TalkRepair
A real-time co-pilot for managers about to have a hard conversation (a performance review, a termination, a peer conflict) that listens in, flags when the other person is shutting down or escalating, and quietly suggests a concrete repair move.
first-time and middle managers who dread difficult 1:1s, performance reviews, and conflict conversations
- Live conversation monitoring (with explicit consent prompts) that detects divergence spikes before they escalate
- Pulls up context: previous 1:1 notes, personality hints, last agreed action items
- Suggests specific repair moves phrased in the manager's own voice and tone
- Post-call debrief with what worked, what missed, and a draft follow-up note
Hybrid and remote work has stripped managers of the casual cues they used to read in person, while layoffs, performance plans, and conflict cases have all spiked post-2024.
Manager burnout, post-2024 layoff/performance-plan spike, and hybrid-work loss of in-person cues are well-documented pain points; demand likely high but could not verify with live web tools.https://www.google.com/search?q=AI+conversation+coach+manager+difficult+conversations+real-time ↗
Adjacent categories are crowded (BetterUp/Lattice/CultureAmp for coaching, Gong/Chorus/Read.AI for call intelligence, Otter/Copilot for meeting AI) but a true real-time 'dialogue repair' co-pilot aimed at hard 1:1s appears less directly served; could not confirm with live searches.https://www.google.com/search?q=AI+conversation+coach+manager+difficult+conversations+real-time ↗
B2B SaaS pricing of $20-50/seat/month is plausible given manager coaching market (BetterUp enterprise contracts), but recording conversations introduces consent/legal friction that suppresses willingness-to-pay; no live pricing evidence obtained.https://www.google.com/search?q=AI+conversation+coach+manager+difficult+conversations+real-time ↗
Hard workplace conversations are evergreen, and LLM real-time reasoning/coaching capability is improving each model cycle — durable category if privacy is solved.
Real-time audio capture + emotion/escalation detection + sub-second suggestion latency is technically demanding; significant two-party-consent legal exposure (CA/EU) and trust barrier of 'boss is running AI on me' are major blockers.