← All modules

Q9 — The Quick Suite AI Layer

Why this module exists

Everything in modules Q0–Q8 existed, in some form, before October 2025. This module is everything that didn't: the AI workspace AWS wrapped around the BI product — spaces and the index that ground it, chat and the agents that front it, redesigned topics that turn questions into SQL, flows and automations that act on other systems (including an agent that autonomously drives a web browser), and research reports that take half an hour and cite their sources.

It exists for a specific person: the one whose boss said "be the Quick expert." That means naming every component, knowing which plan gates it, quoting its limits from the quota table, and answering the three grown-up questions — what did the AI do, who told it to, what did it cost.

The one idea to hold onto

The AI layer retrieves before it reasons — and everything it retrieves from, you curate.

   you curate            it stores & finds          it acts
   ┌──────────────┐      ┌───────────────┐      ┌──────────────────────┐
   │ Spaces        │      │               │      │ Chat & chat agents   │
   │  files        │ ───→ │  Quick Index  │ ───→ │ Flows (per-user)     │
   │  dashboards   │      │  (Q Business  │      │ Automate (org, UI    │
   │  topics       │      │   machinery)  │      │   agents, HITL)      │
   │ Knowledge     │      │  $5/GB over   │      │ Research (20–40 min) │
   │  bases (6 src)│      └───────────────┘      │   ↑ all metered in   │
   └──────────────┘                              │     agent hours      │
                                                 └──────────────────────┘
        …and every API in this picture still says `quicksight`.

Garbage space, garbage answers. Full index, failed uploads. Empty agent-hours pool, three dollars an hour. The layer is only as good as what you feed it and only as governable as the logs you enabled on day one.

What you'll be able to do

1. Name every Quick component, which plan/edition gates it, and which of the three meters it burns.
2. Design a space with the file-permission exception in mind, and explain why Salesforce content can never be indexed.
3. Build and share a custom chat agent, placing knowledge correctly between reference documents and spaces.
4. Explain the legacy-vs-V2 topic split — in the console and in the API — and choose dashboard Q&A when a topic is overkill.
5. Route any automation request to Flows or Automate in one sentence, and describe the UI agent and human-in-the-loop machinery accurately.
6. Answer an auditor: CloudTrail for actions, CloudWatch vended logs for conversations — and hand AWS Support a complete evidence bundle for an AI-layer case.

Lessons

Then: Cheat sheet · Lab · Quiz · Interview

A note on the docs

This module inherits Q0's three-doc-tree problem and adds new contradictions of its own, all live on 2026-08-26:

• Per-space storage: user guide says 10 GB, Service Quotas table says 1 GB (not adjustable).
• Index capacity: quota table still lists the superseded 60 GB figure; the capacity page documents auto-scale to 2 TB.
• Resources per space: "20 per resource type" (user guide) vs "100 per space" (quota table), unreconciled.
• The product's marketing page now titles itself "Amazon Quick" while the launch blog said "Quick Suite" — the third name in under a year.
• Live typos in shipped docs ("agentic worflows", "Upto 100,000 characters") — quoted verbatim in lessons so you recognize the page you're on.

Where the docs contradict themselves, the lessons show both numbers and say which to plan against. Where a fact couldn't be verified (fields-per-topic limit, Research quotas, Fast/Pro model identities), the lessons say "not documented" — that phrase is a feature of this course.

Facts verified 2026-08-26 against the pages cited in each lesson.

Keeps playing into the following modules — 116 min from here to the end of the course.