The numbers from the pilot
Not a projection — a controlled pilot with real field reps using the Sales Agent in their genuine visit-prep routine. The figures below are the only proof points we cite, and they come straight from that pilot:
- 91% peak answer accuracy, reached in the pilot's final stages.
- 100% of participating reps agreed it can support their role and save visit-prep time.
- A measurable lift in pre-visit effectiveness, with fewer information gaps.
What the pilot tested
Reps used the agent during real visit preparation — not a lab. It combined Large Language Models with a structured-data retrieval layer, so a rep could ask a plain-language question and get a precise answer back. Response-retrieval accuracy was tracked as the model was refined, peaking at 91%.
What reps experienced
More effective visit prep with less manual data aggregation. Higher confidence walking into outlets. Fewer information gaps. And unanimous endorsement — every participating rep agreed it has the potential to support their role.
What it means for scaling
A pilot that earns unanimous endorsement and 91% accuracy has cleared the two hardest barriers to scale: trust and adoption. From here, expansion is a question of disciplined rollout, not persuasion. The commercial targets — higher Average Order Volume and reduced Out-of-Stock — are what scaling is designed to deliver; they are goals, not pilot results.
The flagship build The AI Sales Agent — proven with real field reps → 91% answer accuracy · 100% rep endorsement, in a controlled FMCG pilot.