Roll out and improve Spotter

Building memory is not the finish line — it’s the prerequisite. Once Spotter is loaded with context, the next step is getting it in front of users deliberately, not all at once. Roll out in phases, watch how people use it, and use those signals to keep the memory current.

Roll out in phases

Start small. A full Org rollout before you have signal from real users makes problems harder to diagnose — noise from many users obscures the patterns you need to fix things quickly.

Phase	Who	What to watch
---	---	---
Power users / champions	2 to 3 analysts who know the data model well and will give honest feedback	Do they trust the answers? What do they correct? What questions fail entirely?
Target team	The one business team you identified in usecase discovery	Are they asking the questions you scoped? Where do they rephrase or drop off?
Org-wide	All users on that data model	Volume of unanswered questions, quality regressions, downvote patterns

Phase

Who

What to watch

---

Power users / champions

2 to 3 analysts who know the data model well and will give honest feedback

Do they trust the answers? What do they correct? What questions fail entirely?

Target team

The one business team you identified in usecase discovery

Are they asking the questions you scoped? Where do they rephrase or drop off?

Org-wide

All users on that data model

Volume of unanswered questions, quality regressions, downvote patterns

Don’t move to the next phase until the current one produces clean signal. Champions surface the most important gaps fastest. One team’s real usage patterns tell you what the org will do at scale.

Share access

Give champions Manage Memory access

Champions need to correct Spotter in conversation and have those corrections saved as shared memory. That requires Can manage Spotter — the privilege for managing memory and Spotter instructions across the Org.

How to assign it:

In clusters with RBAC enabled: add Can manage Spotter to a role, then assign the role to a group. Add your champions to that group.
In clusters without RBAC: add Can manage Spotter directly to a user group.

Use a group, not individual grants. Assign Can manage Spotter to a group, add champions to it. When you expand to more users later, you add them to the group — you don’t re-configure the privilege.

Can manage Spotter implies Can use Spotter — no separate Spotter privilege is needed.

For reference on permissions, see Memory permissions & governance.

Enable personal memory for your Org

Once you’re ready to expand beyond champions, enable personal memory so all users can teach Spotter about themselves.

Go to Admin > All Orgs > AI settings.
Toggle on "Enable personal memory learning for all users".

This is a separate flag from the main memory toggle. The main memory flag must be on for personal memory to work — but you can have the main flag enabled without enabling personal memory.

See Personal memory for what users can do once this feature is enabled.

Read the signals and act

Once users are active, the Spotter Conversations Liveboard shows you where Spotter is and isn’t working. Check it regularly — weekly during the champion phase, at least monthly after full rollout.

Signal	What it means	What to do
---	---	---
Questions with no answer	Memory coverage gap for that topic	Add a relevant Liveboard source or teach in conversation
Same question rephrased multiple times	Wrong memory applied, or retrieval miss	Correct the rule in conversation
Downvoted answers	A specific memory rule is producing bad output	Find and correct it in Managing memory
Low usage despite enablement	Spotter not trusted, or users don’t know it’s available	Run a live session with champion users to demonstrate

Signal

What it means

What to do

---

Questions with no answer

Memory coverage gap for that topic

Add a relevant Liveboard source or teach in conversation

Same question rephrased multiple times

Wrong memory applied, or retrieval miss

Correct the rule in conversation

Downvoted answers

A specific memory rule is producing bad output

Find and correct it in Managing memory

Low usage despite enablement

Spotter not trusted, or users don’t know it’s available

Run a live session with champion users to demonstrate

See Spotter Conversations Liveboard to learn where to find it, what KPIs it shows, and how to read exact usage data.

The improvement loop:

Spotter Conversations Liveboard
        ↓
  Identify gap or bad rule
        ↓
  Add or correct memory
  ├── New topic gap      →   Add memory from Liveboards
  └── Wrong rule         →   Correct in conversation or Managing memory
        ↓
  Retest with champion users
        ↓
  Monitor again

The loop doesn’t stop. Memory stays current by cycling through it — champions surface issues, you fix them, you verify they’re fixed, you watch for the next round.