How to run a weekly review with Claude Projects

A weekly review with Claude works best when you set it up as a reusable workflow inside Claude Projects, with clear inputs, structured prompts, and Artifacts for your dashboards.

Claude Projects, launched in September 2024 and expanded in June 2025, were built for this: they hold related conversations, uploaded files, and persistent instructions in one workspace so your weekly review can run the same way, every time.¹ Projects are available across Free (with limits), Pro, Max, Team, and Enterprise, which means individuals and teams can standardise their weekly cadence without jumping tools.¹

How does a weekly review with Claude Projects actually work?

A weekly review with Claude Projects is a repeatable workflow where you connect your logs, metrics, and tasks into one Project, add persistent instructions, and have Claude generate structured insights and a plan each week.²

Modern AI workflow guidance is clear: pick one recurring task, define input/output, connect context, add review points, and iterate after each run.² Weekly review is a near-perfect fit for that pattern — the inputs are your calendar, notes, metrics exports, and task lists; the outputs are summaries, priorities, and next‑week commitments, all stored in one place.²¹

Rather than opening a fresh chat every Friday, you:

• Create a dedicated Project called something like “Weekly Review with Claude”.
• Upload or connect all relevant materials (task exports, meeting notes, KPIs).
• Write persistent instructions that describe how you want Claude to review your week.
• Run the same set of prompts (or a single orchestrating prompt) each week.

This solves the usual problems: scattered context, inconsistent questions, and shallow AI outputs that amount to “you were busy”.

What role do Claude Projects play in weekly reviews?

Claude Projects provide the persistent workspace, context, and instructions that make a weekly review with Claude reliable instead of ad hoc.

Projects are more than folders of chats: they let you upload reference documents, define system instructions that persist across conversations, and cache content in ways that don’t hit per‑message limits.¹ Since their context capacity was expanded roughly 10× in June 2025, you can now keep large multi‑file dashboards — OKRs, logs, meeting notes — inside one Project and reuse them week after week.¹

On Claude Enterprise, Projects use RAG-style retrieval: when your content nears context limits, Claude pulls in only relevant snippets for each query, making reviews over hundreds of thousands of pages feasible.³ In practice, that means you can drop in a year’s worth of reports and have Claude surface what matters for this week without hand‑curating every run.

For a solo workflow, a typical setup looks like:

• One master Project for your weekly review.
• Subfolders or tags for areas of life or work (e.g. clients, products, personal).
• A short “ABOUT ME/” file that describes your role, constraints, and goals — a pattern productivity practitioners using Projects already follow.⁴

The Project then becomes your weekly review “hub” that you return to every Friday, not something you rebuild from scratch.

Which Claude model should you use for a weekly review with Claude?

For most professionals, Sonnet‑class models are the default for weekly review with Claude because they balance reasoning quality, speed, and cost.

Claude 3.5 Sonnet, released in June 2024, introduced a mid‑tier model that outperformed previous flagships at lower cost, making it a practical default for recurring workflows.⁵⁶ A later Sonnet 4.6 is described as “Opus‑level intelligence at Sonnet pricing” and as a “daily driver”, which is exactly the profile you want for a weekly review Project you’ll run dozens of times per year.⁶

Across 2024–2026, Claude has moved from a single chat model with a 9k‑token context window to a four‑tier family (Opus, Sonnet, Haiku, Fable 5) with up to 1M‑token context windows, native vision, tool use, computer use, and server‑side memory.⁷ That evolution is what makes deep weekly reviews possible: you can now include multi‑source project data, screenshots of dashboards, and longer narrative reflections without hitting a hard wall.

In practical terms:

• Default to Sonnet 4.6 for weekly review Projects.
• Use Opus if you’re doing heavier strategy or multi‑quarter analysis during the review.
• Drop down to Haiku only for quick pre‑processing of large data exports.

What inputs and outputs should a weekly review with Claude produce?

The weekly review with Claude should ingest your logs, metrics, and tasks, then output summarized insights, priority lists, and a realistic plan for the coming week.

Typical inputs include:²

• Calendar exports (or screenshots) of your meetings and time blocks.
• Notes from 1:1s, project discussions, and personal reflections.
• Metrics dashboards exported as PDFs or CSVs — e.g. revenue, sign‑ups, shipping cadence.
• Task lists from tools like Todoist or Asana.

For example, one user runs a Claude Project that plans their day, tracks habits, and handles daily/weekly reviews by integrating Todoist and Google Calendar; Claude acts as a task auditor, habit scheduler, and schedule composer.⁸ Your weekly review can piggyback on that same data.

The outputs should be similarly concrete:²

• A short narrative summary of the week.
• A priority list with 3–7 items for next week.
• A plan that respects your actual constraints, not your ideal self.

This is where Artifacts matter.

How do Artifacts turn your weekly review into a dashboard?

Artifacts let you turn the weekly review with Claude into an editable dashboard or scorecard instead of a transient chat transcript.

Artifacts — interactive documents, tables, and diagrams — ship inside Projects on all paid plans as of April 2026.¹ They are ideal for weekly reviews because they can render task tables, KPI charts, and routines that you update each week inside the same Project.¹

You might have Claude create an Artifact that includes:

• A Task table: item, status, owner, priority, notes.
• A KPI snapshot: current value vs target, trend arrows.
• A Habit/routine tracker: days hit, blockers, tweaks.

Each Friday, you:

• Upload fresh metrics or confirm they’re synced.
• Ask Claude to refresh the Artifact based on the latest inputs.
• Make manual tweaks where needed, then save.

