Manage Your AI Chat Controls

Chat controls shape how Procore AI and Datagrid respond to your requests. Choose your preferred agent, tools, actions, large language model (LLM), saved prompts, and stored memories — all from the chat window shown below.

By default, chats use Agents (Auto) to handle your requests without requiring you to select an agent first. However, you can manually select a specific agent when you want a task to stay with one specialist. This tutorial walks through both approaches, along with the other chat controls available to you.

• Prerequisites:

  • To see Procore AI's chat controls: Your account must have an active subscription to one of the Procore Digital Coworker packages, and a Company Admin enabled your project and assigned you a role. See Enable Procore AI.

  • To see Datagrid's AI chat controls (without Procore AI): Your account has purchased and enabled the Datagrid product.

• Credit consumption: Credits are consumed by every AI question, agent task, and background process such as storing and syncing your data. Learn more about AI credits.

• Validation: AI can make mistakes. Always check all AI work, results, and answers against your own records to ensure accuracy.

Understand Auto Mode

The Agents (Auto) mode is the AI's default state during conversations. In Auto, Procore AI automatically identifies routes, tools, and specialized agents it can use to fulfill your requests.

Not sure whether to stay in Auto or select a specific agent? Use this guide:

Select a Specific Agent

Manually override Auto mode and select a specialized agent when you want to focus on a specific, known workflow or task:

1. Click Agents (Auto) to open the menu of agents.

2. Click on the name of one or more agents from the list to select them. Or, learn how to create an agent if you have an AI Enterprise account.

   • Stacking multiple agents: If you have two or more agent's selected at the same time, the AI router considers the context of your request and routes it to the agent deemed most appropriate. When you ask follow up questions, the router continues sending those to the same agent unless it decides that the task is clearly outside that agent's scope. In that case it will switch to another selected agent. Alternatively, if you know which agent you prefer to handle a given request, name it in your message to the AI.

3. The agent name(s) appear in the header of the prompt composer, and you can begin chatting with an agent.

4. Optional: To return to Auto mode after working with a specific agent, click on the icon(s) representing the previously selected agents, and then select the auto option at the top, labeled Procore AI or Datagrid AI.

Explore More Chat Options

Whether you're chatting with a specific agent or in Auto mode, when you need to give the AI extra reach or context, use the following features via the plus + menu.

Attach Files

Upload a local PDF or spreadsheet directly into the chat so the AI can consider and use that information the next time you submit a request.

Actions and Tools

Click on Actions and Tools to enable connectors and functions that give the AI more reach.

1. Actions: Consider enabling an Action your company has added, or adding a new Action by clicking the + Add Action button at the bottom of the Actions and Tools menu. Note: Procore Enterprise is an action that is enabled by default for Procore AI customers. This allows the AI to fetch and submit content directly to your Procore tools (RFIs, Submittals, etc.). Learn more about the actions the AI can help you take.
   
   Some of the Actions that may be available:

   • Google Calendar

   • Microsoft Teams

   • Slack

   • Trimble Connect

   • SharePoint

   • Microsoft Planner

   • DocuSign
     
     Connect via MCP: In the same + Add Action window, you can also click Connect via MCP to connect to a third party product via Model Context Protocol (MCP). MCP is an advanced data connector that allows the AI to securely access additional data sources, such as an internal database. This feature is only available to Enterprise accounts. See the 'MCP Servers' section of Configure an Agent for set up instructions.

2. Tools: Consider enabling tools that are not included by default in Auto mode or most agents, such as the following tools:

   • Web Search: Looks up current information on the open internet and cites its sources, which is how you get at things that live outside your own data, such as external regulations or today's weather. Unlike Access Links, you do not need to know the URL — it finds the pages for you.

   • NEC Calculator: Runs real National Electrical Code calculations such as voltage drop and conduit fill, and consults material properties against NEC standards. Calculate does the arithmetic; this applies the code rules that determine which arithmetic is correct. Especially useful for electrical and specialty contractors.

   • Company Research: Gathers public background information on a company (industry, size, locations, recent news) so you can size up a bidder, developer, or vendor without opening a dozen tabs. Without enabling this tool, the AI will only focus on what is already in your own project data. Especially useful for owners and developers.

   • Procore Support: Searches Procore's own help and support documentation (including this tutorial) to answer "how do I do this in Procore" questions. Web Search might land on an outdated forum post; this reads the official documentation.

Use Knowledge

Guide the AI to read and retrieve facts from your project data. Select one or more data sources from @Use Knowledge to focus the AI on specific datasets or page environments.

• When to use it: When you need to research specific details or facts buried in large files—like drawings, specifications, or historical project records. This is your best tool for "deep dives" into the thousands of pages that make up your project’s knowledge base.

Prompt Library

Click Prompt Library to save your most effective and complex prompts for reuse later. Toggle prompts between "Only for me" and "Share with team" to share your most effective prompts with your team and help standardize workflows.

Agent Memories

Click thebrain icon to view the personal knowledge the AI has stored about you.

• This includes stored context like your name, role, time zone, and project preferences to make responses more relevant over time. If the AI is consistently getting your title or location wrong, you can edit or clear these memories here.

LLM (Large Language Model)

Click LLM (Large Language Model) to switch to a different LLM, such as a specific version of Claude or Gemini. Note: Model options can vary by account, depending on company configurations such as cloud platform, and switching your LLM affects your available features as described below.

The LLM default setting is Auto (Recommended). When you keep or return to this setting, the Agent in use (Auto mode or a manually selected agent) will select the best LLM based on its configuration.

Verify and Audit Results

1. Review the results against your project data to verify accuracy.

2. These optional features may help as you audit the AI's work:

• Steps: Click the expandarrow at the top of the AI's response if you'd like to see the actions and logic it used to produce that result. Short AI responses with no tool use do not display these steps.

• Sources:

  • When the AI used sources (your data, a PDF, Procore, or the web) to support answering you, numbered marks appear next to its statements that those sources helped it with. Clicking on one of them opens that source. Replies that did not look anything up will not show these.

  • At the end of the AI's message, it will list the number of sources that ground its answer, if there were any. Click on it to see more details about the sources.

• Thumbs up & thumbs down: Click on these to tell us if the answer was or wasn't helpful. Rating a reply is how that exchange can be used to improve the product. Details that identify people are stripped before anyone looks at what you share. Conversations you don’t rate stay in your chat history and are not reviewed.

Troubleshooting:

Lack of context:
If you're using a specific agent and it seems "project blind" or "context blind," you can click Go to Configuration from themore menu next to its name at the beginning of one of its responses and check whether it has access to the necessary datasets. Learn more about configuring an agent. Or, for a more temporary solution, add the necessary data to the conversation via the plus + menu in the window where you compose your messages to the AI.

The AI says it can't do something (like search the web, open a link, run numbers)
Not every tool is on. Web Search, in particular, is not on by default for Auto mode or for new agents. Open Actions and Tools from the plus + menu in the chat, and switch on what the task needs, then ask again. For something you'll need regularly, turn it on in the agent's configuration instead so it persists.

A different agent answered than you expected
In Auto, the chat picks the agent it thinks fits, and it tends to stay with the one already answering. This also works similarly if you've manually selected more than one agent. If you'd like it to use a different one, name that agent in your chat to the AI. To learn more about how this works, in Auto mode, see Understand Auto Mode section. To learn more about how this works with manual agent selection, see Select a Specific Agent above, and read about stacking multiple agents.