Notifications and AI Insights
How ProLens generates AI insights, how to train them, and how to turn them off.
ProLens uses Anthropic's Claude to read your operational data and surface the single most important finding per domain. You see the result; the AI sees an anonymized view.
What gets analyzed
A separate AI pipeline runs for each domain:
- Revenue Pulse - daily revenue briefing
- KPI Annotations - one-liner per KPI tile
- AR Collections - outstanding-receivables narrative
- Budget Health - over-budget detection
- Project Profitability - margin anomaly detection
- Timesheet Compliance - submission patterns
- Client Activity - dormancy and concentration alerts
- Market Insight - macro context for your pipeline
Each runs on its own morning schedule (see Getting Started).
Anonymization
Before any data leaves ProLens for the AI, all client names, project
names, and employee names are replaced with opaque tokens (Client A,
Project Alpha, Person B). The AI sees the structure of your data
(counts, dollars, ratios) but never the actual names. The response is
deanonymized on the way back so you read real names on the dashboard.
This is enforced in the data pipeline, not a policy. The AI literally cannot link a number to a real client.
Training the AI
Every AI insight has a thumbs-up / thumbs-down at the bottom. Click thumbs-down to open a feedback panel where you can:
- Tell us what is wrong (categorized: amounts wrong, timing wrong, missing entity, irrelevant entity, tone, recommendation)
- Paste what it should say instead
- Apply just to this widget or all widgets that reference the same data
Admins can preview the rewritten output before applying it, roll back to any previous version, and see the full version history in Settings > AI Intelligence.
Refresh on demand
Every AI insight has a small sparkle/refresh icon. Click it to regenerate manually. Rate-limited to three manual refreshes per widget per day.
Turn it off
In Settings > AI Intelligence (AI Configuration), each AI feature has its own on/off toggle. Disabling a feature stops the AI pipeline for that widget entirely. The widget keeps showing your data, it just will not include the AI insight.
Some features are tier-gated (Widget Training, Weekly Pulse, Market Insight require Practice or higher). Locked features show a "Requires Practice plan" badge on the toggle.
Cost
ProLens covers the AI cost on every paid plan. There is no per-token billing or usage cap.