Analytics with an agenda (two of them, actually)
Anthropic rolled out Reflect, a dashboard in Claude's settings that turns chat history into a usage report - part Spotify Wrapped, part screen-time tool. It summarises the last one, three, six, or twelve months: most-discussed topics with percentages, the kinds of tasks you delegate, your most active day and peak hour, all visualised. A recommendation layer then coaches you using the 4D AI Fluency Framework Anthropic co-created with academics - delegation, description, discernment, diligence - with concrete suggestions like grouping repeated-context prompts into Projects, or generating a custom skill. One early reviewer's research pattern prompted Claude to build a fact-checking template listing sources, confidence levels, and caveats. The beta is live for Free, Pro, and Max users with Memory enabled on web and desktop, with mobile and a time-spent metric to follow.
The wellbeing layer is real - and carefully built
Reflect ships with quiet hours, schedulable break nudges, and periodic reflection questions such as what you would want to keep doing yourself even if Claude were faster. Anthropic built the sensitive parts with the MIT Media Lab's AHA programme, Boston Children's Hospital's Digital Wellness Lab, and the Family Online Safety Institute. Incognito chats and connected-tool source files stay out, conversations tied to health integrations are excluded entirely, and the company says insights are not reused elsewhere. Tellingly, Anthropic had not tracked time-spent internally because its product team did not want to maximise it.
The sharper read
TechCrunch's framing deserves its own paragraph: at a moment of AI backlash, a dashboard laying out everything Claude helped you accomplish is quietly persuasive - it frames the product as indispensable, and its suggestion engine recommends only Anthropic features, never alternatives. The historical rhyme it cites is Gmail Meter circa 2012: analytics as sentiment-shaping. Both readings are true at once, which is what makes it worth studying.
What to take from it
- If your team uses Claude, Reflect is a free AI-adoption audit: the task-type breakdown shows where AI already carries load and where usage stays shallow - useful evidence for training plans and licence decisions.
- Product leaders should treat this as a pattern to copy: usage dashboards pairing self-quantification with feature recommendations drive retention while wearing a wellbeing halo. If you run a SaaS product, this is your benchmark for doing it - and disclosing it - credibly.
- Governance note: usage analytics derived from employee chats, even privacy-scoped ones, deserve a policy line. Decide whether staff dashboards like this are personal or manager-reviewable before someone asks.