PolyAI is ahead on repeatable reliability, task completion, infrastructure reliability, tool call accuracy, interruption handling and response time, and Smallest AI on voice tone and clarity.
Last updated Methodology by Luis Ojeda
Pick PolyAI for reliability across repeated runs (31.70 points higher), completing the caller's task (4.47 points higher), calls that connect and stay up (2.85 points higher), accurate tool calls (0.32 higher), handling interruptions (0.08 higher) and fast replies (0.25s faster). Pick Smallest AI for how the agent sounds (0.40 higher).
Bars share one scale across all 10 platforms. The place next to each figure is its rank in the field.
Share of the 82 scenarios that passed on all three runs.
PolyAI, 31.70 points higher
Calls where the expected outcome was fully reached.
PolyAI, 4.47 points higher
Calls with no connection, audio or platform failure.
PolyAI, 2.85 points higher
Mean score for calling the right tool with the right arguments.
PolyAI, 0.32 higher
Mean score for how clear and natural the agent sounds.
Smallest AI, 0.40 higher
Mean score for yielding and recovering when the caller cuts in.
PolyAI, 0.08 higher
Mean time for the agent to start replying after the caller stops.
PolyAI, 0.25s faster
246 calls per platform: 82 scenarios, 3 runs each.
PolyAI is more reliable, by 31.70 points. PolyAI passed 87.80% of the 82 scenarios on all 3 runs and Smallest AI passed 56.10%. A scenario only counts when every run of it passed.
PolyAI is more reliable, by 31.70 points. PolyAI passed 87.80% of the 82 scenarios on all 3 runs and Smallest AI passed 56.10%. A scenario only counts when every run of it passed.
PolyAI starts replying sooner, 0.25s faster. Mean response time is 2.46s for PolyAI and 2.71s for Smallest AI.
PolyAI reached the expected outcome on 98.78% of calls and Smallest AI on 94.31%. PolyAI is ahead on task completion.
Both ran the same Appointment and Medicare agents through the same 82 caller scenarios, 3 times each, with the same evaluators and mock tools. Only the platform and its speech and model components changed.
Cekura runs these simulated calls on your agent and scores every one.
