- “Every Monday at 8am, email me a summary of new leads from last week.”
- “Each morning, find opportunities closing this week with no activity in 5 days and post them to the Sales Chatter group.”
- “On the 1st of the month, create a follow-up task for every account with an open renewal.”
Creating an agent
You don’t fill out a form — you just ask in the chat. Describe what you want done and how often, for example “Every Monday at 8am, email me a summary of new leads.” Logic AI turns that into an agent: a name, the instructions it will follow each run, the schedule, and the tools it’s allowed to use. It confirms the details with you before saving. New agents show up in the Scheduled Agents tab.The Scheduled Agents tab
Each agent appears as a card showing its name, when it last ran and how that went, and a Scheduled badge if it runs on a timer. From a card you can:- View — open the agent to see its schedule, instructions, and full run history
- Edit — change anything about the agent (see below)
- Run Now — trigger a run immediately, without waiting for the schedule
Editing an agent
Press Edit and the agent opens in a chat sheet — just tell Logic AI what to change: “run it every weekday instead of Mondays,” “also include closed-lost deals,” or “stop sending the email, just post to Chatter.” You can change the name, the instructions, the schedule (or remove it to make the agent run-on-demand only), which tools it may use, the model, and whether it’s active. You can also delete an agent from its edit view. Deleting an agent also cancels its schedule.What an agent can do — and can’t
An agent runs with the same permissions and tools as you. It can only see and change records you could, and it can only use tools your admin has enabled for agents. If you ask an agent to use a tool that needs admin approval, Logic AI tells you which one. A few things to know:- Runs cost credits, the same as a normal chat, and count toward your usage limits. An agent stops after a set number of steps (its max turns) so a single run can’t run away.
- Because an agent runs unattended, there’s no one to click “approve” mid-run — so an agent only ever uses the tools it was given, within your permissions.
- Schedules only fire while the agent is active.

