Creatio AI Studio, October 15, 2026: Agents That Run on a Schedule

The most important change in the Creatio AI Studio update of October 15, 2026 is schedules: an agent can now start by itself, once or on a recurring basis, and the result of each run is posted to a conversation in the AI Twin chat.

The other changes concern AI Twin, agent knowledge, HTTP tools and voice calls, among other things. The full official notes: October 15, 2026 release notes. The notes do not mention feature flags, licensing requirements or beta labels. The previous update: October 8. A general description of the product is on our Creatio AI Studio page.

Agent schedules

How to create a schedule

On the agent’s Triggers tab, choose Add trigger, then Add scheduled trigger. A four-step dialog opens. You can also ask AI Twin to create the schedule.

What you can set

• Repeating: run once, or repeat at intervals of 30 minutes or more, or every few hours, days, weeks, months or years. You set a start date and time and a time zone. Recurring and cron schedules have an end condition: a specific date or never.
• Optional rules: days of the week, quiet hours (including ranges that cross midnight), overlap handling, and downtime handling, meaning runs that fall due while AI Studio was unavailable.
• Cron: switch to Advanced and enter a cron expression.
• Additional prompt: text the agent receives on every scheduled run. The agent keeps its own instructions.
• Environment: you assign the schedule to an environment, for example to keep test schedules out of production. You can move it later.
• Automatic stop: the Stop after failed runs field sets the number of failed runs after which the schedule stops. The default is 5. Clearing the field means the schedule never stops because of failures.
• Review step: before saving, you see a plain-language summary of the schedule and the next run times.

Managing schedules

On the Triggers tab, each schedule shows its summary, time zone, environment, status, creator, and last and next run. You can pause, resume, edit or delete it. Run now starts the agent immediately without changing the schedule.

The Triggers section now combines scheduled, event and group triggers. You can search it, filter it (type, status, last outcome, agent, owner, environment) and sort it (Next fire, Last fire, Failures).

Permissions

Scheduled runs use the schedule creator’s own access. Users can schedule their own agents that they can run, and enterprise agents where the organization allows it. They see only their own schedules unless their role can manage all of them. The owner of a personal agent manages its schedules on the agent’s Triggers tab.

Schedules built with AI Twin

AI Twin asks for missing details, states which time zone it uses, and shows the summary and the next runs for your approval before saving. You can ask it to change the time, days, quiet hours, instruction or end date, or to pause or resume a schedule. It keeps the end date exactly as you stated it.

Where this may come in handy

This is our interpretation, not examples from the release notes: recurring summaries (for example a weekly review of cases or sales opportunities), reminders about overdue tasks and regular checks of data that someone does by hand today.

What to watch out for

A scheduled run is a real run of the agent, so it can send messages or change records. That is why it is worth setting the right environment, quiet hours and failure limit, and remembering that the agent acts with the permissions of the person who created the schedule.

AI Twin

• Renaming a skill. You can ask AI Twin to rename a skill. It changes the display title (which can contain spaces and symbols), while the skill’s name stays unchanged.
• Record changes. AI Twin reports a Creatio record change only after saving it. If nothing changed, it says so and explains why. Rich-text formatting, such as in knowledge base articles, is preserved when editing.
• Files across agent switches. Files uploaded earlier in a conversation remain available after you switch agents. Files that cannot be found are skipped.

Agent design and deployment

Deploying a personal agent draft to an environment that accepts direct deploys (such as Sandbox) first publishes that draft. The agent header then shows the version as “Current”, and it is also visible in the Agents list. This covers personal prompt agents and workflow agents you can publish.

Skills and voice agents

SKILL.md files can be up to 157,856 characters (long reference material goes into the references/ folder). In realtime voice calls, the agent opens the needed skill before answering and replies in the caller’s language. This works in the “Platform tools” and “Hybrid (platform + native)” modes.

Knowledge

Agents with knowledge sources answer questions on those topics from the sources. If the answer is not there, the agent says so. If the knowledge service is down, it says it cannot look up the information right now. Source names appear only when citations are enabled for that source.

Channels and integrations

• HTTP tools: requests use the body defined in Body Template. Response Path applies only to successful responses, so on failure the agent gets the service’s error text. Calls whose arguments don’t match the tool’s input schema are refused before any request is sent.
• Safe retries: after a timeout or failure, a tool call is retried only if it is known to be safe to repeat, such as a read-only tool. Tools that create or change data are not retried in that case.
• Deactivated records: lists and totals of Creatio records exclude records marked as deactivated, such as retired lead statuses.
• Outbound call refusals: the agent receives the reason when it cannot place a call, for example a number not in E.164 format or a destination blocked by the channel’s policy. Spaces, dashes, dots and parentheses are removed from numbers before dialing.
• Channel pages have an Open agent link next to Agent Binding (hidden while the channel has unsaved changes).

Runtime and observability

• The health badge on the Deploy tab shows an agent as ready once it serves requests.
• Creating or activating an environment starts setting up AI Twin and other system agents in the background, without waiting for a first sign-in.
• Each realtime voice turn appears on the run timeline as a model step (with the model, duration and tokens when the provider reports them).
• A memory recall that times out appears as a failed step reading “Recalling memory timed out.”

What to check before rolling out schedules

1. Who creates schedules. The agent acts with the creator’s permissions, so for enterprise agents decide who may schedule them.
2. Environment. Keep test schedules out of production.
3. Limits. Check quiet hours and the Stop after failed runs threshold before you turn on a schedule that sends messages or changes data.
4. Tools that change data. Remember that they are not retried automatically after a failure, so their error handling needs to be thought through.

Want to see how this fits your company?

As a certified Creatio Level IV partner, we help plan which tasks are worth handing to agents, in which environment and with what safeguards. If you want to start with one controlled use case, get in touch.

Other Creatio AI Studio updates

See also the other Creatio AI Studio updates from October 2026:

• Creatio AI Studio, October 22, 2026: reasoning level, the Creatio Feed skill, email senders
• Creatio AI Studio, October 8, 2026: dictation in AI Twin, draft-first agent changes
• Creatio AI Studio: product overview, governance and rollout