Q9 cheat sheet — the Quick Suite AI layer

All facts verified 2026-08-26. APIs/IAM/ARNs say quicksight; console and docs say Quick.

Components → lesson → plan gate

Component	What	Gate
Spaces + Quick Index	curated knowledge + retrieval/storage engine	all plans; shared spaces & KBs from Plus
Quick chat + agents	conversational front door; system agent "My assistant" + custom agents	all plans
Topics (V2)	multi-dataset semantic layer, ≤12 datasets, join keys	Enterprise Edition
Dashboard Q&A	1-click "Allow data Q&A" checkbox, no topic needed	enablement fee
Quick Flows	per-user task automation, ≤35 steps	Plus and up (pricing page ambiguous re Free)
Quick Automate	org-level process automation, UI agents, HITL	Professional/Enterprise, console-provisioned accounts only
Quick Research	20–40 min cited research reports	all plans; burns Research agent hours

The three meters (+ hidden ones)

Infrastructure: $250/account/month (Pro/Ent), flat.
Index storage: 25 GB/user (Pro) · 50 (Ent), pooled; $5/GB/mo over; counts source bytes; non-home-Region capacity = all overage.
Agent hours: 4/user (Pro: 2 agentic + 2 Research) · 8 (Ent); pooled; $3/hr over. One Research run ≈ 20–40 min.
Hidden: dashboard Q&A enablement fee; Automate custom-mode Bedrock inference bills to the Bedrock account.

Spaces — the two rules

Linked resources keep their own permissions. Uploaded files are readable by everyone in the space. Upload = access grant.
Delete a space → files deleted, linked assets only unlinked.

Files: 500 MB (PDF/Word/PPT), 50 MB others; 10,000 files/space; chat convo = ~20 files. ⚠️ Space storage: user guide 10 GB vs quota table 1 GB not adjustable — plan on 1 GB, flag it.

Index / knowledge bases

Index = Q Business machinery (CMK sync needs qbusiness and quicksight principals).
KB sources (only these are indexable): Google Drive, OneDrive, Confluence, SharePoint, S3, Web Crawler. ⚠️ Salesforce/ServiceNow/Jira/Slack = Actions only, never indexed.
Sync daily by default; >14 days → FAILED; index full → space uploads fail.
Auto-scale 0.25 GB steps → 2 TB (ticket beyond); quota table still says 60 GB (stale).

Chat & agents

Scoping modes: All data and apps / General knowledge / Specific data and apps. Web search admin-gated, per-query globe toggle.
Explanations panel → generated SQL per numeric claim (datasets).
Reference docs = always in context (10 docs / 50 MB / 100K chars, non-adjustable); spaces = searched at question time. Policy → reference doc; library → space.
Agent API: create-agent etc. (max 10 connectors, 10 spaces, 3 starter prompts; lifecycle PREVIEW|PUBLISHED). No conversation API — embed instead (domain allowlist → Share via embed). Conversations retained 90 days (non-adjustable).

Topics

New: metadata enriched on datasets; topic = datasets + relations (+ instructions); chat generates real SQL (joins/unions/subqueries/self-joins); API = *-topic-v2.
Legacy: metadata inside topic object; fuzzy-search picks one dataset; API = *-topic. No auto-migration — Dataset Enrichment is the path.
Fields-per-topic limit: not documented (nearest: 2,000 fields/dataset).

Flows vs Automate

	Flows	Automate
Runs as	individual user	the org (independent of a user)
Build	NL / visual / duplicate / from chat	NL, process-doc upload, canvas; groups→projects→automations
Scale	35 steps, 100/user, 10k/instance	versioned (Live→Committed→Deployed)
Agents	—	UI agent (drives browser), custom agent (tools incl. MCP, Code, HITL)
Trigger	manual, chat, browser ext., schedules (⚠️ us-east-1/us-west-2/eu-west-1 only)	schedules (10 parallel/trigger, 50/account) + `start-automation-job` API (payload ≤7M chars)
Trap	action list-steps read first batch only	needs console-provisioned account; Fast/Pro models unnamed; Custom = BYO Bedrock

Audit — the split

CloudTrail = actions (API calls for flows/agents/automations/connectors + dashboard views). CloudWatch vended logs = content: CHAT_LOGS, FEEDBACK_LOGS, INDEX_USAGE_LOGS, KB_FILE_SYNC_LOGS, agent-hours usage. No retroactive capture — enable day one. Usage dashboard gated by quicksight:QuickSuiteUsageMetrics.

Support evidence bundle (AI layer)

account + home Region · plan/edition · doc URL incl. which tree · automation: group ID + automation ID + JobId + status + run artifacts · index: capacity setting + INDEX_USAGE_LOGS + KB_FILE_SYNC_LOGS · chat: CHAT_LOGS excerpt + Explanation SQL · CloudTrail event IDs.

CLI quick reference (all `aws quicksight …`)

create-space | list-spaces | update-space-resources | describe-space-permissions
create-knowledge-base | list-knowledge-bases | list-users-index-capacity
create-agent | list-agents | update-agent-permissions
create-topic-v2 | describe-topic-v2            # new topics
create-topic | batch-create-topic-reviewed-answer   # legacy topics
list-flows | search-flows | get-flow-metadata | update-flow-permissions
start-automation-job | describe-automation-job
create-action-connector | list-action-connectors