Jonathon's AI Wiki

Seven Operator Use Cases for Claude's New Stack (Rick Mulready)

13 min read

Source: raw/7_NEW_Mind-Blowing_Use_Cases_of_Claude.md Creator: Rick Mulready (online-business-owner / AI Playbook membership) URL: https://www.youtube.com/watch?v=L6RBrmYlTAg Published: 2026-04-07 Duration: 17:42 Platform: YouTube

A 17-minute operator-perspective tour of Claude features that shipped in March-April 2026, framed as seven business use cases rather than a feature changelog. Filed in claude-ai because the value is in how a non-developer wires the features together — a complement to the surface-by-surface coverage in W16 / W17 and the canonical entries on Cowork, Dispatch, Computer Use, and Scheduled Tasks. Distinct angle: end-to-end Operator-track usage with Cowork Desktop as the home base, no Claude Code involved.

Key Takeaways

  • The seven use cases hang on six features shipped in the recent weeks: Memory 2.0 (auto-memory + auto-dream), Cowork Projects, Dispatch, Computer Use, Scheduled Tasks, Visualizations, plus the Cowork Ideas marketplace.
  • Memory is the foundation. Auto-memory (default on) synthesizes conversations every 24 hours into a memory summary that loads into every new conversation. Auto-dream is the cleanup pass between sessions — Claude reviews its own memory files, merges duplicates, removes outdated entries, and reorganizes for currency. Inspect with /memory from Cowork or Claude Code.
  • Cowork Projects ≠ regular Claude.ai Projects. Both have file uploads + custom instructions. Difference: regular Projects are a chat workspace (you drive every turn); Cowork Projects are a task workspace with persistent memory (you assign, Claude executes, the project remembers what happened on previous runs and computes deltas). Three creation paths: from inside Cowork, by importing an existing claude.ai Project, or by pointing at an existing folder on disk. Plus a “Start a task in Cowork” button now lives inside regular claude.ai Projects.
  • Dispatch is the phone-as-remote-control surface for Cowork. Send tasks from your phone, Claude executes on your computer, conversation thread syncs across devices. Hard limitation: the Mac/PC must stay awake and Cowork desktop must stay open — speaker enables the “Keep awake” setting in Cowork settings → Dispatch as the workaround. Demo prompt: “research three competitors’ pricing pages and build me a comparison spreadsheet” → fire from phone, dinner, sit down to a finished spreadsheet.
  • Dispatch + Computer Use is the qualitative jump. Once browser actions are enabled (Settings → General → “Allow all browser actions” + Computer Use enabled), the prompt menu opens up to anything your logged-in browser can reach — speaker’s demo: “Go to my YouTube analytics, pull my last 10 videos’ performance data, and build me a spreadsheet with views, click-through rate, and average view duration.” Caveat: must be already logged into the target tool in the browser Claude controls.
  • Scheduled Tasks wire daily AI-newsletter triage into Slack — speaker’s “AI news digest” cron runs in Cowork at 7am, pulls 18+ newsletters from a Gmail label, deduplicates, and posts a categorized digest to Slack. Two MCP-style connectors used: Gmail and Slack, both visible under Customize → Connectors.
  • Computer Use UX audit pattern. Point Cowork at a deployed app URL and tell it “go through the full user flow as a new member from the OTP login through to the dashboard, identify any UX issues, confusing steps, or friction points.” Claude opens the browser, retrieves the OTP from Gmail, walks the flow, and emits a written audit. Speaker’s run: ~15 minutes for an audit that “would normally take a team member half a day.”
  • Visualizations generate interactive HTML diagrams inside the chat (no Cowork required) — speaker’s example: a clickable onboarding flow with decision points for success/failure, exportable as a Mermaid chart via the Mermaid connector for paste-into-Notion/etc.
  • Cowork Ideas is the cold-start helper — a marketplace-style menu sortable by role (Marketing / Engineering / etc.) with pre-written prompts. Click “Turn documents into a presentation” → Cowork session opens with the prompt pre-filled and asks for the source assets.
  • Permission discipline matters. Speaker recommends “Ask permissions” mode for first-time users on Computer Use; “Auto accept edits” / “Plan mode” only after you trust a workflow. Per-app deny lists live under Settings → General. Reference Computer Use for the canonical security guidance.
  • Standard editorial caveat — video closes with a pitch for the speaker’s paid AI Playbook membership (“7-day free trial,” “Your Zone of Genius app”). Same lead-magnet-funnel pattern documented in Brandon Storey’s lead-magnet course from earlier today.

Memory 2.0 — auto-memory + auto-dream

Worth pulling out as its own section because the speaker frames it (correctly) as the foundation under everything else.

