Synthetic workload benchmark
Meeting Intelligence Benchmark Methodology
A transparent task-count comparison using the fixed public demonstration scenario. It measures handling surfaces and active synthesis workload, not elapsed human time, dollars saved, or production AI performance.
Benchmark design
Count the work the coordinator must actively assemble or resolve.
The benchmark uses the exact counts exposed by the browser demonstration and applies the same definitions to both routes.
| Benchmark area | Manual baseline | Assisted workflow | Calculated change |
|---|---|---|---|
| Contributor/status checking | 6 source or status checks | 1 consolidated contributor view | 83.3% fewer status-check surfaces |
| Active synthesis items | 10 items assembled or reconciled: 7 routine facts + 3 exceptions | 3 exceptions requiring active resolution | 70.0% fewer active synthesis items |
| Output construction | 3 separate artifacts drafted | 1 coordinated finalization step produces all 3 outputs | 66.7% fewer separate construction steps |
| Judgment-bearing exceptions | 3 reconciliations | 3 explicit review decisions | Judgment is retained, not automated away |
Important boundary: this is a synthetic task-count benchmark. It does not claim measured organizational time savings, financial ROI, production accuracy, or client performance. Routine facts prepared by the workflow remain subject to accountable verification under the organization’s controls.