Real sales. Real money on the line. Labeled for AI training.

Not synthetic. Not scraped. Actual humans selling actual things, with the dollar outcome attached to every dialog.

Why it's different

The reward signal is built into the data. Money in, money out, every time.

Most training data was written by people who've never closed a sale.

Ours wasn't. Every dialog ends with a real payment. The revenue is the label. No annotator opinion, no proxies, just what actually happened.

Built for training, not for reading.

Tactics taxonomy, buyer profiles, objections, conversion turns, outcomes, skill tiers (including the bad ones). 24 fields per dialog.

You can't get this anywhere else.

The big data labelers won't touch this kind of content. We've checked. That's the whole reason this niche is wide open.

What's inside

24 fields per dialog. Drop in and start training.

Annotated dialogs, full datacard, schema docs. Works for SFT, RLHF, DPO, or as the foundation for an RL Gym.

outcome revenue_usd tactics_used[ ] conversion_turn user_profile{ } dialog_id team_id skill_tier sale_type turns_total time_to_conversion_min user_initiated_purchase model_initiated_sale upsell_attempted upsell_success objections[ ] price_negotiation emotional_dynamic key_moments[ ] model_strategy_annotation conversion_catalyst notes

SFT

Sales agent fine-tuning

Real persuasion sequences with known outcomes. Tactics and buyer profiles as structured labels.

RLHF / DPO

Reward modeling

Use revenue_usd directly as your reward signal. Compare wins against losses without asking an annotator what's "good".

RL Gym

Simulation environment

Buyer profiles and turn-level events let you build a gym where your agent plays seller and gets paid in dollars.

Get access

Request a demo or ask a question.

We'll send the demo pack with annotated dialogs, full datacard and schema docs, under NDA. No pitch deck, no sales call required to start.

// Evaluation → Commercial licensing available

✓

Got it. We'll be in touch within 24 hours.