The Problem: Right now, Cheat Sheet operates as a "Black Box." We feed it data (Timeline entries, emails, meetings), 360 data, etc and it gives us an assumed summary. While the concept is great, the execution lacks the steerability required for enterprise CS Ops.
Currently, we have limited visibility into why the model highlights certain things and ignores others. We see internal emails between colleagues being flagged as “insights” but are just noise which is creating a poor signal-to-noise ratio.
We’ve been provided workarounds for excluding internal comms using custom field in timeline to flag as “Internal communication”, but we may want some of those signals when risks are identified or discussed.
We’re lacking documentation or confirmation that the Cheat Sheet model actually respects these custom fields if we were to implement them and at the same time we’re essentially building shadow configurations with no guarantee they will influence the output. Without explicit alignment controls, these workarounds are just guesses.
Feature Function Request:
True Admin Configurability to make Cheat Sheet a viable tool for Ops teams, we need to move from a passive "consumer" model to an active "orchestrator" model. We need a dedicated Cheat Sheet Administration page that allows for:
1. Data Source "Steerability" & Filtering
-
Ability to include Success Plans, custom objects, signals from custom product areas.
-
Object/Field Exclusion: The ability to explicitly exclude specific Activity Types or entries based on logic (e.g., we have a custom field to collect internal vs external attendees.
IF "External Attendees" = Null AND XYZ Field = ABCTHEN Exclude from Cheat Sheet -
Do not opt any objects besides timeline in/out by default - customers who have not configured steerable data sources should have the OOTB default as-is. Notifications should be sent if this is going to change with ample notice.
2. Signal Weighting
-
Contextual Importance: The ability to assign higher weight to specific touchpoints. (e.g., A "QBR" or "Executive Business Review" timeline entry should carry 5x the weight of a standard "Check-in" email).
-
Persona Weighting: The ability to weight signals from key stakeholders (Decision Makers) higher than day-to-day users.
-
This could be a great opportunity to implement something the way AI-"suggestions" for scorecards is built. Like if standard/system objects that are heavily used could be recommended to include or included objects as part of a config.
3. Custom Signals & Sections
-
Define New Signals: Allow Admins to define what constitutes a "Risk" or "Opportunity" signal based on our own data, rather than relying solely on the pre-trained generic model.
-
Custom Sections: Capability to configure a section within Cheat Sheet that looks at specific, Admin-defined datasets or objects and summarizes them alongside the standard text analysis.
-
Signals using Co-Pilot-like Config Elements:
-
There’s also opportunity here to also have signals act like key definitions for Copilot - so having capability to fine-tune vs the standard OOTB. Allowing for signal recommendations, category recommendations
-
And by proxy: potential for some kind of "power user" type capabilities, or like end user "suggestion" capabilities like exist in Copilot config
-
