Configure the Sidekick panel
The Sidekick panel is built from independent modules — each one a different kind of help. You control which appear and in what order so agents see the most useful module first.
Before you start
- Admin permissions on Sidekick settings
- A clear answer to “what does this team need to see first when opening a conversation?”
Steps — module ordering and visibility
- Open Settings → Sidekick → Panel layout.
- Each module has a visibility toggle and a drag handle. Two operations:
Reorder — drag a module up or down to change its position in the panel.
Hide — toggle the visibility off to remove the module from the panel entirely. - Save (or the page may auto-save, depending on UI version).
The change applies to every agent in the tenant. Individual agents don’t currently customize the panel themselves — it’s a tenant-wide configuration.
Recommended orderings by team type
- Email-heavy support, long threads — Summary (agents need to catch up before replying)
- Chat-heavy with quick replies — Suggested Answer (the draft is the fastest path to a send)
- Specialist support (technical, billing) — Knowledge Articles (the right reference is more valuable than a draft)
- Internal-procedure-heavy — Handbook Articles (procedures are the workflow)
- Data-extraction focus (orders, accounts) — Insights (the data is what the agent needs first)
There’s no wrong answer; observe what your agents reach for and reorder accordingly.
Steps — global Sidekick settings
Open Settings → Sidekick → General. Five toggles:
- Master switch — turn Sidekick on or off entirely. Off means no panel, no detection, no auto-tagging, no incident discovery. Use only for tenants that don’t want AI features.
- Incident discovery — toggle the background worker that powers the incident-match hint card. Off saves LLM cost on every inbound message; the trade-off is no automatic incident detection.
- Inline highlights — toggle whether detected values are highlighted in the message body. Off keeps the message text clean; on makes detections visually obvious.
- Auto-save — toggle whether high-confidence detections automatically write to their mapped custom fields. Off means every detection is flagged for agent review; on means trust the AI for above-threshold matches.
- Confidence threshold — the slider that decides green-vs-amber on detected values. Higher = stricter (fewer auto-saves, more agent review). Lower = more permissive (more auto-saves, more risk of bad data).
A reasonable starting point: master switch on, incident discovery on, inline highlights on, auto-save on, confidence threshold around 0.7-0.8.
Verify it worked
- Open any active conversation as an agent (or use an incognito tab logged in as a test agent).
- Open the Sidekick panel — modules should appear in the order you set, with hidden modules absent.
- Look at a recent inbound message — if inline highlights are on and detection rules exist, structured data should be visually highlighted.
Per-tenant vs. per-team configuration
Sidekick configuration is per-tenant. There’s no per-team override today. If two teams have wildly different needs (one wants Summary first, the other wants Insights), pick the configuration that benefits more agents and rely on the modules being toggleable rather than reordered per-agent.
Troubleshooting
- Symptom: A module is enabled but never shows content. Fix: The module relies on a data source that’s empty. Suggested Answer needs Knowledge / Handbook content to draw from; Insights needs detection rules to be configured. Set up the data source first.
- Symptom: Confidence threshold seems too strict — every detection is amber. Fix: Lower the threshold. Or for the specific rule, use Pattern builder / RegExp instead of AI detection (deterministic matches score higher).
- Symptom: Inline highlights are distracting agents. Fix: Turn off inline highlights in General settings. Detections still happen and still flow to custom fields; they just don’t visually pop in the message body.