The challenge
Weekly reporting often involves opening several tables, comparing periods, and turning colleagues' notes into a narrative. When figures and commentary are separate, the explanation becomes difficult to trace.
This scenario assumes defined metrics and an established aggregation process. Start with one recurring meeting and one report format.
Information to prepare
Define the source owners, document dates, and permitted readers before choosing the retrieval scope.
| Input | What to check |
|---|---|
| Aggregated data | Check reporting periods, filters, and refresh times |
| Team notes | Review activities and operational events |
| Report template | Agree on headings and the expected level of detail |
| Metric definitions | Verify units, coverage, and comparison conditions |
The workflow
Adapt the following responsibilities to the team's existing workflow.
| Step | AI and system role | Human role |
|---|---|---|
| Prepare the information | Match relevant tables and notes | Check periods, units, and missing data |
| Surface changes | Present calculated differences alongside related notes | Validate calculations and comparison conditions |
| Draft the report | Separate facts, reported context, and unconfirmed hypotheses | Decide which explanations and discussion points are supported |
| Prepare for the meeting | Retain sources and the edit history | Approve the report and next actions |
An example output
The following example was written around fictional source material. It is not customer data or output from a deployed system.
| Field | Example content | Source or status |
|---|---|---|
| Observed fact | Request volume increased compared with the previous week | Sample summary table |
| Team context | An information page was updated this week | Sample operations note |
| Unconfirmed explanation | Whether the page update caused the increase remains unknown | The available material does not establish causation |
| Discussion point | Review referral sources and request categories | Suggested next investigation |
Human review and operating conditions
Calculations remain in a verifiable aggregation process; AI drafts explanations of the results. Coinciding events are not automatically treated as causes. Retain reporting periods and refresh times, with appropriate access restrictions.
Evaluate a narrow pilot
A possible initial trial covers two weekly reporting cycles. Compare preparation, review, and correction effort with the current process, and check for incorrect figures and unsupported explanations.
Two cycles is an initial evaluation scope. Include unusual or missing data before deciding whether the workflow is repeatable and whether review effort is acceptable.
| Measure | What to record |
|---|---|
| Preparation time | Data checks, drafting, and correction effort |
| Accuracy | Agreement on figures, periods, and units; unsupported causal claims |
| Repeatability | Whether another reviewer can check the report and repeat the process |
Explore the next step with Atlas Support
Describe the report fields, data sources, and preparation steps. We can help distinguish calculations handled by existing tools from explanations that AI can assist with.
Depending on the agreed scope, we can map the workflow, review data and access, build a prototype, and define evaluation criteria. The evidence then informs the operating scope and remaining issues.
Discuss this workflow
Tell us about one recurring task and the types of documents your team uses. We can help define the scope and the evidence needed for a first evaluation.