  • Auto-memory — On by default. Every 24 hours, Claude synthesizes recent conversations into a memory summary that loads into every new conversation across surfaces. Inspect via /memory (Cowork or Claude Code; speaker demos in Claude Code, “continue in terminal” prompt → editable). Distinct from the August-2025 “search past chats” feature, which was on-demand only.
  • Auto-dream — Anthropic’s name (per the speaker, confirmed at /memory), not a community nickname. Between sessions, Claude reviews its memory files, merges duplicates, removes outdated content, and reorganizes for relevance. The Wikipedia-trick equivalent for memory state.
  • Why it matters for everything else in this article: When Cowork Projects, Scheduled Tasks, or Dispatch routines run on a recurring basis, they need to know your business without you re-uploading context every time. Memory 2.0 is the under-layer that makes “weekly review that gets smarter every week” coherent rather than amnesiac.

The Seven Use Cases

1. Weekly CEO review (Cowork Projects)

Cowork Project named “[Business] Weekly Review.” Each Friday: prompt “Run this week’s report” → Claude pulls all key business metrics, summarizes what happened, computes week-over-week deltas. The deltas only happen because the Project has memory of previous runs — a regular Claude.ai Project would have to be re-fed last week’s data manually.

Speaker’s demo report ships with: dates / headline / key numbers / green flags / watch list / focus. Same prompt next week → automatic delta calculation against this week’s baseline.

2. End-of-day handoff (Dispatch)

Walk away from desk → fire prompt from phone → return next morning to a finished deliverable. Demo prompt: “research three competitors’ pricing pages and build me a comparison spreadsheet.” Cowork asks clarifying questions, runs in background, spreadsheet waiting on return.

Setup checklist:

  • Cowork desktop installed + logged in
  • Open Dispatch from Cowork side panel → walk through phone pairing
  • Settings → Dispatch → enable “Keep awake” so the laptop doesn’t sleep mid-task
  • Phone is now a remote control for the Cowork-on-desktop instance

3. Phone-triggered Computer Use (Dispatch + Computer Use)

Same Dispatch surface but with Computer Use enabled — Cowork can interact with browsers and websites, not just files. Demo prompt: “Go to my YouTube analytics, pull my last 10 videos’ performance data, and build me a spreadsheet with views, click-through rate, and average view duration.”

Pre-reqs:

  • Already logged into the target tool in Cowork’s browser
  • Settings → General → “Allow all browser actions” enabled
  • Computer Use enabled
  • (Recommended) per-app deny list configured for sensitive apps
  • (Recommended) Ask-before-acting permissions for first-time use

The qualitative shift: from “phone is a remote for file tasks” to “phone is a remote for any web-app task you can describe.”

4. Daily AI news digest (Scheduled Tasks → Slack)

Cowork Scheduled Task running at 7am daily. Prompt instructs Claude to look in a specific Gmail label (“AI Newsletters”), filter for criteria the operator defined (story types worth reading), deduplicate (“most newsletters say the same thing”), categorize, and post a digest to Slack. Two connectors used: Gmail + Slack, configured under Customize → Connectors.

Operator-track win: Slack is open every morning anyway — the digest is there before the operator opens their laptop.

5. Full UX audit of a live app (Computer Use)

Point Cowork at the production URL of an app. Prompt: “Go through the full user flow as a new member from the OTP login through to the dashboard, identify any UX issues, confusing steps, or friction points that a new user might encounter.” Speaker uses Opus 4.6 explicitly (“more complex workflow”).

Cowork’s plan: load browser tools → walk through login → fetch the OTP from Gmail → step through every screen → write an audit. Demo run completed in <15 minutes against an app the speaker said would normally consume a team member’s half-day.

Output structure: executive summary, what works well, issues identified, “four things I’d fix first.” Equivalent to a junior product designer’s first-pass audit document.

6. Interactive onboarding-flow visualization (Visualizations)

In regular claude.ai (no Cowork required). Prompt: “Create an interactive process flow for how users get added to the [app name]. I’ve listed out the steps. Make each step clickable to show what happens. Include decision points for success/failure.” Output: clickable HTML chart inside the chat — click a node to see its command endpoint and decision point.

Export path: connect the Mermaid chart MCP connector (Plus button → Connectors → Manage → Browse → “Mermaid” → install) → export the entire flow as a Mermaid sharable link or as code suitable for embedding in Notion / docs / a dashboard.

7. Cowork Ideas (cold-start helper)

For first-time operators staring at the empty Cowork window. Click “Ideas” in the left rail → sort by role (Marketing, Engineering, …) → pick a recipe like “Turn documents into a presentation” → Cowork session opens with the prompt pre-filled and a clarifying-questions step asking for the source assets.

