Billable work shouldn’t depend on memory.

The problem: work scattered across the day

A call, a meeting, a few client emails, then the next project. When it was time to complete timesheets, people had to reconstruct those fragments from memory. Small pieces of billable work were easy to overlook.

The useful information already existed in Outlook calendars and email. The missing step was bringing it together into something a person could check.

What J2J built

1. Gather the day’s activity. Read calendar and email signals from the tools the team already uses.
2. Suggest the relevant client. Use AI to classify activity and assemble draft time entries.
3. Deliver a nightly report. Give each person a starting point instead of a blank timesheet.
4. Keep the person in charge. Review and correct the suggestions before using them for billing.

What the evaluation showed

That measures classification agreement in one evaluation sample. It is not a measure of time-estimate accuracy, recovered revenue, or a guarantee for future reports. People still need to review the output.

The practical change is a draft to check rather than a week to reconstruct. Calendar and email activity can surface work someone might otherwise forget to record.

What stays a human decision

An email is evidence of activity, not proof of how long the work took or whether it is billable. The team decides the correct client, duration, and billing treatment. This is assistance with reporting, not automatic invoicing.

For a similar project, we scope access permissions, the information sent to AI services, retention, and account ownership before connecting client systems. Using an AI service is different from keeping all processing inside an email account.