The Practical AI Benchmark

The test battery

5 tasks from ordinary working days, each aimed at one way AI commonly fails. Critical checks are errors that could materially affect the usefulness or correctness of the result; the rest are standard checks.

1. Task 01Everyday productivity

   Messy brain dump → realistic day plan

   Turn 18 unordered to-dos into a plan for one 8-hour day, with breaks and honest time estimates.

   Watching for

   Cramming more work into the day than fits, or silently dropping tasks.

   Checks

   4

   Critical

   3

   Standard

   1

   View scoring criteria

   • Planned work fits 09:00–17:00CriticalInterpretive review
   • Every task placed or explicitly deferredCriticalInterpretive review
   • 30-minute lunch and at least one short breakStandardInterpretive review
   • Fixed 14:00 dentist keptCriticalInterpretive review
2. Task 02Everyday productivity

   Email and chat catch-up

   Read 20 scattered messages and return the decisions, the action items and who owns each one.

   Watching for

   Missing an action item, inventing an owner, or reporting an open idea as decided.

   Checks

   4

   Critical

   4

   Standard

   0

   View scoring criteria

   • All 4 action items foundCriticalInterpretive review
   • Owners match the sourceCriticalInterpretive review
   • No invented deadlinesCriticalInterpretive review
   • Open proposal not reported as decidedCriticalInterpretive review
3. Task 03Communication

   Polite refusal under pressure

   Decline a demanding customer within company policy, in under 150 words.

   Watching for

   Over-apologising, promising things outside policy, or running long.

   Checks

   4

   Critical

   2

   Standard

   2

   View scoring criteria

   • ≤ 150 wordsStandardCode checked
   • No refund or free express promisedCriticalInterpretive review
   • Only offers what policy allowsCriticalInterpretive review
   • Clear next step offeredStandardInterpretive review
4. Task 04Operations

   Receipts → clean table

   Turn 12 messy receipts into a fixed table with VAT split out and categories assigned.

   Watching for

   Wrong arithmetic, a changed table layout, or receipts that never existed.

   Checks

   5

   Critical

   3

   Standard

   2

   View scoring criteria

   • Exact columns, in orderStandardCode checked
   • Each of the 12 receipts appears exactly once, none inventedCriticalCode checked
   • Net and VAT correct for every receiptCriticalCode checked
   • Gross total stated after the table = 12 674.50CriticalCode checked
   • Every category is one of the four allowedStandardCode checked
5. Task 05Analysis

   Contradiction spotter

   Compare a supplier's formal quote with their sales email and flag every place they disagree.

   Watching for

   Smoothing over discrepancies or quietly picking one version.

   Checks

   5

   Critical

   4

   Standard

   1

   View scoring criteria

   • Price discrepancy flaggedCriticalInterpretive review
   • Delivery-date discrepancy flaggedCriticalInterpretive review
   • Installation-scope discrepancy flaggedCriticalInterpretive review
   • No invented resolutionCriticalInterpretive review
   • Both sources cited for eachStandardInterpretive review