Those principles pointed at consumer marketplaces — where AI-originated demand is already landing and the platform has an obvious P&L line tied to getting the handoff right. Somewhere between 10% and 30% of re-book, match-conversion, and completion-rate leaks at that handoff today. 60–90 day pilots, priced against the platform's own metric.
Three categories cleared all five principles; each stretches a different part of the problem.
Does context-portability matter most when the information is emotionally loaded and hard to summarize? Parents carry the most to the conversation — allergies, routines, anxieties, a specific caregiver philosophy. The leak is also most expensive here: a bad match is a safety incident, not a missed conversion.
Does the same hold for operational context — property history, job specs, schedule constraints? Homeowners carry things like the furnace's short-cycling history, HOA rules, the apartment's ceiling height. This is also the category most aggressively shipping agent integrations (ChatGPT apps, Alexa+, Operator), so the gap widens fastest here.
Does the same hold when the agent is already part of the service — an AI tutor handing off to a human tutor? Preply runs that handoff in production today. That's where we can actually measure impact.
Ordered by decision-window closing speed.
| # | Account | Lead | Decision-window anchor |
|---|---|---|---|
| 01 | Care.com | Brad Wilson CEO |
Pacific Avenue $320M PE close +32 days; Year-One deck being drafted this week. |
| 02 | Angi | Jeff Kip CEO |
Q1 earnings May 5–6; Feb 11 LLM-handoff quote on the record. |
| 03 | Thumbtack | Chris Patalano CTO |
6 months in-seat, scope-what-to-build window still open. |
| 04 | Preply | Dmytro Voloshyn CTO, co-founder |
$150M Series D architecture being written now. |
| 05 | TaskRabbit | Ania Smith CEO |
Alexa+ shipped NRF 2026; Tasker-briefing upgrade not on the roadmap. |
| 06 | UrbanSitter | Lynn Perkins Founder-CEO |
Post-Kinside multi-modal matching; founder-direct decision loop. |
| 07 | Wyzant | Paul Mishkin CEO, IXL Learning |
MyTutor UK acquisition opens a cross-marketplace integration lever. |
| 08 | Rover | Brent Turner CEO |
Dec 2025 SF sitter incident + Blackstone PE-exit narrative. |
| 09 | Outschool | Xueyan Wang COO, co-lead |
Membership pivot + Feb 2026 CEO transition. |
| 10 | Handy | Umang Dua CRO, ANGI Services |
Companion pilot on the Angi conversation. |
Three waves, 3-4-3, one vertical per wave. Vertical grouping isn't theme — it's learning: a Care.com reply tightens UrbanSitter's send, an Angi reply tightens TaskRabbit's.
| Wave | Vertical | Days | Accounts | Why this wave here |
|---|---|---|---|---|
| 01 | Care | T+1 → T+4 | Care.com · UrbanSitter · Rover | Tightest decision window; Care.com leads the sequence. |
| 02 | Home | T+6 → T+9 | Angi · Thumbtack · TaskRabbit · Handy | Highest signal density; voice calibrates on Wave 1 replies before the biggest sends. |
| 03 | Tutoring | T+11 → T+14 | Preply · Wyzant · Outschool | Clearest live handoff (Preply's AI Tutor → human tutor); benefits from earlier waves. |
A wedge-reframe. The pain is vetting and capability-match, not context loss between agent and human. Pitch is quality/safety, not context-bridging. If the reframe doesn't land in 30 days, drop to Q2.
Weakest against the five principles. CEO transition + membership pivot = thin internal bandwidth. Kept because live-synchronous handoff is a shape we need in the first ten. First-reply signal decides whether it stays through Wave 3.