Managed Agents

• Run automatically on triggers - fire based on real events in your account like a tag change, form submission, new contact, appointment booked or updated, or a set schedule

• Take action in your account - update contact fields and tags, start workflows, create tasks, notes, or appointments

• Follow a repeatable process every time - teach it a Custom Skill for a task it needs to handle consistently, separate from its general Instructions

• Draw on a Knowledge Base you connect - base its actions and any answers it gives you or your team on your real business information, pricing, policies and procedures, rather than a guess

• Connect to external tools - take actions on connected platforms, or utilise external tools or your own systems your Agent needs to do its job

• Search the web and show its sources - with sources you can click to verify

• Generate media - create images, audio, and short videos as part of a task, saved to your Media Library for reuse

Ask AI is your own in-account assistant, helping you build and configure things, draft content, and take quick actions as you work.

Conversation AI talks directly with your leads and clients across various messaging channels (e.g. SMS, email, chat, etc.).

Voice AI manages all of your inbound phone calls, and can place outbound calls for you as well.

Both Conversation AI and Voice AI Agents can answer questions, book appointments, and more, in real time while chatting with your contacts.

Managed Agents work in the background. They aren’t built to hold live conversations with your customers, or manage your inboxes. Managed Agents are set off by specific triggers (e.g. a tag, a form, an appointment, a schedule) and carry out a specific job for you, start to finish. You and your team can also open an Agent directly and ask it questions, drawing on its own connected Knowledge Base.

If you want an AI to speak with a customer in the moment, that's Conversation AI or Voice AI. If you want a copilot to help you get something done right now, that's Ask AI. If you want an AI teammate to take a repeatable task off your plate, that's a Managed Agent.

1. From the main Ivorey™ menu, go to AI Agents > Agent Studio > and click [+ Create Managed Agent]

2. Choose how you'd like to start:

   1. Start from a template - select one of the pre-built use cases on the right hand side, as a starting point to customise

   2. Describe it and let AI create a draft - in the prompt area on the left-hand side, select one of the available pre-written prompts or write your own prompt to explain the Agent's role, including:

      1. Its primary goal (e.g. manage my sales pipeline)

      2. What it's allowed to do (e.g. tagging, starting workflows, updating stages, sending emails)

      3. Any boundaries it should respect

      4. Click the [↑] arrow icon or press Enter to send the prompt

3. Ivorey™ will generate a first draft of your Agent, with a suggested Trigger and full Instructions for you to review and refine (we’ll cover Triggers in the next step)

4. Review your Agent's Instructions in the panel on the right - this is the core of your Agent, covering its role, its goals, how it should work, and rules it should follow:

   1. Manually edit as needed, to remove, adjust or add in any additional instructions

   2. In the chat panel on the left, describe any adjustments you'd like to the tone, rules, or actions
      
      Note: Ask for one change at a time, (e.g. “always check for upcoming appointments” or “only suggest changes for my approval”) and the builder will update only that part.

5. Next, connect a Knowledge Base so your Agent has something real to draw from, rather than guessing:

   1. In the main Agent settings area, above the Instructions, next to Knowledge Base, click [+ Add] > and select the Knowledge Base you'd like this Agent to search from

   2. Repeat this step if you'd like to add more than one Knowledge Base to the same Agent

Note: Instructions are the Agent's general behaviour and rules, a Knowledge Base is factual reference material it looks things up in, and Custom Skills (further down) are for a specific repeatable process it should follow the same way every time. A pricing policy belongs in the Knowledge Base, but the exact steps for qualifying a lead belong in a Custom Skill.

1. From the main Ivorey™ menu, go to AI Agents > Agent Studio > click the [⋮] three dots on the Agent you want to configure > and click [Edit]

2. In the panel on the right, next to Triggers, check the suggested trigger - click the [x] on it, if it’s not right for your setup

3. Click [+ Add trigger] to replace the suggested trigger, or to add an additional trigger

4. Browse the list of available triggers, use the category tabs along the top or search for the trigger you want:

   1. Contact created - when a new contact is created in the CRM

   2. Lead tag - when tags are added or removed from a contact

      1. Under Action, select whether it should trigger from [Tag added] or [Tag removed]

      2. Select which tags should trigger it from the drop-down (you can add multiple) or click [+ Add New] to open Tags > and create a new tag

      3. Click [Save]

   3. Opportunity created - when a new opportunity is created (manually or automatically)

      1. Select which pipelines should trigger it from the drop-down (you can add multiple)

      2. Select which stages in that pipeline should trigger it or leave it empty to trigger from an opportunity created at any stage

      3. Click [Save]

   4. Opportunity status changed - when an opportunity changes status or stage

      1. Select which pipelines should trigger it from the drop-down (you can add multiple)

      2. Select which stages in that pipeline should trigger it or leave it empty to trigger from an opportunity moving to any stage

      3. Select which statuses should trigger it (open, won, lost, and/or abandoned) or leave it empty to trigger from any status
         
         Note: If you only want to trigger your Agent when an opportunity changes stages, select the relevant stages and leave statuses blank - and vice versa if you only want to trigger based on status change.

