For AI agents: the complete documentation index is available at https://docs.clickmax.io/en/llms.txt, the full documentation bundle is available at https://docs.clickmax.io/en/llms-full.txt, and this page is available as Markdown at https://docs.clickmax.io/en/features/max/temperature-score-with-ai.md.
  • English

      • How do I ask the AI to configure or explain contact Temperature and Score?

      Describe what you want to know or change ("why is Ana cold?", "create a rule that gives 20 points to anyone with a director job title", "how many hot contacts have a high Score?"). The AI reads your account's current configuration before answering or proposing any change.

      Before you start:

      • Role: any role that can already see the contact can ask for explanations and the health of the base. To change Temperature or Score, the role must be owner, admin, editor or coordinator attendant — the same rule as the screens under CRM settings.
      • Temperature measures what the contact did (behaviour) and cools down on its own over time. Score measures who the contact is, through rules you define.
      • To use your own AI instead of Max, see Connect your AI to Clickmax (MCP).

      1. Ask for the number of one contact to be explained

      Give the contact's name or email and say which indicator you want to understand.

      • Temperature: the AI shows the saved number and band (Cold, Warm or Hot), the events that weighed most, what the contact has not done yet that would add points, and the history of changes. The number recalculated right now can differ slightly from the saved one, because Recency and Frequency decay every day. See What does each temperature score number mean on a contact? for the four partial scores.
      • Score: the AI lists every rule and says whether it scored, did not score, is switched off or has an invalid criterion, and tells you when the floor (0) or the ceiling of the Score hid part of the sum.

      A contact that was never calculated shows as not calculated, not as zero.

      2. Ask for the health of the base

      Ask for a picture of the whole base or of an audience ("only people from Instagram"). The AI answers how many contacts are Cold, Warm and Hot, how many have a Low, Medium and High Score, the cross of the two, and how many are Hot with a high Score — the ones to call first. If the Score is switched off, every contact shows without a Score.

      3. Ask to adjust Temperature

      Say the change in words ("set the window to 21 days", "make meeting held weigh more", "use the High-ticket model"). The AI reads the current configuration, changes only what you asked and summarises before and after.

      Before switching model, ask for the impact preview ("how many contacts would be hot if I used the Perpetual model?"). Without saving anything, the AI shows how many contacts are Cold, Warm and Hot today and how many would be with the other model, with up to three example contacts. On large bases the count is an estimate built from a sample of the active contacts — the AI tells you when that is the case. The preview uses the chosen model's own cut-offs, window and channel weights.

      Three things to know before you confirm:

      • Switching the model does not copy its numbers. The model sets the weight of Recency, Frequency, Depth and Declaration; window, cut-offs and channel weights stay as they are unless you ask to use the model's values too. For the result to match the preview, ask to apply the model's numbers as well.
      • Changing Temperature recalculates the whole base, in the background and over a few minutes. Contacts can change band, lists filtered by temperature change size, and automations that use a temperature change as a trigger can fire in bulk.
      • You can ask to save the calibration as your own model (up to 20 per workspace). Saving a model does not activate it; the AI only uses it when you ask to activate it. The active model cannot be deleted.

      4. Ask to build the Score

      Describe the criterion and the points. Before saving, the AI measures how many contacts the rule reaches and shows where they would land: how many more or fewer contacts would sit in each tier (Low, Medium, High) and up to three examples with today's Score and the new one. When editing an existing rule, the preview already takes out what that rule adds today. It is an approximation based on the saved Score; the final number comes from the recalculation after saving. Then the AI creates the rule and, if the Score is still switched off, warns you that nobody gets a number until you switch it on.

      • Each rule gives or takes points (−1000 to 1000); a contact's Score is the sum of the enabled rules it meets, never below 0 nor above the ceiling.
      • The criterion looks at the contact record — job title, fields, tags, origin, purchases. Behaviour ("opened an email") is Temperature's job.
      • To pause a rule, ask to switch it off ("switch off the Director rule") instead of deleting it: switching off leaves its name, points and criterion untouched, and you can switch it back on later. Deleting removes the rule and its criterion. Before switching it back on, ask for the preview to see how many contacts would change tier.
      • Switching the Score on or off and changing cut-offs or rules recalculates the Score of every contact. Switching the Score off clears everyone's number; switching it back on recalculates. This recalculation does not fire Score-change automations.

      Actions that delete, such as removing a rule or a saved model, ask for your confirmation before they happen.

      If it didn't work

      • The AI says it cannot change something: your role is read-only. Ask an owner, admin, editor or coordinator attendant.
      • I created rules but contacts still have no Score: the Score is switched off. Ask to switch it on and wait for the recalculation to finish before checking the numbers.
      • I saved Temperature and the list still shows the old values: the recalculation runs in the background. Wait a few minutes and ask for the health of the base again.
      • The model changed but cut-offs and channel weights are the old ones: expected; ask to apply the model's numbers too.
      • A rule shows an invalid criterion: the field or tag used in the criterion no longer exists. Ask to fix the criterion or switch the rule off.