Voice AI Agents Struggle to Convert Leads, Forbes Explains

A Forbes Technology Council op-ed by Chao-Ping Wu, published July 7, 2026, argues that voice AI agents are failing to convert real estate and mortgage leads not because of weak language models but because of operational gaps: the piece cites internal platform data showing it takes an average of 4.7 attempts before a lead answers the phone, with nearly three of those attempts often happening on day one. Wu contrasts current voice AI with earlier stopgaps like offshore call centers and frames the near-term future as a hybrid of human and AI agents rather than full automation. For practitioners, the piece is a reminder that persistence, scheduling, and reliable handoffs to humans often matter more for ROI than incremental gains in speech or intent-recognition accuracy.
The useful signal here is not a new model or benchmark, but an operational framing: for phone-based voice agents, the bottleneck described is contact persistence and handoff design, not raw conversational quality.
What happened
Forbes Technology Council published an op-ed by Chao-Ping Wu on July 7, 2026, examining why AI voice agents underperform in outbound customer outreach, with a focus on real estate and mortgage lead conversion. Citing data described as drawn from "millions of calls processed across our platform," the piece reports it takes an average of 4.7 attempts before a lead answers the phone, with nearly three attempts often occurring on the first day alone. The article contrasts current voice AI approaches with earlier stopgaps such as offshore call centers, which it says introduced management overhead, telephony costs, staff turnover, and training burden.
For practitioners
The article's operational framing points to two practical failure modes for voice agents: insufficient persistence and scheduling logic to reach transient contacts, and brittle intent understanding or domain gaps that break handoffs. It suggests engineering priorities should include scalable telephony integration, retry and queuing logic for outbound attempts, domain-tuned speech and intent recognition, and deterministic escalation paths to human agents, rather than optimizing intent-classifier accuracy alone. The author frames the near-term future of customer engagement as neither fully human nor fully automated, but collaborative.
What to watch
This is a single-author opinion piece from an industry practitioner, not an independent study or benchmark; the specific "4.7 attempts" figure and the comparison to offshore call centers are self-reported claims attributed to the author's own platform data, not independently verified. Watch for published case studies or vendor-neutral data that separate the operational gains described here, such as retry logic and scheduling, from actual model-quality improvements.
Key Points
- 1A Forbes Technology Council op-ed argues voice-AI lead conversion often fails due to call persistence and handoff gaps, not weak language understanding.
- 2The author cites platform data showing it takes an average of 4.7 attempts before a real estate or mortgage lead answers the phone.
- 3The claims are self-reported, single-author opinion content; no independent data corroborates the specific attempt-count figures cited.
Scoring Rationale
A single-author, self-published opinion piece with a useful operational framing but no independent data, benchmark, or named company behind the 'millions of calls' claim. Held in the minor-to-solid range because it offers genuine practitioner value but is not a reported news event or verified study.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