Functions as a discovery surface for what Cowork can actually do without forcing the operator to write the first prompt from scratch.

Try It

For a WEO Marketly operator (no Claude Code required) wanting to apply this in 30 minutes:

  1. Verify Memory 2.0 is on. Open Cowork → start a chat → run /memory. Confirm “auto-memory: on” and inspect what Claude already knows about you. Edit if anything is wrong or out-of-date.
  2. Pick one recurring task that already lives on a calendar reminder. Weekly client report? Monday content review? Friday outreach digest? That’s your first Cowork Project.
  3. Build it as a Cowork Project (not a regular claude.ai Project). New Project → from-scratch → write the prompt as if you were briefing a contractor. Run it once manually. The next run will compute deltas against this baseline.
  4. Schedule it. Convert the same Project’s prompt into a Scheduled Task at the cadence that matches the calendar reminder. Connect Gmail / Slack / Drive as needed.
  5. Add Dispatch only if you have phone-time when desk-time isn’t possible. Setting up Dispatch without a real out-of-the-house workflow is overhead without payoff.
  6. Save Computer Use for the audit-and-extract category specifically. It earns its setup time on tasks like the YouTube-analytics pull or the UX audit, not on tasks where an MCP connector or a Connector already exists.
  7. Apply WEO governance before connecting client tools. Permission discipline matters — start with Ask-before-acting for any connector touching client data.

Implementation

Tool/Service: Claude Cowork desktop (free with Pro/Max), no Claude Code required. Optional: Claude Code for the /memory inspection if preferred.

Setup notes:

  • All seven use cases run on the Operator track described in the surfaces decision framework — Cowork is the home base, not Code.
  • Dispatch requires Cowork desktop staying open; “Keep awake” is the workaround for laptop-sleep, not a substitute for actually leaving the machine running.
  • Computer Use must be enabled per-session (not always-on) and benefits from Ask-permissions mode while the operator is still building intuition for the workflow.
  • The Mermaid connector for Visualizations export is one of several MCP-style integrations now available via Cowork → Manage Connectors. Worth surveying the connector list once Cowork is set up.
  • “Cloud Code” / “auto-dream” / “co-work” / “Arizona Genius” / “Zona Genius” all appear as auto-caption mishearings in the source transcript — true forms used in this article: Claude Code, auto-dream, Cowork, Your Zone of Genius (the speaker’s app name).

Cost: Cowork is bundled with Pro (100-200/mo) / Team ($20/seat) per Cowork product overview. Computer Use, Dispatch, Memory 2.0, Scheduled Tasks, and Visualizations are all included — no separate metering disclosed by speaker.

Integration notes:

  • This video is a “feature highlight” framed as use cases. Useful complement to the canonical W16 / W17 release-note articles, which document the same features from a release-engineering perspective.
  • Editorial caveat — Rick Mulready’s framing pivots toward a sales pitch for his “AI Playbook” membership and “Your Zone of Genius app” in the closing minute. Standard lead-magnet-funnel pattern (see Brandon Storey’s lead magnet course for the documented version of this template). Treat the use cases as legitimate, the membership pitch as orthogonal.
  • For WEO-Marketly use, the strongest fits look like: Cowork Project for weekly client status (use case 1), Scheduled Task for recurring competitive-intel digests (use case 4), Computer Use for site UX audits at the dental-practice tier (use case 5), and Visualizations for client-facing process explanations (use case 6). Dispatch (2 + 3) is operator-comfort dependent.

Open Questions

  • The video presents Memory 2.0 as default-on across all Pro/Max accounts. Is this actually the case for Team / Enterprise plans, or are there admin-level memory toggles for compliance reasons (HIPAA, SOC2)? Worth checking against WEO AI Policies before assuming auto-memory is on for everyone.
  • “Cowork Projects vs. claude.ai Projects” is a real distinction but the migration story between them is unclear from the video. If a team has 50+ existing claude.ai Projects, is the import-to-Cowork flow lossless? Speaker shows the button but doesn’t run a real migration.
  • The UX audit demo (use case 5) had Cowork retrieve a one-time-passcode from Gmail mid-flow. That’s a useful capability and a meaningful auth boundary — what guardrails exist to prevent Computer Use from authenticating into accounts the operator didn’t intend to touch?
  • Auto-dream’s behavior is described in conceptual terms (“merges duplicates, removes outdated stuff”) but not specified — what’s the actual diff algorithm, what’s evicted vs. kept, and is there an audit log of memory edits the auto-dream pass made?
  • The Cowork Ideas marketplace (use case 7) is a discovery surface — is there an extension model where teams can add private “ideas” for their own organization? Would unlock significant onboarding leverage at WEO scale.

Graph View

Backlinks