AI Insights & Coaching

Deskify's AI microservice analyzes activity patterns across time to identify meaningful signals. It looks at focus score trends (not just current scores but direction and rate of change), app usage patterns (changes in tool usage that might indicate shifting priorities or emerging problems), hour patterns (overwork signals, after-hours activity, weekend work), and team-level patterns (whether individuals are outliers relative to their peers).

Every Monday, the AI generates a weekly digest for each manager. The digest summarizes the previous week for each direct report: focus score and trend, hours worked, notable app usage patterns, and any signals worth the manager's attention. The AI writes this in plain language, contextualizing the numbers rather than just reporting them. Digests are delivered by email (if email reports are configured) and available in the dashboard.

The AI monitors for burnout signals across the team: consecutive high-hour weeks (3+ weeks above the individual's normal), persistent after-hours work, focus score decline over time, and sudden drops in core work tool usage. When burnout signals are detected, the AI flags them in the weekly digest and in the individual user view with a recommendation for the manager to check in.

When the AI encounters applications that have not been categorized, it uses its knowledge of common software to suggest a category and productivity classification. Suggestions appear in Settings > App Categories > AI Suggestions, where admins can approve or reject each one. Approved suggestions are applied immediately. The AI learns from approvals and rejections to improve future suggestions.

Deskify's AI chat (available on Business plan) allows managers to ask natural-language questions about their team's productivity data. Ask "Who had the highest focus score last week?" or "What apps is the engineering team spending the most time in?" The AI retrieves and interprets the relevant data, providing answers with the underlying numbers for verification.

Deskify's AI processes activity metadata — application names, durations, and derived scores — not content. It does not process screenshot images, window content, or any personally identifying content beyond names and roles. AI processing happens within Deskify's infrastructure and is not shared with third-party AI providers. The Gemini model used for categorization receives only application names, never user data.