AI Chat

Found in: Performance > AI Chat

AI Chatbot is the next step in making your data easier to access and understand.

If your company has subscribed to our AI & Analytics Package, this chat interface becomes available.

The AI feature is currently in open beta and available to super admin users in your organization. During this testing phase, it comes at no additional cost.

Interface Overview

The page offers a familiar chat-style interface. You type your question, and the assistant replies in real time.

Additional Configuration Options

  • Preserve chat history: If enabled, the full conversation history is passed to the AI instead of just the last message. This provides more context but can increase usage costs. Leave this on when refining a line of questioning. Turn it off for one-off queries.
  • Model selection: Choose the model that fits your needs. Use a faster, lower-cost model for simpler questions. Switch to a more advanced one when deep data analysis is required. Balance cost vs. outcome.

Functionality Overview

Start by asking the assistant for help—try “What can you do?”

Then follow up with specific data questions:

Screenshot

(Example: If your system has a “product category” field but no “product type,” make sure your query reflects that. Word choice matters.)

Once you’re happy with the text response, you can add it to a dashboard using the “Add to Dashboard” button:

Add to Dashboard Add to Dashboard

Want a visual? Ask for a graph instead:

Graph Representation

Know the type of chart you need? Be specific—for example, say “line chart”:

Line Chart Example

You can add visuals to dashboards just like text answers.

Query Writing Tips

Today’s AI models are powerful—but not perfect. Think critically and review results carefully. Use exact field names from SynControl when possible. While the AI might sometimes guess your meaning correctly, being specific helps get better results. Think of this like giving clear instructions to a new team member.

Example of a well-written request:

“List the acronyms of the clients associated with bookings that haven’t been invoiced yet and that are over 4 man-days in aggregate, that have a current step greater than 1, and that have a start date in the past. Include the client acronym, comma-separated booking UUIDs and URLs to each booking, and the count of uninvoiced bookings, sorted by count.”

Example of a query that may be misunderstood:

  • Say you have a “qualified” property on suppliers, with values like “yes”, “no”, or blank.
  • If you ask for “unqualified suppliers,” the AI might interpret this as “suppliers with a blank value.”
  • A better query: “Include count of unqualified suppliers. An unqualified supplier is one where the qualified column is ‘no’.”

Data Privacy & Security

Your data remains confidential. We allow the ChatGPT engine to read your database to generate replies, but OpenAI has confirmed that your data is not used to train their models.

Source: How We Use Your Data

We may support additional language models in the future, but data privacy will always remain a top priority.

AI Notifications

You can set up automatic alerts by simply asking the AI questions like:

  • Notify me when there are more than 5 unpaid invoices overdue by 30+ days
  • Alert me when any open issue is created with priority = high
  • Notify when a new defect is reported classified as critical
  • Notify me of new bookings created that are still unconfirmed after 24 hours
  • Notify me daily at 10am about number of bookings created yesterday with sum of man-days

These will become active notifications (automations). Ask the AI to:

  • “List my notifications” to review
  • “Delete notification #5” or “Delete booking notification” to remove

MCP Server Integration

Our AI now supports connections to an MCP (Model Context Protocol) server. The endpoint uses OAuth 2.0 dynamic client registration for token issuance and adheres to modern security standards.

Connecting with Claude

To connect Claude to the MCP server:

  • Follow the tutorial here: Use a Remote MCP Server
  • Give the server a clear name (e.g., syncontrol) — you’ll need to reference it in questions. For example: “count total bookings in syncontrol”
  • The server address will be [your-domain]/mcp — make sure to use HTTPS!

Connecting with Perplexity

To connect Perplexity to the MCP server you need to firstly issue an Oauth token since perplexity doesn’t yet follow MCP authentication protocol.

Creating OAuth token

  • Open the syncontrol platform and log in
  • Click on your user profile icon in topright corner -> OAuth -> Personal Access Tokens
  • Click on Create new Token. Check all scopes and click Create
  • Copy the contents of the token in input field

Adding server to perplexity

  • Follow the tutorial here: Local and Remote MCPs for Perplexity
  • In Add Connector menu click Advanced tab and paste the JSON replacing:
    • YOUR_SERVER with your server
    • YOUR_TOKEN with the OAuth token you copied
      {
"command": "npx",
"args": [
"-y",
"supergateway",
"--streamableHttp",
"https://YOUR_SERVER.syncontrol.com/mcp",
"--oauth2Bearer",
"YOUR_TOKEN"
]
}
  • Choose meaningful name for Server Name — you’ll need to reference it in questions. For example: “count total bookings in syncontrol”