Power Platform Community Forum Thread Details

We are in the process of introducing copilot to our users but have come across a few set backs. I have a few questions regarding copilot configuration and limitations.

1) Copilot Prompt in Sales Hub

- Does this only look at data fields selected in copilot settings? For example, if we ask when was this lead created on, and created on isnt a selected field, copilot return anything?

- Is copilot aware of the current open record or do you have to define what record prior to retrieving data?

- Are the selected data fields in copilot settings used for Summary and Copilot prompt or is there seperate configuration for copilot?

- Is copilot prompt generative? i.e. draft and email to lead X

2) Row Summary

- Is row summary for lead no longer supported? Why would this be removed? We have summaries in lead entity that are way off and is including data that is irrelevant.

2) Lead and Opportunity Scoring

- Is this still supported? I ask because Sales Insights Settings is no longer in the change area. I can only access through a link. Im sure I can add it back, but was not removed by us.

- We recently updated the fields within lead scoring, retrained the model and published. We have several examples of lead records, created within the designated timeframe, that have not been re-scored and are still displaying the old fields. The info hover text says it is only retrained when values change that contribute to the scoring, which makes sense, but in this case the fields have changed so I would think it would have to re-score? Or are we stuck with the current grade until the previously selected field is updated?

Thank you!!

Categories:

Copilot Studio skills development

Hi there,

You've got several distinct questions mixing Sales Copilot (in Sales Hub) and Predictive Lead/Opportunity Scoring (Sales Insights). Let me address each, grounded in Microsoft Learn docs:

1) Copilot Prompt in Sales Hub

Does Copilot only look at fields selected in Copilot settings? — Yes. The fields you select in Copilot for Sales settings define the dataset Copilot grounds its answers against. If "Created On" isn't in the selected fields, asking "when was this lead created?" won't reliably return data — the field needs to be exposed to Copilot first.
Is Copilot aware of the open record? — Yes, when invoked from the side pane on a record form, Copilot has the current record's context. When invoked outside that context, you generally need to specify the record.
Are selected fields shared between Summary and Copilot prompt? — The Record Summary and Copilot prompt features use the same admin-controlled field selection in Copilot for Sales settings; the field list is shared, though some prompts are scoped per scenario.
Is the Copilot prompt generative? — Yes. Tasks like "draft an email to lead X" use generative AI grounded in the record and configured fields.

2) Row Summary for Lead

Row summaries are still supported in Sales, but quality issues like irrelevant data being summarized typically come down to:

Too many fields being included in Copilot's record summary scope (noise → irrelevant summaries).
The lead has linked records or activities that aren't useful for the summary but are being pulled in.

Trim the included field list in Copilot for Sales admin settings and re-test. If you have customized fields with little business meaning, exclude them.

3) Lead and Opportunity Scoring (Predictive Lead Scoring)

Is it still supported? — Yes, predictive lead scoring is fully supported and is a core Sales Insights feature. The settings panel has been restructured under the new admin experience — if "Sales Insights Settings" isn't visible in your sitemap, you can re-add it via App Designer, or access it via Sales Hub → Change area → App Settings → Sales Insights settings.
Why aren't leads being re-scored after retraining? — Per the Configure predictive lead scoring docs: scoring runs incrementally. Re-scoring of existing open leads happens only when the values in the contributing fields change, OR when a new model is published. If you retrained and published but old leads still show old scores, check:
- Confirm the new model is in Published (not Draft) state.
- The model has at least 40 qualified + 40 disqualified closed leads in the training window; otherwise the AUC can be too low to be usable.
- There can be processing delay before existing open leads are rescored against the new model. Wait up to a few hours after publish.
- If a contributing field's value hasn't changed on a given lead, the lead's score won't be recalculated until either the field is touched or the next batch run picks it up.

References (Microsoft Learn):

Hope this clears things up — happy to drill into any single point further.

Raghav Mishra — LinkedIn | PowerAI Labs

Found this helpful? Please mark ✅ "Does this answer your question?" so others searching for the same issue can find it quickly. A 👍 on "Was this reply helpful?" or a ♥ Like is also much appreciated!