Independent coverage of AI agents and the work they do
Est.
Testing agent claims against inspectable outcomes.
12 reports
Compare 46 receptionist products and regional offers by documented workflows, plan conditions, human fallback and billing units. Includes source-linked scores and downloadable data.
Signals and Simulations enter public preview, connecting production traces to candidate tests. Supported harnesses and mocked actions define the limits.
Wonder reaches the API and Muse Code; Full requires a Power or Maximum subscription. The broader Muse rollout is still ahead.
MASS reports better research scores per output token after two training cycles. Software benchmarks expose the limits of that improvement.
The new model scores permitted choices from mixed media. Its documented context limits and uneven benchmark results matter before those choices trigger actions.
Project Beacon aims to flag risky behavior during execution. The buying question is whether earlier detection produces an alert your team can act on.
Test invoice preparation with sample job records, duplicate checks and a human approval step before connecting an agent to real billing.
Compare recovered work with the full bill and supervision time. Use a downloadable checklist to test the paid plan, bookings, human handoffs and rollback.
A practical first workflow for small teams: assign ownership, prepare AI drafts, check for new messages and test handoffs before allowing automatic sends.
Give a small-business agent a named owner, backup, change log and pause procedure. Includes a handover card and maintenance test sheet.
Count setup, checking, repairs and software costs before calling an AI pilot a success. Includes a blank ledger and a worked example.
Our testing method: a real task, explicit permissions, a result you can inspect, and a record of the work still left to you.