sources, and skills can carry much longer instructions. Learn more about all the new capabilities below.

Scheduling​

Scheduled triggers. Agents can run on a schedule instead of waiting for someone to start them. For example, an agent can send a daily digest of overdue invoices, sweep stale leads every week, or check in with customers every month. You can build a schedule in a short dialog on the agent's Triggers tab. AI Twin can also build it for you in the chat.

To add a schedule, open the agent, go to the Triggers tab, click Add trigger, and select Add scheduled trigger. A dialog takes you through four steps: you name the schedule, set when it runs, add an optional prompt, and review it.

  • Repeat pattern. An agent can run once, or repeat every 30 minutes or more, or every few hours, days, weeks, months, or years. Choose the start date and time, and the time zone. For a recurring or cron schedule, under End condition, select On a specific date and pick the date, or select Never.

  • Optional rules. Add only the options you need:

    • Days of the week limits the schedule to certain days.
    • Quiet hours keeps the agent silent during a time range, including a range that crosses midnight, such as 22:00 to 06:00.
    • Overlap handling sets what happens when the previous run is still going.
    • Downtime handling sets what happens to a run that was due while Creatio AI Studio was unavailable.

  • Cron. If you know cron, switch the schedule to Advanced and enter a Cron expression.

  • Instruction for every run. In Additional prompt, add text the agent receives on every scheduled run, for example, "focus on disputes older than five days." The agent keeps its own instructions.

  • Environment. Select the Environment the schedule runs in, so a schedule you build for testing stays out of Production. You can move the schedule to another environment later.

  • Automatic stop. Under Stop after failed runs, set the number of failed runs that stops the schedule. The default is 5. To never stop on failures, clear the field.

  • Preview. Before you save, the Review step describes the schedule in plain words in the Schedule summary and lists the next runs.

On the Triggers tab, each schedule shows its summary, time zone, environment, status, creator, and its last and next run. You can pause, resume, edit, or delete a schedule. Run now starts the agent right away and leaves the schedule unchanged.

Every run of a schedule posts its result into one conversation in the AI Twin chat, next to your other conversations. Reply in that conversation to keep working with the agent.

A scheduled run uses your own access, as if you had started the conversation with the agent yourself. Every user can schedule the personal agents they are allowed to run. Where your organization allows it, users can also schedule enterprise agents that are available in the AI Twin chat. Unless your role can manage all scheduled triggers, you see and manage only the schedules you created. The owner of a personal agent manages its schedules on the agent's Triggers tab.

Schedules set up with AI Twin. You can also ask AI Twin to set up a schedule, for example, "Every weekday at 9, have the collections agent summarize overdue invoices." AI Twin asks for anything that is missing, such as the start or the end of the schedule. AI Twin tells you which time zone it uses, taken from your profile or your current time zone. Before it saves the schedule, AI Twin shows the schedule summary and the next run times and waits for your approval. To change a schedule, ask AI Twin to change its time, days, quiet hours, instruction, or end date, or to pause or resume it. AI Twin keeps the end date exactly as you state it.

All triggers in one list. In the Triggers section, every scheduled, event, and group trigger you can see is in one list. Search the list, and filter by type, status, last outcome, agent, owner, or environment. Click a column header, such as Next fire, Last fire, or Failures, to sort the list. From the list, you can pause, resume, run now, edit, or delete a schedule.

AI Twin​

Rename skills through AI Twin. You can ask AI Twin to rename a skill. AI Twin changes the title that people see and keeps the skill's name. Titles can include spaces and symbols.

Record changes confirmed only when made. When AI Twin changes a Creatio record for you, it says so only after it saves the new value. If nothing changed, AI Twin says so and explains why. When AI Twin edits a rich-text field, such as a knowledge base article, the formatting is kept.

Files kept when you switch agents. Files you uploaded earlier in an AI Twin conversation stay available after you switch to another agent in that conversation. A file that cannot be found is skipped.

Agent design and configuration​

Personal agent drafts published on deploy. You can deploy a draft of a personal agent in one step. When you deploy the draft to an environment that accepts direct deploys, such as Sandbox, Creatio AI Studio publishes the draft first. The agent header then shows the version as "Current," and the Agents list shows its version. This applies to personal prompt agents and workflow agents that you have permission to publish.

Skills​

Longer skill instructions. Skills can carry much longer instructions. The SKILL.md file can have up to 157,856 characters. To keep your skills efficient, move long reference material into files in the references/ folder.

Voice agents answer in the caller's language. In realtime voice conversations, agents open the skill they need before they answer, and they reply in the caller's language. This works in the "Platform tools" and "Hybrid (platform + native)" modes.

Knowledge​

Answers based on your knowledge sources. An agent with knowledge sources answers questions on their topics from what it finds in those sources. If the sources do not contain the answer, the agent says so. If the knowledge service is not available, the agent says that it cannot look up the information right now. The agent names a source only when citations are on for that source.

Channels and integrations​

HTTP tool requests built from your template. HTTP tools send the request body that you define in Body Template. Response Path applies to successful responses, so when a request fails, the agent sees the error text from the service. When the arguments of a call do not match the tool's input schema, the call is refused before any request is sent.

Open the bound agent from a channel. Channel pages have an Open agent link next to Agent Binding. Click it to open the agent that the channel is bound to. The link is hidden while the channel has unsaved changes.

Safe retries for tool calls. Sometimes a tool call times out or fails after it may have reached the service. In that case, Creatio AI Studio repeats the call only when the tool is known to be safe to repeat. For example, a tool that only reads data is safe to repeat. A tool that creates or changes data is not repeated in that case.

Deactivated records left out of answers. When agents list or total Creatio records, they leave out records that Creatio marks as deactivated, such as retired lead statuses.

Reasons for refused outbound calls. When an agent cannot place an outbound call, the agent gets the reason. For example, the number must be in E.164 format, or the channel's outbound destination policy does not allow the destination. Spaces, dashes, dots, and parentheses are removed from the number before the call.

Runtime and deployment​

Health badge for serving agents. The health badge on the Deploy tab shows an agent as ready as soon as it serves requests.

AI Twin in new environments. When you create or activate an environment, Creatio AI Studio starts setting up AI Twin and the other system agents in it. This happens in the background, without waiting for a first sign-in.

Observability​

Voice turns in the run timeline. Each turn of a realtime voice call appears as a model step in the run timeline. The step shows the model and the duration. It also shows the tokens when the model provider reports them.

Failed memory recall shown as a failed step. When recalling memory times out, the run timeline shows a failed memory step, "Recalling memory timed out."