Anthropic launches "Review Dashboard": see your interactions with Claude, learn to collaborate smarter with AI

Anthropic Launches Beta "Reflection Dashboard" for Claude – Users Can Review Past Months of Chat Habits in Settings, with Feedback Based on the 4D AI Fluency Framework (Delegation, Description, Discernment, Diligence)

(Previous: From Forcing Employees to Use AI to Worrying About Burning Too Many Tokens – More Companies Are Tightening Internal AI Usage Quotas)
(Context: Anthropic Launches New Slack Integration Tool Claude Tag – Colleagues Can Take Over Your Progress, Team-Shared Memory)

Last night, Anthropic launched a new feature called the Reflection dashboard (currently in beta, only available to Free, Pro, and Max users who have already enabled memory; plans to extend to Cowork conversation history later). It's hidden in the Settings of Claude's web and desktop apps. It doesn't record what you chatted about, but it tells you "what are you using Claude for?"

How often is "just right" to use AI?

Anthropic says this feature originated from three recurring questions in user interviews: How often should I use AI? How can I use it most effectively? Which tasks should I delegate to AI and which should I keep for myself?

After enabling it, users first see a summary covering their most discussed topics, common interaction patterns, and the types of tasks they frequently entrust to Claude.

Scrolling down, you can review conversation activity over the past 1, 3, 6, and 12 months, seeing during which periods you used Claude the most and where you spent your time. Anthropic also announced that a "how much time you spent on Claude" statistic will be added soon.

Beyond statistics, the dashboard periodically presents reflection questions, such as "Is there anything you would still prefer to do yourself, even if Claude could do it faster?" and leaves this question in the dialogue box so you can discuss it directly with Claude – not a one-way report, but a two-way inquiry.

The dashboard also includes two behavior adjustment tools: quiet hours (can set a period of time not to be disturbed) and usage reminders (nudges that suggest you take a break after using it for a while). Both features are voluntary and can be turned off at any time – the decision remains with the user.

To get you to use Claude correctly

The core of the dashboard is calibration. Anthropic packages this logic into the 4D AI Fluency Framework, divided into four dimensions:

• Delegation – deciding whether and how to delegate tasks to AI
• Description – clearly defining your goal to get useful output
• Discernment – judging whether AI's output is usable
• Diligence – taking responsibility for how you use AI

The review report gives specific examples based on these four dimensions, such as "You often modify Claude's draft emails to match your own tone," or "You only delegate tasks after you've thought through the strategy yourself." This feedback carries a clear product intent: rather than letting users use AI blindly, it tells them where they are doing well and where they could improve – for example, suggesting they use the Project feature to avoid repeating background context each time.