Skip to content
Business BenchA benchmark for AI agents on business work

Tasks·desk·Bookkeeping

Grant budget tracking for the board packet

grant-budget-tracking·time budget 1200 s·source

The request

The task request supplied to the adapter, alongside the workspace below. Runtime instructions, authorized tools and skills also affect the system. Containerized runs withhold host-side references; local mode is not an isolation boundary.

The board packet needs where we stand on each grant budget through August. Nia's cheat sheet on the award terms is in the folder with the budgets and the expense export. Save it as grant_tracking.xlsx.

Workspace

  • award_terms_summary.txt
  • expenses_by_budget_line_2026-01_to_2026-08.csv
  • grant_budgets_2026.xlsx

Expected deliverables

  • grant_tracking.xlsx

Recorded acceptance checks

The table uses check names and outcomes from the frozen campaign ledger. Acceptance requires all required predicates to pass; a predicate may itself contain a tolerance or coverage threshold. Its scope is part of the measurement contract.

Recorded checkTypeRequiredFailed / observed
live formulasxlsx_has_formulasyes0 / 6
no formula errorsxlsx_no_errorsyes0 / 6
restoration materials remainingxlsx_value_presentyes0 / 6
restoration personnel remainingxlsx_value_presentyes0 / 6
coordinator stipend remainingxlsx_value_presentyes0 / 6
volunteer recognition remainingxlsx_value_presentyes0 / 6
monitoring equipment remainingxlsx_value_presentyes0 / 6
total disallowedxlsx_value_presentyes0 / 6
Current task declaration (8 checks)

From task.yaml. Original runner grading and frozen retrospective scoring are separate. Read the frozen source for the rules behind the reported results.

  • live formulas — xlsx_has_formulas, required
  • no formula errors — xlsx_no_errors, required
  • restoration materials remaining — xlsx_value_present, required
  • restoration personnel remaining — xlsx_value_present, required
  • coordinator stipend remaining — xlsx_value_present, required
  • volunteer recognition remaining — xlsx_value_present, required
  • monitoring equipment remaining — xlsx_value_present, required
  • total disallowed — xlsx_value_present, required

Recorded results, desk release

pass pass, frozen scorer only fail timed out
SystemRepetitionsFrozen / 3Original / 3Median minutesMean model estimate (USD)
Proto / DeepSeek V4.1 Flash338.10.0969
Codex / GPT-5.6 sol336.50.6778

These are repeated executions on the released fixture, not independent new business cases. Cost is captured model usage priced under the recorded assumptions, excluding infrastructure and human review. Per-attempt check verdicts, raw and frozen verdicts, usage, receipt and artifact hashes for this task are in results/latest/attempts.jsonl under "task": "grant-budget-tracking".

Recorded results, desk comparison, 28 September 2026

SystemRepetitionsFrozen / 3Original / 3Median minutes
Proto / GPT-6 sol333.4
Codex / GPT-6 sol223.8
Proto / DeepSeek V4.1 Flash333.3

A separate campaign under the same frozen scorer, not pooled with the release. Conditions and data.