When does the report arrive?

On Monday morning, covering the seven days that ended the day before. It arrives by email and as an in-app notification linking to the analytics screen.

Does the AI compute any of the numbers?

No. The numbers are computed from the daily rollup first. The AI is given those numbers and asked only for the sentences and three suggested actions.

What if the narrative fails?

The email waits for the next run and the in-app entry still arrives with the headline numbers.

Can the report and the analytics screen disagree?

No. Both read the same rollup rows through the same queries.

The Monday morning owner report

Every Monday morning Tasked sends the owner a report of the past seven days: tasks completed, open and overdue now, hours waiting on clients, the three slowest stages, the three projects waiting longest on a client, the three busiest people and the three riskiest projects. The numbers come from the same daily rollup the analytics screens use. The AI adds three to five sentences and three suggested actions, and it is never asked for a number.

Published September 15, 20265 min read

What is in it

The report covers a seven-day window ending the day before it is sent. Its facts are fixed; the same eleven things arrive every week, in the same order, so the owner learns to read it in a few minutes.

Tasks completed this week.
Open now, and overdue now.
Hours spent waiting on clients.
Mean cycle time for the week, or a note that it was not measured.
Days with no data, if any, so a quiet number is not mistaken for a good one.
The three slowest stages with their median time in stage.
The three projects that waited longest on a client, in minutes.
The three busiest people, with their workload in minutes.
The three riskiest projects, with score and band.

Where the numbers come from

A daily job rolls every organisation's events into one row per day for the organisation, for each project and for each member, and scores each project for late-delivery risk. The analytics screens read those rows. The weekly report reads the same rows, through the same five queries the screens use: portfolio, bottlenecks, clients, load and risk.

That is the point of the design. The report and the screens cannot disagree, because there is one computation and two readers. If the report says 14 completed, the portfolio tab says 14.

What the AI is allowed to do

After the facts are assembled they are handed to the AI as a list of lines, one number each, together with two rules: use only the numbers above, and invent nothing. It returns three to five sentences and exactly three suggested actions, most useful first, each with a one-line reason tied to one of the listed numbers.

The AI is never asked for a number, so a made-up figure has nowhere to enter. It writes in the recipient's language, Hebrew or English. The AI reaches the organisation's data only through the same interface every other AI feature uses, under the acting user and the organisation's own budget.

If the narrative fails to generate, the report is not sent as a table of numbers with a blank where the words should be. The email waits for the next run, and the in-app entry still lands with the headline numbers, so the owner is never without the facts.

By email and in the app

The email carries the organisation name, the date range, the four headline numbers, the narrative, the three actions and a button to the analytics screen. The in-app notification carries the date range as its title, the narrative as its body, and opens the same screen.

Both are keyed to the organisation and the week, so a rerun of the job does not produce a second copy. The scheduled job runs on Monday morning, and the same code path can be run on demand.

Reading it in five minutes

Read the four headline numbers against last week's. Then read the three lists from the bottom up: the riskiest projects first, because that is where a client call may be due; the busiest people next, because that is where a reassignment helps this week; the slowest stages last, because those are process questions for the month, not the morning.

The three suggested actions are a starting point, each pinned to a number. Take one, ignore two, and the report has done its job.

Headline numbers: compare with last week.
Riskiest projects: who needs a call today.
Busiest people: what to move this week.
Slowest stages: what to raise in the process audit.

What it is not

The report is not a dashboard you configure; its shape is fixed and it draws the same five views the analytics screen already shows. It is not a replacement for opening the boards; it tells you which board to open first. And it is not a forecast: every number in it is a count, a total, a median or a stored risk score, all of which you can find on the screens it links to.

Questions that keep coming up

When does the report arrive?: On Monday morning, covering the seven days that ended the day before. It arrives by email and as an in-app notification linking to the analytics screen.
Does the AI compute any of the numbers?: No. The numbers are computed from the daily rollup first. The AI is given those numbers and asked only for the sentences and three suggested actions.
What if the narrative fails?: The email waits for the next run and the in-app entry still arrives with the headline numbers.
Can the report and the analytics screen disagree?: No. Both read the same rollup rows through the same queries.
Which language is the report in?: The recipient's language, Hebrew or English.