      4. Click [Save]

   5. Appointment booked - when a contact books an appointment (good for sending confirmations, pre-appointment instructions, or pre-qualifying questions)

      1. Select which calendars should apply from the drop-down (you can select multiple)

      2. Click [Save]

   6. Appointment status - when an appointment status changes (good for outcome-based follow-ups, like a win-back message after a no-show, or a thank-you after a completed appointment)

      1. Select which calendars should apply from the drop-down (you can select multiple)

      2. Select which statuses should trigger it (New, Confirmed, Cancelled, Showed, No-Show, and/or Invalid) or leave it empty to trigger from any status

      3. Click [Save]

   7. Form submitted - when a form is submitted

      1. Select which forms should trigger it from the drop-down (you can add multiple)

      2. Click [Save]

   8. Survey submitted - when a survey response is received

      1. Select which surveys should trigger it from the drop-down (you can add multiple)

      2. Click [Save]

   9. Facebook lead form - when a Facebook lead form is submitted

      1. Select which Facebook page should apply from the drop-down

      2. Select which lead forms should trigger it (you can add multiple) or leave it empty to trigger from any lead form from that page

      3. Click [Save]

   10. Chat message - when your Agent is asked something through the Managed Agents chat (internal to your team only, not a client-facing chat)

   11. Schedule - run your Agent once at a set time or on a recurring basis:

       1. Once - select a future date & time that your Agent should be triggered to run

       2. Set interval - set how often the trigger should repeat (e.g. repeat every 5 days, or repeat every 1 month) > select a future date and time for the schedule to Start > and choose when it Ends (Never, On date, After occurances)

       3. Cron expression - set a custom schedule using Cron syntax > use the expression guide to confirm everything is set correctly > or select a preset option

       4. Click [Save]
          
          Note: You can describe the schedule you want in the prompt window (e.g. run this every Tuesday, Wednesday and Thursday at 9am until the end of the month) and the builder will configure it for you - review the result before publishing. Schedules run on your account's local timezone. You can add more than one schedule to the same Agent if it needs to run at different cadences (e.g. a daily check plus a weekly summary)

   12. Invoked from a workflow - when the Invoke Agent action fires in a workflow

       1. Click [Save]
          
          Note: Your workflow will need to be published in order to trigger this.

5. Once a trigger is saved, you can click into it to edit the configuration at any time, or click the [x] to remove it

1. From the main Ivorey™ menu, go to AI Agents > Agent Studio > click the [⋮] three dots on the Agent you want to configure > and click [Edit]

2. In the panel on the right, next to Apps, click on [Default] - this is the default connection to all of the areas in your Ivorey™ account

   1. Use the drop-downs on each feature (e.g. Calendars, Contacts, Pipelines) to review the enabled skills > and deselect any actions you don’t want this Agent to be able to take in your account

   2. Or use the checkboxes next to each main feature to disable all the skills for it (e.g. if you don’t want your Agent to take any actions under Payments, you can disable all of them at once)
      
      Note: Default is connected to every Managed Agent automatically and can't be disconnected - you're only controlling which of its actions this Agent is allowed to use.

   3. Click [Save]
      
      Note: This is also where you can connect external apps via MCP server - we’ll cover that in upcoming steps.

3. (optional) Next to Capabilities, toggle ON any of the following to enable them:

   1. Web search - your Agent can look things up online and show you the sources behind its answer

   2. Image generation - your Agent can create visuals

      1. Select your preferred Quality level (low, medium or high) - keep in mind that higher quality will use more tokens

   3. Audio generation - your Agent can create audio (delivered as a separate MP3 file)

      1. Select your voice preference from the drop-down (you can preview in testing) > and include any guidance on Tone (e.g. professional and warm)

   4. Video generation - your Agent can create short-form videos (max. 8 seconds, in 16:9 or 9:16)
      
      Note: Generating images, audio and video uses more AI tokens than text, which will mean higher charges. Turn these on only for Agents that need them. Once these are enabled, your team will be able to prompt your Agent to generate media in the internal chat interface in Managed Agents.

4. (optional) Toggle ON Custom API to connect your Agent to your own systems via API > click the [+] to the left of the toggle:

   1. On the Connect to API tab, click [Import cURL] if you already have a cURL command or configure manually:

      1. Set the Endpoint URL > and method (GET, POST, DELETE, etc.)

      2. Select the Content-Type from the drop-down (json or x-www-form-urlencoded)

      3. Click [+ Add header] > and enter the Key and Value > repeat to add as many headers as needed

      4. Click [+ Add query params] > and enter the Key and Value > repeat to add as many parameters as needed

      5. Choose if you want to send a fixed JSON body

      6. (optional) Create a custom Form or switch to JSON Schema to specify data collection