Because the Artifact lives inside the Project, the history of your reviews accumulates into something like a lightweight “business pulse” system — mirroring pre‑built weekly workflows that are emerging inside Claude Enterprise agents.⁹³

How do you design a weekly review with Claude workflow that actually improves productivity?

A weekly review with Claude meaningfully improves productivity when you explicitly redesign the workflow around human–AI collaboration instead of bolting AI onto an existing checklist.

Research on AI workflows suggests that redesigning processes around human–AI collaboration can produce 3× better results than simply adding AI as a helper.¹⁰ Implementation quality determines whether you see around 5% vs 40% productivity gains, which is a stark reminder that “ask Claude for a summary” is not a workflow.¹⁰

For weekly review, that means:

• Mapping the steps: capture, reflect, analyse, decide, schedule.
• Deciding where Claude leads (analysis, pattern detection, synthesis).
• Deciding where you lead (value judgments, commitments, trade‑offs).

The MIRROR system described in one practical guide — Map, Interrogate, Reality‑check, Reveal, Optimize, Reset — is an example of a structured six‑stage Claude review that takes ~30 minutes weekly.¹¹ Each stage is a distinct prompt run in sequence within a single conversation, building context as you go.

Example weekly review with Claude prompt flow

Inside your Project, you might run a sequence like:

1. Map – “Summarize what I worked on this week across projects, meetings, and deep‑work blocks, using the uploaded calendar, notes, and task exports.”
2. Interrogate – “Compare what I planned vs what I executed; highlight gaps and plausible reasons using patterns in the notes.”
3. Reality‑check – “Given my constraints and energy patterns, what seems sustainable? Where am I overcommitting?”
4. Reveal – “Identify one underlying signal from this week that I may be ignoring.”
5. Optimize – “Turn this into three specific behaviour commitments, one stop‑doing, and one thing to protect.”
6. Reset – “Render next week’s priorities and schedule as an Artifact table I can adjust.”

Run these in the same Project each week to build a consistent review muscle.

How can you automate or schedule a weekly review with Claude?

You can automate a weekly review with Claude by pairing Projects with scheduling tools like Claude Cowork or Claude Code, so the review kicks off at the same time every week.

In practice, users already schedule weekly reports that pull from Microsoft 365, CRMs, and notes using Claude Cowork: they create a Project, define a rules prompt, and then set a recurring task that references the Project to generate a draft report every Friday.¹² Cowork provides scheduled tasks and task history tied to Projects, with a “weekly on Friday at 5pm” cadence that runs as long as your desktop is open.⁴

Another route for technically inclined users is Claude Code. Skills like the Weekly Review skill automatically scan workspace files — PRDs, meeting notes, daily plans — and synthesize weekly progress into a structured report.¹³ You can also use the Knowledge Base Builder skill to turn recurring review questions and artifacts into FAQs and scripts Claude can reuse.¹⁴

A simple, low‑friction setup:

• Schedule a 30‑minute slot on your calendar.
• Use Cowork to trigger a starter prompt and pull that week’s files into the Project.
• Let Claude generate the first pass, then you edit the Artifact and commit.

How does a weekly review with Claude compare to manual or other AI workflows?

Weekly review with Claude differs from manual review and generic AI summaries by offering persistent context, structured outputs, and deeper pattern detection.

AI project‑management tools like ClickUp and Asana already emphasise weekly summaries and automated updates, showing the demand for recurring review workflows.¹⁵ But they typically summarise activity inside their own ecosystem. A Claude Project can sit above multiple tools — calendars, docs, CRMs — and analyse the whole picture.

Guidance on AI productivity stresses moving beyond single‑step use (e.g. summarising memos) towards end‑to‑end workflows.¹⁶ Weekly review with Claude is a canonical example of this shift: one Project, many inputs, several reasoning stages, and concrete outputs.

Here’s a comparison at a glance:

Approach	Pros	Cons
Manual weekly review	Deep personal insight; no setup	Time‑consuming; easy to skip; limited perspective
Simple AI summary in chat	Fast; no configuration	Shallow; little pattern detection; no persistent context
Weekly review with Claude Project	Structured, repeatable workflow; cross‑tool context; Artifacts dashboards	Requires initial setup and light maintenance

If you’re running a business or leading a team, there is also ecosystem support: Claude Enterprise now ships managed agents and pre‑built “business pulse” weekly workflows that mirror this pattern and scale it across departments.⁹³

What are the common mistakes to avoid when setting up a weekly review with Claude?

The biggest mistakes with weekly review with Claude are treating Projects like simple folders, asking only for a summary, and ignoring context design.

First, Projects are not just collections of chats; they provide persistent instructions, cached content, and retrieval‑based reasoning across files — all of which you should explicitly use in your workflow.¹ Second, simply asking Claude for a weekly summary tends to deliver marginal (~5%) productivity gains, whereas structured workflows with clear inputs, checkpoints, and outputs can reach ~40% gains.¹⁰¹⁶

Third, context limits are often overstated: with Projects’ expanded capacity and modern Claude models offering up to 1M tokens plus RAG retrieval, detailed weekly reviews over multi‑source knowledge are not only possible but practical.¹⁷³

The antidote is simple:

• Treat the weekly review with Claude as a design problem, not just a prompt.
• Start small, on one area of work, and iterate monthly.
• Let Claude handle the pattern detection and synthesis, while you stay responsible for judgment and commitments.