Negotiation Coach

The call-negotiation-coach skill acts as a pre-call strategy generator and post-call debrief analyzer. It arms AI agents (or human agents) with customized negotiation tactics based on the counterparty's DISC profile and tracks adherence to those tactics in the post-call transcript.

Experimental Scope

This offline reference organizes familiar negotiation concepts into tactic cards and heuristic debriefs. It does not implement a validated negotiation policy, enforce a BATNA, accept deals, or execute financial actions. Research attributions and performance targets are not validation of this prototype; see research scope.

How it works

1. Pre-Call (Prepare Mode)

Generates a structured strategy_card.json containing:

• Goal and BATNA floor threshold.
• A customized sequence of tactics matched to the counterparty's DISC personality type.
• A Dual Concern Mode profile (e.g. Collaborating).

2. Post-Call (Debrief Mode)

Analyzes the transcript against the strategy_card to:

• Score tactic execution (EXECUTED vs SKIPPED).
• Detect Anti-Patterns (e.g., pre-emptive concessions, emotional capitulation).
• Flag if the BATNA was violated (i.e. the agent accepted a deal worse than the walk-away point).

Decision Matrix

DISC Tactic Mapping

DISC Profile	Core Tactics
Dominant	Anchor high, Timed concession, BATNA reference, Conditional close
Influential	Rapport building, Interest exploration, Label emotion, Conditional close
Steady	Rapport building, Label emotion, Timed concession, Interest exploration
Analytical	Interest exploration, Anchor high, Timed concession, BATNA reference

Anti-Patterns Detected

Anti-Pattern	Description
pre_emptive_concession	Conceding before the counterparty responds to the anchor.
positional_bargaining	Arguing strictly on numbers instead of underlying interests.
batna_reveal	Prematurely disclosing the walk-away alternative.
emotional_capitulation	Conceding due to emotional pressure (e.g. "I'm sorry you are angry, let me lower the price").

Expected Outcomes & Metrics

The following numbers are unvalidated aspirations, not measured results or guarantees.

Metric	Target	Notes
Anti-pattern detection TPR	> 85%	Pattern-matching on well-known concession phrases.
BATNA Violation Flags	100% precision	If numerical BATNA is strictly breached, it must flag.

Limitations & Known Constraints

• Heuristic Pattern Matching: The debrief engine relies on regex-based heuristics which may miss nuanced, highly sophisticated phrasing.
• Numeric BATNA Extraction: The _batna_violated function currently only supports percentage-based BATNAs (e.g. "15%"). Currency-based extraction requires future NLP upgrades.