   2. On the Authentication tab, toggle ON if your service requires it:

      1. Select the Authentication type from the drop-down (API key, Bearer Token, or Client Secret)

      2. Enter the API Header and Value, Bearer Token, or Secret Client ID and Password

   3. On the General tab:

      1. Enter a Name for the API

      2. Describe when your Agent should use it

      3. Toggle ON Retry on failure as a backup try for failed API runs

      4. Set a timeout (in seconds) before the API is considered failed

   4. On the Test tab, send a sample request and check the response before relying on it

   5. Once you’ve tested successfully, click [Done] to save

1. From the main Ivorey™ menu, go to AI Agents > Agent Studio > click the [⋮] three dots on any Agent you want to test > and select [Edit]

2. In the top right corner, click the [Test Agent] button to open the testing panel, which simulates the internal chat interface in Managed Agents

3. Type a prompt into the box, to test your Agent’s response and functionality (e.g. create an opportunity for [email protected] in [Stage Name] stage):

   1. Click the 🎙 microphone icon to speak your message instead of typing it

   2. Click the [+] plus icon to attach an image or file to your message

   3. Click the [↑] arrow icon or press Enter to send in the test prompt

4. Review the response and the actions your Agent took, checking they match what you described:

   1. Click the drop-down above the response, to expand your Agent's reasoning and see how it reached its answer

   2. If the Agent has Web Search enabled, review the Sources displayed beneath the response and open any to verify the information

5. If media generation is enabled, ask the Agent to create a sample asset:

   1. For image generation - test the quality and aspect ratio you plan to use

   2. For audio or video generation - confirm that the generated file appears correctly

   3. Confirm that the generated file is saved to the Media Library in Ivorey™

6. Use the chat panel on the left-hand side to prompt AI to make adjustments to your Agent’s Instructions or other configurations if needed

7. Continue prompting your Agent to review its functionality, until all of its responses and actions look correct

8. Click the [↺] arrow icon, in the top right corner, to clear the conversation and start a fresh test

9. Click the [X] button in the top right corner to exit the testing panel

10. When you're happy with your Agent’s setup and have tested successfully, click the [Publish] button in the top right corner

11. To move your Agent back into draft mode, at any time, click the [⋮] three dots in the top right corner > click [Unpublish] > and click [Unpublish] again

1. From the main Ivorey™ menu, go to AI Agents > Agent Studio to see your full list of drafted or published Agents

2. In the main table, you can:

   1. View each Agent's Status (Draft or Published), when it was Last Updated, and the date it was Created On

   2. Click the [⋮] three dots on any Agent:

      1. Edit - Reconfigure your Agent’s Instructions, Knowledge Base, Triggers, Capabilities, Skills, or connected Apps at any time (changes save automatically)

      2. Move to folder - Organise your Agents into folders

      3. Unpublish - Move your Agent back into draft-mode, without deleting its configuration. When an Agent is unpublished:

         1. Automated triggers stop immediately

         2. Anything already in progress runs to completion

         3. The Agent's internal chat keeps working, using the current draft

         4. Publishing again restores the automated triggers as a new version

      4. Delete - Permanently delete an Agent you no longer need (this is not reversible, we recommend unpublishing to save an Agent’s Instructions and configuration for future use)

3. To view your Agent’s activity, or to prompt it directly or ask a question, click on the Agent’s name or click the 💬 chat bubble icon:

   1. Prompt the Agent the same way you did when testing, to start a new chat - type in your prompt, use voice commands or upload attachments > and click the [↑] arrow icon or press Enter to send a prompt

      1. Click the drop-down above any response to view the Agent’s thought process

      2. In the top right corner, click [Edit Agent] to make any adjustments, or click [View in Agent Logs]

      3. Click [← Back] in the top left corner, to leave the chat

   2. The Activity tab, below the prompt window, will display every time your Agent was triggered, as well as any previous chat threads:

      1. Each entry shows whether it was a Trigger run and what set it off (e.g. "Form: Consultation Request" or "Tag added: new-lead"), or a Chat message and a summary of the initial prompt

      2. View the date of the run or the last message

      3. Click any entry to open its full execution log - what steps the Agent took, which skills or connectors it used, and its final result (useful for confirming an Agent behaved as expected, or troubleshooting when it didn't)

   3. In the top right corner, click [Edit Agent] to make any adjustments

   4. Click on [Agents] in the top left corner, to return to the main list of Agents

No. Managed Agents don't require coding, flowcharts or complex configurations. You only need to describe the job you want your agent to do, or write out its Instructions directly.

No. A Managed Agent works in the background, responding to things that happen in your account, or answering questions when you or your team ask it directly through its own chat.

If you want an AI Agent that talks live with your customers on a channel like SMS, email, or your website chat widget, consider setting up a Conversation AI bot.

Instructions are your Agent's general behaviour and rules. This includes its role, tone, and boundaries overall. A custom skill is a specific, repeatable process for one particular kind of task, which can also be shared across more than one Agent.

If you update a Knowledge Base that is connected to a Managed Agent, your Agent will pull the latest version the next time it runs.